HTML/UI text → contrast/label issues + non-color cues + alt-text.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: A11y Issues, Non-Color Cues, Alt Text. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (3):

• A11y Issues (qwen2.5-coder:1.5b) — Review the HTML/UI for accessibility issues: missing labels, alt, roles, contrast risks. Bullets.
• Non-Color Cues (llama3.2:1b) — Suggest non-color ways to convey meaning (icons, text, patterns) for any color-only cues. Bullets.
• Alt Text (qwen2.5:1.5b) — Write concise alt text for any images/icons mentioned. Bullets.

Idea → endpoints → example request/response payloads.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Routes, Payloads. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Routes (qwen2.5-coder:3b) — Design a minimal REST API for the request: list endpoints (method + path + purpose).
• Payloads (qwen2.5-coder:1.5b) — For these endpoints, give one example JSON request and response each.

Error/log → root-cause → fix → test idea.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnosis, Fix, Test. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnosis (qwen2.5-coder:3b) — From the error/log, state the most likely root cause in 1-2 sentences.
• Fix (qwen2.5-coder:3b) — Propose a concrete fix (code or steps) for this root cause.
• Test (qwen2.5-coder:1.5b) — Suggest one test that would catch this bug. Be specific.

Code → plain English + complexity + risks.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Plain English, Complexity, Risks. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Plain English (qwen2.5-coder:1.5b) — Explain what this code does in plain English, step by step. Brief.
• Complexity (llama3.2:1b) — State the time/space complexity and any performance concerns. Brief.
• Risks (qwen2.5-coder:1.5b) — List correctness/security risks in this code. Bullets.

Two coders write the same thing; a judge picks the cleaner.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (qwen2.5-coder:3b) — Write clean, correct code for the request. Code first, minimal prose.
• Coder B (starcoder2:3b) — Write clean, correct code for the request. Code only.

3 specialists audit code: bugs, docs, security.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Documenter, Security. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (qwen2.5-coder:1.5b) — List ONLY syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Documenter (llama3.2:1b) — Output a concise Markdown README: what it does + how to use. No code changes.
• Security (qwen2.5-coder:1.5b) — Report ONLY security issues: missing validation, injection, plaintext secrets. Bullets.

Port code between languages (PHP↔Pascal↔Bash): analyze → emit → lint.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Analyzer, Porter, Linter. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Analyzer (qwen2.5-coder:3b) — Extract the logic of the given code as terse pseudo-steps. State source + intended target language if named.
• Porter (qwen2.5-coder:3b) — Using the pseudo-steps, write the equivalent code in the requested target language. Code only.
• Linter (qwen2.5-coder:1.5b) — Check the code for obvious syntax/style problems. Output the corrected final code only.

Diff/notes → summary → Conventional Commit + body.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Summary, Commit. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Summary (qwen2.5-coder:1.5b) — Summarize what this change does in one sentence.
• Commit (qwen2.5-coder:1.5b) — Write a Conventional Commits message: a <=50-char subject line, then a short body. Output the message only.

Schedule in words → cron + systemd timer + explanation.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Cron, systemd, Explain. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Cron (qwen2.5-coder:1.5b) — Output ONLY the crontab expression for the described schedule, plus the command line.
• systemd (qwen2.5-coder:1.5b) — Output a systemd OnCalendar= line equivalent to the described schedule.
• Explain (qwen3:0.6b) — Restate when this schedule fires, in plain words.

Librarian condenses RAG → coder writes the code.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Librarian, Coder. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Librarian (qwen2.5:3b) — From the context, extract ONLY the syntax rules, names, and signatures relevant to the task. Terse bullets. Do NOT write code.
• Coder (qwen2.5-coder:3b) — Write the code using only the rules below. Code only.

App description → Dockerfile → compose → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile, Compose. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile (qwen2.5-coder:3b) — Write a Dockerfile for the described app. File only.
• Compose (qwen2.5-coder:1.5b) — Write a minimal docker-compose.yml for it, plus the run command.

Need → regex + sed/awk + bash one-liners side by side.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Regex, sed/awk, Bash. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Regex (qwen2.5-coder:1.5b) — Output ONLY a regex that satisfies the request, plus a one-line explanation.
• sed/awk (qwen2.5-coder:1.5b) — Output ONLY a sed or awk one-liner for the request, plus one line of explanation.
• Bash (qwen2.5-coder:3b) — Output ONLY a bash one-liner for the request, plus one line of explanation.

Refactor for clarity → add comments → sanity check.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Commenter, Checker. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Refactorer (qwen2.5-coder:3b) — Refactor the code for clarity and structure without changing behavior. Code only.
• Commenter (qwen2.5-coder:1.5b) — Add concise, useful comments/docstrings to the code. Output the commented code.
• Checker (qwen2.5-coder:1.5b) — List any remaining issues in 1-3 bullets, then output the final code.

Plain English → SQL for MySQL + SQLite + explanation.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: MySQL, SQLite, Explain. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• MySQL (qwen2.5-coder:3b) — Write the SQL query for MySQL. Query only.
• SQLite (qwen2.5-coder:1.5b) — Write the SQL query for SQLite. Query only.
• Explain (llama3.2:1b) — Explain in 2-3 bullets what the requested query does.

Code → unit tests + edge cases + failure modes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Unit Tests, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Unit Tests (qwen2.5-coder:3b) — Write unit tests for the given code. Tests only.
• Edge Cases (qwen2.5-coder:1.5b) — List tricky edge cases this code should handle. Bullets.
• Failure Modes (llama3.2:1b) — List ways this code could fail at runtime. Bullets.

One prompt → environment + factions + tech + NPCs.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Architect, Sociologist, Quartermaster, NPCs. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (4):

• Architect (llama3.2:1b) — Describe the physical environment/infrastructure of the setting. 1 paragraph.
• Sociologist (tinyllama:1.1b) — Describe factions, culture, conflicts. 1 paragraph.
• Quartermaster (smollm2:360m) — List available technology, weapons, vehicles. Bullets.
• NPCs (llama3.2:1b) — Create 3 short character profiles that fit the setting.

Premise → outline → next chapter → cliffhanger.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outline, Chapter, Cliffhanger. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Outline (qwen2.5:3b) — Outline the next chapter for the premise. Bullets.
• Chapter (llama3.2:3b) — Write the next chapter from the outline.
• Cliffhanger (llama3.2:1b) — Write a one-paragraph cliffhanger ending.

Draft → edit for consistency → wrap as clean HTML.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Writer, Editor, Formatter. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Writer (qwen2.5:3b) — Write a rich draft for the request. Avoid relying on color names for meaning.
• Editor (llama3.2:3b) — Edit the draft for consistency and flow. Output edited text.
• Formatter (qwen2.5-coder:3b) — Wrap the text in clean semantic HTML (headings, paragraphs). HTML only, no commentary.

Messy tabular text → schema + cleaned rows + anomalies.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Schema, Cleaned, Anomalies. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Schema (qwen2.5-coder:1.5b) — Infer the columns and types from the data. JSON.
• Cleaned (qwen2.5-coder:3b) — Output the data as clean, normalized CSV (consistent delimiters, trimmed).
• Anomalies (qwen3:0.6b) — List rows/values that look wrong or inconsistent. Bullets.

Messy text → tech tokens + human text + timestamps.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Tech Parser, Linguist, Timekeeper. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Tech Parser (qwen2.5-coder:1.5b) — Extract ONLY IPs, file paths, URLs, error codes. JSON array.
• Linguist (llama3.2:1b) — Extract ONLY the human-readable messages/complaints. Bullets.
• Timekeeper (qwen3:0.6b) — Extract ONLY timestamps/dates, chronological. JSON array.

Vet claims: skeptic + timeline + related concepts.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Skeptic, Timeline, Concepts. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Skeptic (llama3.2:1b) — List claims that sound dubious or lack evidence. Bullets.
• Timeline (qwen3:0.6b) — Extract every date/event, chronological. JSON.
• Concepts (smollm2:1.7b) — Suggest 5 related topics or works based on the themes.

Article → summary + claims + counterpoints + to-verify.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summary, Claims, Counterpoints, To Verify. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (4):

• Summary (llama3.2:3b) — Summarize the article in 4 bullets.
• Claims (qwen2.5:1.5b) — List the main claims the article makes. Bullets.
• Counterpoints (llama3.2:1b) — List plausible counterpoints or limitations. Bullets.
• To Verify (qwen3:0.6b) — List specific facts worth verifying. Bullets.

Idea → 3 style prompts: photoreal + illustration + cinematic.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Photoreal, Illustration, Cinematic. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Photoreal (llama3.2:3b) — Write a photorealistic image prompt for the idea: camera, lens, lighting, detail tags. One line.
• Illustration (qwen2.5:1.5b) — Write an illustration/art-style image prompt for the idea: medium, style, palette tags. One line.
• Cinematic (llama3.2:1b) — Write a cinematic image prompt for the idea: mood, composition, film tags. One line.

Brand → 3 distinct logo prompt directions.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Minimal Mark, Emblem, Wordmark. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Minimal Mark (llama3.2:1b) — Describe a minimalist logo concept + an image prompt for the brand. One line prompt.
• Emblem (qwen2.5:1.5b) — Describe an emblem/badge logo concept + an image prompt. One line prompt.
• Wordmark (llama3.2:1b) — Describe a typographic wordmark logo concept + an image prompt. One line prompt.

Idea → detailed image-generation prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Expander. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (1):

• Prompt Expander (llama3.2:3b) — Expand the idea into a detailed, comma-separated image prompt: subject, lighting, lens, style tags. One line.

Idea → detailed prompt → negative prompt → suggested params.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt, Negative, Params. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Prompt (llama3.2:3b) — Write a detailed image prompt for the idea. One line.
• Negative (qwen2.5:1.5b) — Write a negative prompt (things to avoid) for this image. One line.
• Params (qwen3:0.6b) — Suggest CFG scale, steps, and size for this image. JSON.

Concept → scenes → shot list → image prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scenes, Shot List, Prompts. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Scenes (llama3.2:3b) — Break the concept into 4-6 scenes. One line each.
• Shot List (qwen2.5:1.5b) — For each scene, give a shot (angle + action).
• Prompts (llama3.2:1b) — For each shot, write a one-line image prompt.

Announcement → Spanish + Hebrew + hashtags.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Spanish, Hebrew, Hashtags. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Spanish (qwen2.5:1.5b) — Localize to engaging Castilian Spanish. Text only.
• Hebrew (llama3.2:1b) — Localize to natural Hebrew (RTL). Text only.
• Hashtags (tinyllama:1.1b) — Generate relevant social hashtags.

Text → grammar fix → clarity → tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar, Tighten. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar (qwen2.5:3b) — Fix grammar and spelling. Output corrected text only.
• Tighten (llama3.2:3b) — Improve clarity and remove redundancy. Output the final text.

UI text → Hebrew + Spanish + tooltips.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Hebrew, Spanish, Tooltips. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Hebrew (llama3.2:1b) — Translate to natural Hebrew (RTL). Translation only.
• Spanish (qwen2.5:1.5b) — Translate to natural Castilian Spanish. Translation only.
• Tooltips (qwen3:0.6b) — Write short English UI tooltips for the text. Bullets.

One text → Hebrew + Spanish + Arabic + French at once.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Hebrew, Spanish, Arabic, French. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (4):

• Hebrew (llama3.2:1b) — Translate to Hebrew (RTL). Translation only.
• Spanish (qwen2.5:1.5b) — Translate to Spanish. Translation only.
• Arabic (llama3.2:3b) — Translate to Arabic (RTL). Translation only.
• French (qwen2.5:1.5b) — Translate to French. Translation only.

Translate to JSON, then validate the JSON.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Validator. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator (llama3.2:3b) — Translate the text to Hebrew and Spanish. Output JSON {"he":"...","es":"..."} only.
• Validator (qwen2.5-coder:1.5b) — Validate and fix the JSON so it is well-formed UTF-8. Output ONLY the corrected JSON.

Transcript lines → translate → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translate, Tidy. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translate (llama3.2:3b) — Translate each subtitle line to the requested language, keeping line breaks and any timecodes intact.
• Tidy (qwen2.5:1.5b) — Ensure line lengths are subtitle-friendly (short lines). Output the final lines.

Text → formal + casual + concise versions.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal, Casual, Concise. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Formal (qwen2.5:1.5b) — Rewrite the text in a formal, professional tone.
• Casual (llama3.2:1b) — Rewrite the text in a friendly, casual tone.
• Concise (qwen2.5:1.5b) — Rewrite the text as concisely as possible without losing meaning.

Translate → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translate, Back-translate, Final. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translate (llama3.2:3b) — Translate the text to the requested target language. Translation only.
• Back-translate (qwen2.5:3b) — Translate the following back to English literally, to check fidelity.
• Final (llama3.2:3b) — Given the original request, the translation, and its back-translation, output the corrected final translation only.

3 agents propose; a judge synthesizes the best.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Creative, Logical, Minimalist. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• Creative (llama3.2:1b) — Give a creative, unorthodox solution. Brief.
• Logical (qwen2.5:1.5b) — Give the most logically sound solution. Brief.
• Minimalist (qwen3:0.6b) — Give the simplest solution. Brief.

Concept → ELI5 + intermediate + expert at once.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5, Intermediate, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 (llama3.2:1b) — Explain the concept simply, as if to a curious child. Short.
• Intermediate (qwen2.5:1.5b) — Explain the concept for a knowledgeable beginner. Short.
• Expert (qwen2.5:3b) — Explain the concept precisely for an expert, with the key nuance.

Text → key facts → Q/A flashcards (JSON).

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Facts, Cards. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Facts (llama3.2:1b) — Extract the key facts from the text. Bullets.
• Cards (qwen2.5:1.5b) — Turn the facts into Q/A flashcards. Output a JSON array of {"q":..,"a":..}.

Idea → 3 brainstorms → judge shortlists.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Playful, Professional, Quirky. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• Playful (llama3.2:1b) — Brainstorm 8 playful, catchy names for the idea.
• Professional (qwen2.5:1.5b) — Brainstorm 8 clean, professional names for the idea.
• Quirky (tinyllama:1.1b) — Brainstorm 8 unusual, memorable names for the idea.

Dilemma → pros + cons + risks → verdict.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Pros, Cons, Risks. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• Pros (llama3.2:1b) — List the strongest pros/arguments for the option. Bullets.
• Cons (qwen2.5:1.5b) — List the strongest cons/arguments against. Bullets.
• Risks (qwen3:0.6b) — List the key risks/unknowns. Bullets.

Commits/notes → changelog → release post.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Changelog, Release Post. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Changelog (qwen2.5-coder:1.5b) — Turn the notes into a grouped changelog (Added/Fixed/Changed). Markdown.
• Release Post (llama3.2:3b) — Write a short, friendly release announcement from this changelog.

Long doc → TL;DR + outline + key terms + questions it answers.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR, Outline, Key Terms, Answers. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (4):

• TL;DR (llama3.2:3b) — Give a 2-3 sentence TL;DR of the document.
• Outline (llama3.2:1b) — Give a bulleted outline of the document's structure.
• Key Terms (qwen2.5:1.5b) — List the key terms/definitions in the document.
• Answers (qwen3:0.6b) — List 5 questions this document answers.

Messy text → 3-bullet summary + action items + tone.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summary, Actions, Sentiment. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summary (llama3.2:1b) — Summarize the text in exactly 3 bullet points.
• Actions (qwen3:0.6b) — Extract action items / tasks / dates as a JSON checklist.
• Sentiment (smollm2:360m) — State the tone in 1-3 words (e.g. urgent, formal).

Intent → draft → tone polish → subject + TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Draft, Polish, Subject + TL;DR. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Draft (llama3.2:3b) — Write a clear email for the user's intent.
• Polish (qwen2.5:1.5b) — Polish the email for tone and concision. Output the email.
• Subject + TL;DR (qwen3:0.6b) — Give a subject line and a one-line TL;DR for this email.

Product/topic → FAQ + objections + support replies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: FAQ, Objections, Support Replies. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• FAQ (llama3.2:3b) — Write 5 FAQ question/answer pairs for the topic.
• Objections (qwen2.5:1.5b) — List likely objections/concerns and a short rebuttal each.
• Support Replies (llama3.2:1b) — Write 3 friendly canned support replies for common issues.

Messy text → line items + totals + dates.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Line Items, Totals, Dates/Parties. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Line Items (qwen2.5-coder:1.5b) — Extract line items (description, qty, price) as a JSON array.
• Totals (qwen3:0.6b) — Extract subtotal, tax, and total amounts. JSON.
• Dates/Parties (qwen3:0.6b) — Extract dates, invoice number, and parties. JSON.

Transcript → summary + decisions + action items + attendees.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summary, Decisions, Action Items, Attendees. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (4):

• Summary (llama3.2:3b) — Summarize the meeting in 3-5 bullets.
• Decisions (llama3.2:1b) — List decisions made. Bullets.
• Action Items (qwen2.5:1.5b) — List action items with owner + due if stated. JSON.
• Attendees (qwen3:0.6b) — List the people/speakers mentioned.

Idea → outline → draft → exec summary.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outline, Draft, Exec Summary. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Outline (qwen2.5:3b) — Create a proposal outline for the idea.
• Draft (llama3.2:3b) — Expand the outline into a short proposal draft.
• Exec Summary (qwen2.5:1.5b) — Write a 3-sentence executive summary of this proposal.

Rough bio → polish → ATS keywords.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polish, Keywords. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polish (llama3.2:3b) — Rewrite the bio/resume text to be crisp and professional.
• Keywords (qwen2.5:1.5b) — List relevant ATS keywords/skills implied by this text.

Describe need → Excel + Google Sheets formula + explanation.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Excel, Google Sheets, Explain. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Excel (qwen2.5-coder:1.5b) — Output ONLY the Excel formula for the request.
• Google Sheets (qwen2.5-coder:1.5b) — Output ONLY the Google Sheets formula for the request.
• Explain (llama3.2:1b) — Explain in 1-2 lines what the formula computes.

HTML/UI → keyboard navigation issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (gemma-3-1b-it) — Scan this HTML/UI text ONLY for keyboard navigation issues. List concrete problems as bullets, cite the element.
• Fix Suggester (gemma-3-4b-it) — For each keyboard navigation issue found, suggest a concrete fix (markup/attribute/CSS change).

Generate Bash unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (gemma-3-1b-it) — Write Bash unit tests for the normal/expected inputs only.
• Edge Cases (gemma-3-4b-it) — Write Bash unit tests for boundary/edge-case inputs only.
• Failure Modes (gemma-4-e2b-it-qat-mtp) — Write Bash unit tests for invalid input / error conditions only.

Two C# coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (gemma-3-1b-it) — Solve the given task in C#, straightforward/imperative style. Code only.
• Coder B (gemma-3-4b-it) — Solve the given task in C#, a more functional/idiomatic style. Code only.

Refactor C++ code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (qwen_qwen3.5-2b) — Rewrite the given C++ code for clarity and idiom. Output only the revised code.
• Verifier (gemma-3-1b-it) — Sanity-check the refactored C++ code against the original — same behavior? List any risk in 3 bullets max.

Generate Go unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (gemma-4-e2b-it-qat-mtp) — Write Go unit tests for the normal/expected inputs only.
• Edge Cases (lfm2.5-1.2b-instruct) — Write Go unit tests for boundary/edge-case inputs only.
• Failure Modes (qwen_qwen3.5-0.8b) — Write Go unit tests for invalid input / error conditions only.

Error/log in Java → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (qwen_qwen3.5-0.8b) — Given a Java error/log, state the most likely root cause in 2-3 sentences.
• Fixer (qwen_qwen3.5-2b) — Propose a concrete Java code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (gemma-3-1b-it) — Write one Java regression test that would have caught this bug.

Refactor JavaScript code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (gemma-3-4b-it) — Rewrite the given JavaScript code for clarity and idiom. Output only the revised code.
• Verifier (gemma-4-e2b-it-qat-mtp) — Sanity-check the refactored JavaScript code against the original — same behavior? List any risk in 3 bullets max.

Kotlin code → plain-English explanation + complexity + risk notes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Explainer, Complexity, Risk Spotter. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Explainer (lfm2.5-1.2b-instruct) — Explain what this Kotlin code does in plain English, no jargon.
• Complexity (qwen_qwen3.5-0.8b) — State the time/space complexity of this Kotlin code in one line, with reasoning.
• Risk Spotter (qwen_qwen3.5-2b) — List any risk (security, correctness, maintainability) in this Kotlin code. Bullets.

Idea → REST endpoints in PHP conventions → example payloads.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: API Designer, Example Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• API Designer (qwen_qwen3.5-2b) — Given a feature idea, propose REST endpoints following PHP-ecosystem conventions.
• Example Writer (gemma-3-1b-it) — Write example request/response JSON payloads for the proposed endpoints.

Three specialists audit Python code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (gemma-3-1b-it) — List ONLY Python syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker (gemma-3-4b-it) — Point out non-idiomatic Python — what a senior Python dev would flag in review. Bullets only.
• Performance (gemma-4-e2b-it-qat-mtp) — Report ONLY performance concerns for this Python code — hot loops, allocations, complexity. Bullets.

App description (Ruby) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author (gemma-4-e2b-it-qat-mtp) — Write a production-sane Dockerfile for this Ruby app description.
• Notes Writer (lfm2.5-1.2b-instruct) — Write 3-5 short run/build notes for the Dockerfile above.

Turn Rust code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer (lfm2.5-1.2b-instruct) — Write a concise Markdown README for this Rust code: what it does, how to use it.
• Editor (qwen_qwen3.5-0.8b) — Tighten the README — cut fluff, keep only what a new user actually needs.

Turn SQL code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer (gemma-3-4b-it) — Write a concise Markdown README for this SQL code: what it does, how to use it.
• Editor (gemma-4-e2b-it-qat-mtp) — Tighten the README — cut fluff, keep only what a new user actually needs.

Three specialists audit Swift code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (qwen_qwen3.5-0.8b) — List ONLY Swift syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker (qwen_qwen3.5-2b) — Point out non-idiomatic Swift — what a senior Swift dev would flag in review. Bullets only.
• Performance (gemma-3-1b-it) — Report ONLY performance concerns for this Swift code — hot loops, allocations, complexity. Bullets.

TypeScript diff/notes → summary → Conventional Commit message.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Summarizer, Commit Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Summarizer (gemma-3-4b-it) — Summarize what this TypeScript diff/change actually does, in plain English.
• Commit Writer (gemma-4-e2b-it-qat-mtp) — Write a Conventional Commits message (type, scope, subject, body) for this change.

One prompt → hard sci-fi setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect (gemma-3-1b-it) — Describe a hard sci-fi setting for this prompt — geography, mood, era.
• Faction Writer (gemma-3-4b-it) — Invent 2-3 factions/groups that fit this hard sci-fi setting.
• NPC Writer (gemma-4-e2b-it-qat-mtp) — Invent 2 memorable NPCs that fit this hard sci-fi setting.

Premise → high fantasy outline → next chapter → cliffhanger.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outliner, Writer, Hook Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Outliner (gemma-3-4b-it) — Outline the next 3 beats of this high fantasy story premise.
• Writer (gemma-4-e2b-it-qat-mtp) — Write the next chapter/scene in high fantasy style, following the outline.
• Hook Writer (lfm2.5-1.2b-instruct) — Write a one-paragraph cliffhanger ending for this high fantasy chapter.

One prompt → noir detective setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect (gemma-4-e2b-it-qat-mtp) — Describe a noir detective setting for this prompt — geography, mood, era.
• Faction Writer (lfm2.5-1.2b-instruct) — Invent 2-3 factions/groups that fit this noir detective setting.
• NPC Writer (qwen_qwen3.5-0.8b) — Invent 2 memorable NPCs that fit this noir detective setting.

customer support transcripts → summary + notable claims + what to verify, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Claim Spotter, Verifier. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Summarizer (gemma-4-e2b-it-qat-mtp) — Summarize the key findings visible in this customer support transcripts sample.
• Claim Spotter (lfm2.5-1.2b-instruct) — List any strong/surprising claims implied by this customer support transcripts sample.
• Verifier (qwen_qwen3.5-0.8b) — List what in this customer support transcripts sample would need independent verification before acting on it.

Messy e-commerce order exports → schema guess + cleaned rows + anomalies, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Schema Guesser, Cleaner, Anomaly Spotter. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Schema Guesser (gemma-3-4b-it) — Guess a reasonable schema (column names + types) for this e-commerce order exports sample.
• Cleaner (gemma-4-e2b-it-qat-mtp) — Reformat this e-commerce order exports sample into clean, consistent rows.
• Anomaly Spotter (lfm2.5-1.2b-instruct) — Flag rows/values in this e-commerce order exports sample that look wrong or inconsistent.

Messy sensor telemetry readings → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor (lfm2.5-1.2b-instruct) — Pull structured fields out of this sensor telemetry readings sample, as a table.
• Summarizer (qwen_qwen3.5-0.8b) — Summarize what's happening in this sensor telemetry readings sample, in plain English.
• Timestamp Finder (qwen_qwen3.5-2b) — List every timestamp/date found in this sensor telemetry readings sample.

Messy server access logs → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor (gemma-3-1b-it) — Pull structured fields out of this server access logs sample, as a table.
• Summarizer (gemma-3-4b-it) — Summarize what's happening in this server access logs sample, in plain English.
• Timestamp Finder (gemma-4-e2b-it-qat-mtp) — List every timestamp/date found in this server access logs sample.

Concept → scenes → cyberpunk neon-style shot prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scene Breakdown, Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Scene Breakdown (gemma-4-e2b-it-qat-mtp) — Break this concept into 4-6 numbered scenes.
• Prompt Writer (lfm2.5-1.2b-instruct) — Write a cyberpunk neon-style image prompt for each scene above.

Idea → detailed flat vector illustration image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer (lfm2.5-1.2b-instruct) — Turn this idea into a detailed flat vector illustration-style image generation prompt.
• Negative Prompt Writer (qwen_qwen3.5-0.8b) — Write a negative prompt to pair with the flat vector illustration prompt above — what to avoid.

Idea → 3 distinct oil painting prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A (qwen_qwen3.5-0.8b) — Write a oil painting-style image prompt for this idea, direction A (bold/dramatic).
• Concept B (qwen_qwen3.5-2b) — Write a oil painting-style image prompt for this idea, direction B (subtle/minimal).
• Concept C (gemma-3-1b-it) — Write a oil painting-style image prompt for this idea, direction C (unexpected angle).

Idea → detailed photorealistic image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer (gemma-3-1b-it) — Turn this idea into a detailed photorealistic-style image generation prompt.
• Negative Prompt Writer (gemma-3-4b-it) — Write a negative prompt to pair with the photorealistic prompt above — what to avoid.

Idea → 3 distinct watercolor painting prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A (gemma-3-4b-it) — Write a watercolor painting-style image prompt for this idea, direction A (bold/dramatic).
• Concept B (gemma-4-e2b-it-qat-mtp) — Write a watercolor painting-style image prompt for this idea, direction B (subtle/minimal).
• Concept C (lfm2.5-1.2b-instruct) — Write a watercolor painting-style image prompt for this idea, direction C (unexpected angle).

Text in Arabic → grammar fix → clarity tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar Fixer, Clarity Editor. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar Fixer (qwen_qwen3.5-0.8b) — Fix grammar/spelling errors in this Arabic text. Output corrected text only.
• Clarity Editor (qwen_qwen3.5-2b) — Tighten this Arabic text for clarity — cut redundancy, keep meaning.

UI text → French + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer (gemma-3-4b-it) — Translate this UI text into French, concise and natural for a UI label.
• Tooltip Writer (gemma-4-e2b-it-qat-mtp) — Write a short French tooltip/help text version of this UI text.

Text → formal + casual German versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice (gemma-4-e2b-it-qat-mtp) — Translate/rewrite this text into formal, professional German.
• Casual Voice (lfm2.5-1.2b-instruct) — Translate/rewrite this text into casual, everyday German.

Transcript lines → Hebrew translation → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Format Checker. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator (lfm2.5-1.2b-instruct) — Translate these subtitle lines into Hebrew, keeping line breaks intact.
• Format Checker (qwen_qwen3.5-0.8b) — Verify the Hebrew lines still match the original line count/timing markers.

UI text → Italian + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer (gemma-3-1b-it) — Translate this UI text into Italian, concise and natural for a UI label.
• Tooltip Writer (gemma-3-4b-it) — Write a short Italian tooltip/help text version of this UI text.

Translate to Japanese → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator (qwen_qwen3.5-2b) — Translate the given text into Japanese. Natural, not literal.
• Back-Translator (gemma-3-1b-it) — Translate the Japanese translation back into English.
• Drift Fixer (gemma-3-4b-it) — Compare original vs back-translation; fix the Japanese version if meaning drifted.

Text → formal + casual Portuguese versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice (gemma-3-4b-it) — Translate/rewrite this text into formal, professional Portuguese.
• Casual Voice (gemma-4-e2b-it-qat-mtp) — Translate/rewrite this text into casual, everyday Portuguese.

Translate to Spanish → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator (gemma-3-1b-it) — Translate the given text into Spanish. Natural, not literal.
• Back-Translator (gemma-3-4b-it) — Translate the Spanish translation back into English.
• Drift Fixer (gemma-4-e2b-it-qat-mtp) — Compare original vs back-translation; fix the Spanish version if meaning drifted.

Explain blockchain consensus at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher (lfm2.5-1.2b-instruct) — Explain blockchain consensus like the reader is 10 years old. No jargon.
• Intermediate Teacher (qwen_qwen3.5-0.8b) — Explain blockchain consensus for someone with basic background — some technical terms OK.
• Expert (qwen_qwen3.5-2b) — Explain blockchain consensus at expert depth, precise terminology, no hand-holding.

3 agents argue about compound interest; a judge synthesizes the strongest take.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Advocate A, Advocate B. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Advocate A (gemma-4-e2b-it-qat-mtp) — Give the strongest argument/explanation of compound interest, angle A.
• Advocate B (lfm2.5-1.2b-instruct) — Give the strongest argument/explanation of compound interest, angle B, different framing.

Explain photosynthesis at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher (gemma-3-1b-it) — Explain photosynthesis like the reader is 10 years old. No jargon.
• Intermediate Teacher (gemma-3-4b-it) — Explain photosynthesis for someone with basic background — some technical terms OK.
• Expert (gemma-4-e2b-it-qat-mtp) — Explain photosynthesis at expert depth, precise terminology, no hand-holding.

Text about quantum entanglement → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor (qwen_qwen3.5-0.8b) — Extract the key facts about quantum entanglement from the given text, as bullets.
• Flashcard Writer (qwen_qwen3.5-2b) — Turn those facts into Q/A flashcards about quantum entanglement (JSON array of {q,a}).

Text about recursion → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor (gemma-3-4b-it) — Extract the key facts about recursion from the given text, as bullets.
• Flashcard Writer (gemma-4-e2b-it-qat-mtp) — Turn those facts into Q/A flashcards about recursion (JSON array of {q,a}).

Messy a client proposal text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (gemma-3-1b-it) — Summarize this a client proposal in exactly 3 bullets.
• Action Extractor (gemma-3-4b-it) — List only concrete action items found in this a client proposal, as bullets.
• Tone Reader (gemma-4-e2b-it-qat-mtp) — State the tone of this a client proposal in one sentence (e.g. urgent, neutral, frustrated).

Messy an internal memo text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (lfm2.5-1.2b-instruct) — Extract structured line items from this an internal memo as a table.
• Date Finder (qwen_qwen3.5-0.8b) — List every date/deadline mentioned in this an internal memo.
• Anomaly Spotter (qwen_qwen3.5-2b) — Flag anything in this an internal memo that looks inconsistent or wrong.

Messy a job posting text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (qwen_qwen3.5-2b) — Summarize this a job posting in exactly 3 bullets.
• Action Extractor (gemma-3-1b-it) — List only concrete action items found in this a job posting, as bullets.
• Tone Reader (gemma-3-4b-it) — State the tone of this a job posting in one sentence (e.g. urgent, neutral, frustrated).

Rough a partnership agreement summary → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (lfm2.5-1.2b-instruct) — Polish this rough a partnership agreement summary draft for grammar and flow.
• Keyword Pass (qwen_qwen3.5-0.8b) — Add/verify relevant keywords a reader or search would expect in this a partnership agreement summary.

Intent → draft a performance review → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (gemma-3-1b-it) — Write a first draft of a performance review given the stated intent.
• Polisher (gemma-3-4b-it) — Polish the tone of this a performance review draft — professional, clear, no filler.
• Summarizer (gemma-4-e2b-it-qat-mtp) — Write a one-line subject and a 2-sentence TL;DR for this a performance review.

Long a policy document → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (gemma-3-4b-it) — Write a 2-sentence TL;DR of this a policy document.
• Outliner (gemma-4-e2b-it-qat-mtp) — Extract a section-by-section outline of this a policy document.
• Term Extractor (lfm2.5-1.2b-instruct) — List key terms/jargon used in this a policy document, with one-line definitions.

Rough a press release → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (qwen_qwen3.5-0.8b) — Polish this rough a press release draft for grammar and flow.
• Keyword Pass (qwen_qwen3.5-2b) — Add/verify relevant keywords a reader or search would expect in this a press release.

Intent → draft a status report → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (gemma-3-4b-it) — Write a first draft of a status report given the stated intent.
• Polisher (gemma-4-e2b-it-qat-mtp) — Polish the tone of this a status report draft — professional, clear, no filler.
• Summarizer (lfm2.5-1.2b-instruct) — Write a one-line subject and a 2-sentence TL;DR for this a status report.

Long a support ticket reply → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (gemma-4-e2b-it-qat-mtp) — Write a 2-sentence TL;DR of this a support ticket reply.
• Outliner (lfm2.5-1.2b-instruct) — Extract a section-by-section outline of this a support ticket reply.
• Term Extractor (qwen_qwen3.5-0.8b) — List key terms/jargon used in this a support ticket reply, with one-line definitions.

Messy a user manual section text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (gemma-4-e2b-it-qat-mtp) — Extract structured line items from this a user manual section as a table.
• Date Finder (lfm2.5-1.2b-instruct) — List every date/deadline mentioned in this a user manual section.
• Anomaly Spotter (qwen_qwen3.5-0.8b) — Flag anything in this a user manual section that looks inconsistent or wrong.

HTML/UI → keyboard navigation issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner O (qwen2.5-coder:1.5b) — Scan this HTML/UI text ONLY for keyboard navigation issues. List concrete problems as bullets, cite the element.
• Fix Suggester L (gemma-3-1b-it) — For each keyboard navigation issue found, suggest a concrete fix (markup/attribute/CSS change).

Generate Bash unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path O (llama3.2:3b) — Write Bash unit tests for the normal/expected inputs only.
• Edge Cases L (gemma-3-4b-it) — Write Bash unit tests for boundary/edge-case inputs only.
• Failure Modes O (qwen2.5:1.5b) — Write Bash unit tests for invalid input / error conditions only.

Two C# coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A O (qwen2.5:3b) — Solve the given task in C#, straightforward/imperative style. Code only.
• Coder B L (gemma-3-4b-it) — Solve the given task in C#, a more functional/idiomatic style. Code only.

Refactor C++ code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer O (llama3.2:1b) — Rewrite the given C++ code for clarity and idiom. Output only the revised code.
• Verifier L (gemma-3-1b-it) — Sanity-check the refactored C++ code against the original — same behavior? List any risk in 3 bullets max.

Generate Go unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path O (llama3.2:1b) — Write Go unit tests for the normal/expected inputs only.
• Edge Cases L (lfm2.5-1.2b-instruct) — Write Go unit tests for boundary/edge-case inputs only.
• Failure Modes O (gemma2:2b) — Write Go unit tests for invalid input / error conditions only.

Error/log in Java → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser O (gemma2:2b) — Given a Java error/log, state the most likely root cause in 2-3 sentences.
• Fixer L (qwen_qwen3.5-2b) — Propose a concrete Java code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer O (qwen2.5:3b) — Write one Java regression test that would have caught this bug.

Refactor JavaScript code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer O (qwen2.5-coder:3b) — Rewrite the given JavaScript code for clarity and idiom. Output only the revised code.
• Verifier L (gemma-4-e2b-it-qat-mtp) — Sanity-check the refactored JavaScript code against the original — same behavior? List any risk in 3 bullets max.

Kotlin code → plain-English explanation + complexity + risk notes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Explainer, Complexity, Risk Spotter. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Explainer O (qwen2.5-coder:1.5b) — Explain what this Kotlin code does in plain English, no jargon.
• Complexity L (qwen_qwen3.5-0.8b) — State the time/space complexity of this Kotlin code in one line, with reasoning.
• Risk Spotter O (llama3.2:1b) — List any risk (security, correctness, maintainability) in this Kotlin code. Bullets.

Idea → REST endpoints in PHP conventions → example payloads.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: API Designer, Example Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• API Designer O (qwen2.5:1.5b) — Given a feature idea, propose REST endpoints following PHP-ecosystem conventions.
• Example Writer L (gemma-3-1b-it) — Write example request/response JSON payloads for the proposed endpoints.

Three specialists audit Python code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter O (qwen2.5-coder:1.5b) — List ONLY Python syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker L (gemma-3-4b-it) — Point out non-idiomatic Python — what a senior Python dev would flag in review. Bullets only.
• Performance O (llama3.2:1b) — Report ONLY performance concerns for this Python code — hot loops, allocations, complexity. Bullets.

App description (Ruby) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author O (smollm2:1.7b) — Write a production-sane Dockerfile for this Ruby app description.
• Notes Writer L (lfm2.5-1.2b-instruct) — Write 3-5 short run/build notes for the Dockerfile above.

Turn Rust code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer O (llama3.2:3b) — Write a concise Markdown README for this Rust code: what it does, how to use it.
• Editor L (qwen_qwen3.5-0.8b) — Tighten the README — cut fluff, keep only what a new user actually needs.

Turn SQL code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer O (gemma2:2b) — Write a concise Markdown README for this SQL code: what it does, how to use it.
• Editor L (gemma-4-e2b-it-qat-mtp) — Tighten the README — cut fluff, keep only what a new user actually needs.

Three specialists audit Swift code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter O (qwen2.5-coder:3b) — List ONLY Swift syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker L (qwen_qwen3.5-2b) — Point out non-idiomatic Swift — what a senior Swift dev would flag in review. Bullets only.
• Performance O (llama3.2:3b) — Report ONLY performance concerns for this Swift code — hot loops, allocations, complexity. Bullets.

TypeScript diff/notes → summary → Conventional Commit message.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Summarizer, Commit Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Summarizer O (deepseek-r1:1.5b) — Summarize what this TypeScript diff/change actually does, in plain English.
• Commit Writer L (gemma-4-e2b-it-qat-mtp) — Write a Conventional Commits message (type, scope, subject, body) for this change.

One prompt → hard sci-fi setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect O (qwen2.5-coder:1.5b) — Describe a hard sci-fi setting for this prompt — geography, mood, era.
• Faction Writer L (gemma-3-4b-it) — Invent 2-3 factions/groups that fit this hard sci-fi setting.
• NPC Writer O (llama3.2:1b) — Invent 2 memorable NPCs that fit this hard sci-fi setting.

Premise → high fantasy outline → next chapter → cliffhanger.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outliner, Writer, Hook Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Outliner O (qwen2.5-coder:3b) — Outline the next 3 beats of this high fantasy story premise.
• Writer L (gemma-4-e2b-it-qat-mtp) — Write the next chapter/scene in high fantasy style, following the outline.
• Hook Writer O (llama3.2:3b) — Write a one-paragraph cliffhanger ending for this high fantasy chapter.

One prompt → noir detective setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect O (llama3.2:1b) — Describe a noir detective setting for this prompt — geography, mood, era.
• Faction Writer L (lfm2.5-1.2b-instruct) — Invent 2-3 factions/groups that fit this noir detective setting.
• NPC Writer O (gemma2:2b) — Invent 2 memorable NPCs that fit this noir detective setting.

customer support transcripts → summary + notable claims + what to verify, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Claim Spotter, Verifier. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Summarizer O (llama3.2:1b) — Summarize the key findings visible in this customer support transcripts sample.
• Claim Spotter L (lfm2.5-1.2b-instruct) — List any strong/surprising claims implied by this customer support transcripts sample.
• Verifier O (gemma2:2b) — List what in this customer support transcripts sample would need independent verification before acting on it.

Messy e-commerce order exports → schema guess + cleaned rows + anomalies, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Schema Guesser, Cleaner, Anomaly Spotter. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Schema Guesser O (qwen2.5-coder:3b) — Guess a reasonable schema (column names + types) for this e-commerce order exports sample.
• Cleaner L (gemma-4-e2b-it-qat-mtp) — Reformat this e-commerce order exports sample into clean, consistent rows.
• Anomaly Spotter O (llama3.2:3b) — Flag rows/values in this e-commerce order exports sample that look wrong or inconsistent.

Messy sensor telemetry readings → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor O (llama3.2:3b) — Pull structured fields out of this sensor telemetry readings sample, as a table.
• Summarizer L (qwen_qwen3.5-0.8b) — Summarize what's happening in this sensor telemetry readings sample, in plain English.
• Timestamp Finder O (qwen2.5:1.5b) — List every timestamp/date found in this sensor telemetry readings sample.

Messy server access logs → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor O (qwen2.5-coder:1.5b) — Pull structured fields out of this server access logs sample, as a table.
• Summarizer L (gemma-3-4b-it) — Summarize what's happening in this server access logs sample, in plain English.
• Timestamp Finder O (llama3.2:1b) — List every timestamp/date found in this server access logs sample.

Concept → scenes → cyberpunk neon-style shot prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scene Breakdown, Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Scene Breakdown O (llama3.2:1b) — Break this concept into 4-6 numbered scenes.
• Prompt Writer L (lfm2.5-1.2b-instruct) — Write a cyberpunk neon-style image prompt for each scene above.

Idea → detailed flat vector illustration image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer O (llama3.2:3b) — Turn this idea into a detailed flat vector illustration-style image generation prompt.
• Negative Prompt Writer L (qwen_qwen3.5-0.8b) — Write a negative prompt to pair with the flat vector illustration prompt above — what to avoid.

Idea → 3 distinct oil painting prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A O (gemma2:2b) — Write a oil painting-style image prompt for this idea, direction A (bold/dramatic).
• Concept B L (qwen_qwen3.5-2b) — Write a oil painting-style image prompt for this idea, direction B (subtle/minimal).
• Concept C O (qwen2.5:3b) — Write a oil painting-style image prompt for this idea, direction C (unexpected angle).

Idea → detailed photorealistic image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer O (qwen2.5-coder:1.5b) — Turn this idea into a detailed photorealistic-style image generation prompt.
• Negative Prompt Writer L (gemma-3-4b-it) — Write a negative prompt to pair with the photorealistic prompt above — what to avoid.

Idea → 3 distinct watercolor painting prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A O (qwen2.5-coder:3b) — Write a watercolor painting-style image prompt for this idea, direction A (bold/dramatic).
• Concept B L (gemma-4-e2b-it-qat-mtp) — Write a watercolor painting-style image prompt for this idea, direction B (subtle/minimal).
• Concept C O (llama3.2:3b) — Write a watercolor painting-style image prompt for this idea, direction C (unexpected angle).

Text in Arabic → grammar fix → clarity tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar Fixer, Clarity Editor. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar Fixer O (gemma2:2b) — Fix grammar/spelling errors in this Arabic text. Output corrected text only.
• Clarity Editor L (qwen_qwen3.5-2b) — Tighten this Arabic text for clarity — cut redundancy, keep meaning.

UI text → French + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer O (qwen2.5-coder:3b) — Translate this UI text into French, concise and natural for a UI label.
• Tooltip Writer L (gemma-4-e2b-it-qat-mtp) — Write a short French tooltip/help text version of this UI text.

Text → formal + casual German versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice O (llama3.2:1b) — Translate/rewrite this text into formal, professional German.
• Casual Voice L (lfm2.5-1.2b-instruct) — Translate/rewrite this text into casual, everyday German.

Transcript lines → Hebrew translation → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Format Checker. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator O (llama3.2:3b) — Translate these subtitle lines into Hebrew, keeping line breaks intact.
• Format Checker L (qwen_qwen3.5-0.8b) — Verify the Hebrew lines still match the original line count/timing markers.

UI text → Italian + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer O (qwen2.5:3b) — Translate this UI text into Italian, concise and natural for a UI label.
• Tooltip Writer L (gemma-3-4b-it) — Write a short Italian tooltip/help text version of this UI text.

Translate to Japanese → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator O (qwen2.5:1.5b) — Translate the given text into Japanese. Natural, not literal.
• Back-Translator L (gemma-3-1b-it) — Translate the Japanese translation back into English.
• Drift Fixer O (deepseek-r1:1.5b) — Compare original vs back-translation; fix the Japanese version if meaning drifted.

Text → formal + casual Portuguese versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice O (deepseek-r1:1.5b) — Translate/rewrite this text into formal, professional Portuguese.
• Casual Voice L (gemma-4-e2b-it-qat-mtp) — Translate/rewrite this text into casual, everyday Portuguese.

Translate to Spanish → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator O (qwen2.5-coder:1.5b) — Translate the given text into Spanish. Natural, not literal.
• Back-Translator L (gemma-3-4b-it) — Translate the Spanish translation back into English.
• Drift Fixer O (llama3.2:1b) — Compare original vs back-translation; fix the Spanish version if meaning drifted.

Explain blockchain consensus at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher O (llama3.2:3b) — Explain blockchain consensus like the reader is 10 years old. No jargon.
• Intermediate Teacher L (qwen_qwen3.5-0.8b) — Explain blockchain consensus for someone with basic background — some technical terms OK.
• Expert O (qwen2.5:1.5b) — Explain blockchain consensus at expert depth, precise terminology, no hand-holding.

3 agents argue about compound interest; a judge synthesizes the strongest take.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Advocate A, Advocate B. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Advocate A O (llama3.2:1b) — Give the strongest argument/explanation of compound interest, angle A.
• Advocate B L (lfm2.5-1.2b-instruct) — Give the strongest argument/explanation of compound interest, angle B, different framing.

Explain photosynthesis at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher O (qwen2.5-coder:1.5b) — Explain photosynthesis like the reader is 10 years old. No jargon.
• Intermediate Teacher L (gemma-3-4b-it) — Explain photosynthesis for someone with basic background — some technical terms OK.
• Expert O (llama3.2:1b) — Explain photosynthesis at expert depth, precise terminology, no hand-holding.

Text about quantum entanglement → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor O (gemma2:2b) — Extract the key facts about quantum entanglement from the given text, as bullets.
• Flashcard Writer L (qwen_qwen3.5-2b) — Turn those facts into Q/A flashcards about quantum entanglement (JSON array of {q,a}).

Text about recursion → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor O (qwen2.5-coder:3b) — Extract the key facts about recursion from the given text, as bullets.
• Flashcard Writer L (gemma-4-e2b-it-qat-mtp) — Turn those facts into Q/A flashcards about recursion (JSON array of {q,a}).

Messy a client proposal text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer O (qwen2.5-coder:1.5b) — Summarize this a client proposal in exactly 3 bullets.
• Action Extractor L (gemma-3-4b-it) — List only concrete action items found in this a client proposal, as bullets.
• Tone Reader O (llama3.2:1b) — State the tone of this a client proposal in one sentence (e.g. urgent, neutral, frustrated).

Messy an internal memo text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor O (llama3.2:3b) — Extract structured line items from this an internal memo as a table.
• Date Finder L (qwen_qwen3.5-0.8b) — List every date/deadline mentioned in this an internal memo.
• Anomaly Spotter O (qwen2.5:1.5b) — Flag anything in this an internal memo that looks inconsistent or wrong.

Messy a job posting text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer O (qwen2.5:1.5b) — Summarize this a job posting in exactly 3 bullets.
• Action Extractor L (gemma-3-1b-it) — List only concrete action items found in this a job posting, as bullets.
• Tone Reader O (deepseek-r1:1.5b) — State the tone of this a job posting in one sentence (e.g. urgent, neutral, frustrated).

Rough a partnership agreement summary → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher O (qwen2.5-coder:1.5b) — Polish this rough a partnership agreement summary draft for grammar and flow.
• Keyword Pass L (qwen_qwen3.5-0.8b) — Add/verify relevant keywords a reader or search would expect in this a partnership agreement summary.

Intent → draft a performance review → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter O (qwen2.5:3b) — Write a first draft of a performance review given the stated intent.
• Polisher L (gemma-3-4b-it) — Polish the tone of this a performance review draft — professional, clear, no filler.
• Summarizer O (smollm2:1.7b) — Write a one-line subject and a 2-sentence TL;DR for this a performance review.

Long a policy document → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer O (deepseek-r1:1.5b) — Write a 2-sentence TL;DR of this a policy document.
• Outliner L (gemma-4-e2b-it-qat-mtp) — Extract a section-by-section outline of this a policy document.
• Term Extractor O (qwen2.5-coder:1.5b) — List key terms/jargon used in this a policy document, with one-line definitions.

Rough a press release → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher O (gemma2:2b) — Polish this rough a press release draft for grammar and flow.
• Keyword Pass L (qwen_qwen3.5-2b) — Add/verify relevant keywords a reader or search would expect in this a press release.

Intent → draft a status report → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter O (qwen2.5-coder:3b) — Write a first draft of a status report given the stated intent.
• Polisher L (gemma-4-e2b-it-qat-mtp) — Polish the tone of this a status report draft — professional, clear, no filler.
• Summarizer O (llama3.2:3b) — Write a one-line subject and a 2-sentence TL;DR for this a status report.

Long a support ticket reply → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer O (llama3.2:1b) — Write a 2-sentence TL;DR of this a support ticket reply.
• Outliner L (lfm2.5-1.2b-instruct) — Extract a section-by-section outline of this a support ticket reply.
• Term Extractor O (gemma2:2b) — List key terms/jargon used in this a support ticket reply, with one-line definitions.

Messy a user manual section text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor O (smollm2:1.7b) — Extract structured line items from this a user manual section as a table.
• Date Finder L (lfm2.5-1.2b-instruct) — List every date/deadline mentioned in this a user manual section.
• Anomaly Spotter O (qwen2.5-coder:3b) — Flag anything in this a user manual section that looks inconsistent or wrong.

HTML/UI → color contrast issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (llama3.2:1b) — Scan this HTML/UI text ONLY for color contrast issues. List concrete problems as bullets, cite the element.
• Fix Suggester (llama3.2:3b) — For each color contrast issue found, suggest a concrete fix (markup/attribute/CSS change).

HTML/UI → focus indicators issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (llama3.2:3b) — Scan this HTML/UI text ONLY for focus indicators issues. List concrete problems as bullets, cite the element.
• Fix Suggester (gemma2:2b) — For each focus indicators issue found, suggest a concrete fix (markup/attribute/CSS change).

HTML/UI → keyboard navigation issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (qwen2.5-coder:1.5b) — Scan this HTML/UI text ONLY for keyboard navigation issues. List concrete problems as bullets, cite the element.
• Fix Suggester (qwen2.5-coder:3b) — For each keyboard navigation issue found, suggest a concrete fix (markup/attribute/CSS change).

HTML/UI → screen-reader labeling issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (qwen2.5-coder:3b) — Scan this HTML/UI text ONLY for screen-reader labeling issues. List concrete problems as bullets, cite the element.
• Fix Suggester (llama3.2:1b) — For each screen-reader labeling issue found, suggest a concrete fix (markup/attribute/CSS change).

Generate Bash unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (qwen2.5:1.5b) — Write Bash unit tests for the normal/expected inputs only.
• Edge Cases (qwen2.5:3b) — Write Bash unit tests for boundary/edge-case inputs only.
• Failure Modes (deepseek-r1:1.5b) — Write Bash unit tests for invalid input / error conditions only.

Generate Bash unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (llama3.2:3b) — Write Bash unit tests for the normal/expected inputs only.
• Edge Cases (gemma2:2b) — Write Bash unit tests for boundary/edge-case inputs only.
• Failure Modes (qwen2.5:1.5b) — Write Bash unit tests for invalid input / error conditions only.

Two C# coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (smollm2:1.7b) — Solve the given task in C#, straightforward/imperative style. Code only.
• Coder B (qwen2.5-coder:1.5b) — Solve the given task in C#, a more functional/idiomatic style. Code only.

Two C# coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (qwen2.5:3b) — Solve the given task in C#, straightforward/imperative style. Code only.
• Coder B (deepseek-r1:1.5b) — Solve the given task in C#, a more functional/idiomatic style. Code only.

Two C# coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (qwen2.5-coder:3b) — Solve the given task in C#, straightforward/imperative style. Code only.
• Coder B (llama3.2:1b) — Solve the given task in C#, a more functional/idiomatic style. Code only.

Refactor C++ code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (gemma2:2b) — Rewrite the given C++ code for clarity and idiom. Output only the revised code.
• Verifier (qwen2.5:1.5b) — Sanity-check the refactored C++ code against the original — same behavior? List any risk in 3 bullets max.

Refactor C++ code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (qwen2.5:3b) — Rewrite the given C++ code for clarity and idiom. Output only the revised code.
• Verifier (deepseek-r1:1.5b) — Sanity-check the refactored C++ code against the original — same behavior? List any risk in 3 bullets max.

Refactor C++ code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (llama3.2:1b) — Rewrite the given C++ code for clarity and idiom. Output only the revised code.
• Verifier (llama3.2:3b) — Sanity-check the refactored C++ code against the original — same behavior? List any risk in 3 bullets max.

App description (Dart) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author (llama3.2:1b) — Write a production-sane Dockerfile for this Dart app description.
• Notes Writer (llama3.2:3b) — Write 3-5 short run/build notes for the Dockerfile above.

App description (Dart) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author (qwen2.5-coder:1.5b) — Write a production-sane Dockerfile for this Dart app description.
• Notes Writer (qwen2.5-coder:3b) — Write 3-5 short run/build notes for the Dockerfile above.

Error/log in Elixir → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (deepseek-r1:1.5b) — Given a Elixir error/log, state the most likely root cause in 2-3 sentences.
• Fixer (smollm2:1.7b) — Propose a concrete Elixir code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (qwen2.5-coder:1.5b) — Write one Elixir regression test that would have caught this bug.

Error/log in Elixir → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (qwen2.5:1.5b) — Given a Elixir error/log, state the most likely root cause in 2-3 sentences.
• Fixer (qwen2.5:3b) — Propose a concrete Elixir code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (deepseek-r1:1.5b) — Write one Elixir regression test that would have caught this bug.

Generate Go unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (llama3.2:1b) — Write Go unit tests for the normal/expected inputs only.
• Edge Cases (llama3.2:3b) — Write Go unit tests for boundary/edge-case inputs only.
• Failure Modes (gemma2:2b) — Write Go unit tests for invalid input / error conditions only.

Generate Go unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (gemma2:2b) — Write Go unit tests for the normal/expected inputs only.
• Edge Cases (qwen2.5:1.5b) — Write Go unit tests for boundary/edge-case inputs only.
• Failure Modes (qwen2.5:3b) — Write Go unit tests for invalid input / error conditions only.

Generate Go unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (qwen2.5:3b) — Write Go unit tests for the normal/expected inputs only.
• Edge Cases (deepseek-r1:1.5b) — Write Go unit tests for boundary/edge-case inputs only.
• Failure Modes (smollm2:1.7b) — Write Go unit tests for invalid input / error conditions only.

Error/log in Java → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (qwen2.5:3b) — Given a Java error/log, state the most likely root cause in 2-3 sentences.
• Fixer (deepseek-r1:1.5b) — Propose a concrete Java code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (smollm2:1.7b) — Write one Java regression test that would have caught this bug.

Error/log in Java → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (gemma2:2b) — Given a Java error/log, state the most likely root cause in 2-3 sentences.
• Fixer (qwen2.5:1.5b) — Propose a concrete Java code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (qwen2.5:3b) — Write one Java regression test that would have caught this bug.

Error/log in Java → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (smollm2:1.7b) — Given a Java error/log, state the most likely root cause in 2-3 sentences.
• Fixer (qwen2.5-coder:1.5b) — Propose a concrete Java code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (qwen2.5-coder:3b) — Write one Java regression test that would have caught this bug.

Refactor JavaScript code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (qwen2.5-coder:3b) — Rewrite the given JavaScript code for clarity and idiom. Output only the revised code.
• Verifier (llama3.2:1b) — Sanity-check the refactored JavaScript code against the original — same behavior? List any risk in 3 bullets max.

Refactor JavaScript code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (llama3.2:3b) — Rewrite the given JavaScript code for clarity and idiom. Output only the revised code.
• Verifier (gemma2:2b) — Sanity-check the refactored JavaScript code against the original — same behavior? List any risk in 3 bullets max.

Refactor JavaScript code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (qwen2.5:1.5b) — Rewrite the given JavaScript code for clarity and idiom. Output only the revised code.
• Verifier (qwen2.5:3b) — Sanity-check the refactored JavaScript code against the original — same behavior? List any risk in 3 bullets max.

Kotlin code → plain-English explanation + complexity + risk notes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Explainer, Complexity, Risk Spotter. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Explainer (qwen2.5-coder:1.5b) — Explain what this Kotlin code does in plain English, no jargon.
• Complexity (qwen2.5-coder:3b) — State the time/space complexity of this Kotlin code in one line, with reasoning.
• Risk Spotter (llama3.2:1b) — List any risk (security, correctness, maintainability) in this Kotlin code. Bullets.

Kotlin code → plain-English explanation + complexity + risk notes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Explainer, Complexity, Risk Spotter. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Explainer (gemma2:2b) — Explain what this Kotlin code does in plain English, no jargon.
• Complexity (qwen2.5:1.5b) — State the time/space complexity of this Kotlin code in one line, with reasoning.
• Risk Spotter (qwen2.5:3b) — List any risk (security, correctness, maintainability) in this Kotlin code. Bullets.

Kotlin code → plain-English explanation + complexity + risk notes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Explainer, Complexity, Risk Spotter. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Explainer (llama3.2:1b) — Explain what this Kotlin code does in plain English, no jargon.
• Complexity (llama3.2:3b) — State the time/space complexity of this Kotlin code in one line, with reasoning.
• Risk Spotter (gemma2:2b) — List any risk (security, correctness, maintainability) in this Kotlin code. Bullets.

Lua diff/notes → summary → Conventional Commit message.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Summarizer, Commit Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Summarizer (smollm2:1.7b) — Summarize what this Lua diff/change actually does, in plain English.
• Commit Writer (qwen2.5-coder:1.5b) — Write a Conventional Commits message (type, scope, subject, body) for this change.

Lua diff/notes → summary → Conventional Commit message.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Summarizer, Commit Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Summarizer (qwen2.5-coder:3b) — Summarize what this Lua diff/change actually does, in plain English.
• Commit Writer (llama3.2:1b) — Write a Conventional Commits message (type, scope, subject, body) for this change.

Idea → REST endpoints in PHP conventions → example payloads.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: API Designer, Example Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• API Designer (deepseek-r1:1.5b) — Given a feature idea, propose REST endpoints following PHP-ecosystem conventions.
• Example Writer (smollm2:1.7b) — Write example request/response JSON payloads for the proposed endpoints.

Idea → REST endpoints in PHP conventions → example payloads.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: API Designer, Example Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• API Designer (qwen2.5-coder:1.5b) — Given a feature idea, propose REST endpoints following PHP-ecosystem conventions.
• Example Writer (qwen2.5-coder:3b) — Write example request/response JSON payloads for the proposed endpoints.

Idea → REST endpoints in PHP conventions → example payloads.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: API Designer, Example Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• API Designer (qwen2.5:1.5b) — Given a feature idea, propose REST endpoints following PHP-ecosystem conventions.
• Example Writer (qwen2.5:3b) — Write example request/response JSON payloads for the proposed endpoints.

Two Perl coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (deepseek-r1:1.5b) — Solve the given task in Perl, straightforward/imperative style. Code only.
• Coder B (smollm2:1.7b) — Solve the given task in Perl, a more functional/idiomatic style. Code only.

Two Perl coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (qwen2.5-coder:1.5b) — Solve the given task in Perl, straightforward/imperative style. Code only.
• Coder B (qwen2.5-coder:3b) — Solve the given task in Perl, a more functional/idiomatic style. Code only.

Three specialists audit Python code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (qwen2.5-coder:1.5b) — List ONLY Python syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker (qwen2.5-coder:3b) — Point out non-idiomatic Python — what a senior Python dev would flag in review. Bullets only.
• Performance (llama3.2:1b) — Report ONLY performance concerns for this Python code — hot loops, allocations, complexity. Bullets.

Three specialists audit Python code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (gemma2:2b) — List ONLY Python syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker (qwen2.5:1.5b) — Point out non-idiomatic Python — what a senior Python dev would flag in review. Bullets only.
• Performance (qwen2.5:3b) — Report ONLY performance concerns for this Python code — hot loops, allocations, complexity. Bullets.

Three specialists audit Python code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (llama3.2:1b) — List ONLY Python syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker (llama3.2:3b) — Point out non-idiomatic Python — what a senior Python dev would flag in review. Bullets only.
• Performance (gemma2:2b) — Report ONLY performance concerns for this Python code — hot loops, allocations, complexity. Bullets.

R code → plain-English explanation + complexity + risk notes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Explainer, Complexity, Risk Spotter. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Explainer (llama3.2:3b) — Explain what this R code does in plain English, no jargon.
• Complexity (gemma2:2b) — State the time/space complexity of this R code in one line, with reasoning.
• Risk Spotter (qwen2.5:1.5b) — List any risk (security, correctness, maintainability) in this R code. Bullets.

R code → plain-English explanation + complexity + risk notes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Explainer, Complexity, Risk Spotter. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Explainer (qwen2.5-coder:3b) — Explain what this R code does in plain English, no jargon.
• Complexity (llama3.2:1b) — State the time/space complexity of this R code in one line, with reasoning.
• Risk Spotter (llama3.2:3b) — List any risk (security, correctness, maintainability) in this R code. Bullets.

App description (Ruby) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author (smollm2:1.7b) — Write a production-sane Dockerfile for this Ruby app description.
• Notes Writer (qwen2.5-coder:1.5b) — Write 3-5 short run/build notes for the Dockerfile above.

App description (Ruby) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author (llama3.2:3b) — Write a production-sane Dockerfile for this Ruby app description.
• Notes Writer (gemma2:2b) — Write 3-5 short run/build notes for the Dockerfile above.

App description (Ruby) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author (qwen2.5-coder:3b) — Write a production-sane Dockerfile for this Ruby app description.
• Notes Writer (llama3.2:1b) — Write 3-5 short run/build notes for the Dockerfile above.

Turn Rust code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer (deepseek-r1:1.5b) — Write a concise Markdown README for this Rust code: what it does, how to use it.
• Editor (smollm2:1.7b) — Tighten the README — cut fluff, keep only what a new user actually needs.

Turn Rust code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer (llama3.2:3b) — Write a concise Markdown README for this Rust code: what it does, how to use it.
• Editor (gemma2:2b) — Tighten the README — cut fluff, keep only what a new user actually needs.

Turn Rust code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer (qwen2.5:1.5b) — Write a concise Markdown README for this Rust code: what it does, how to use it.
• Editor (qwen2.5:3b) — Tighten the README — cut fluff, keep only what a new user actually needs.

Turn SQL code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer (gemma2:2b) — Write a concise Markdown README for this SQL code: what it does, how to use it.
• Editor (qwen2.5:1.5b) — Tighten the README — cut fluff, keep only what a new user actually needs.

Turn SQL code into a short README, then tighten it.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Doc Writer, Editor. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Doc Writer (qwen2.5:3b) — Write a concise Markdown README for this SQL code: what it does, how to use it.
• Editor (deepseek-r1:1.5b) — Tighten the README — cut fluff, keep only what a new user actually needs.

Idea → REST endpoints in Scala conventions → example payloads.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: API Designer, Example Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• API Designer (smollm2:1.7b) — Given a feature idea, propose REST endpoints following Scala-ecosystem conventions.
• Example Writer (qwen2.5-coder:1.5b) — Write example request/response JSON payloads for the proposed endpoints.

Idea → REST endpoints in Scala conventions → example payloads.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: API Designer, Example Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• API Designer (qwen2.5:3b) — Given a feature idea, propose REST endpoints following Scala-ecosystem conventions.
• Example Writer (deepseek-r1:1.5b) — Write example request/response JSON payloads for the proposed endpoints.

Three specialists audit Swift code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (qwen2.5:1.5b) — List ONLY Swift syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker (qwen2.5:3b) — Point out non-idiomatic Swift — what a senior Swift dev would flag in review. Bullets only.
• Performance (deepseek-r1:1.5b) — Report ONLY performance concerns for this Swift code — hot loops, allocations, complexity. Bullets.

Three specialists audit Swift code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (qwen2.5-coder:3b) — List ONLY Swift syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker (llama3.2:1b) — Point out non-idiomatic Swift — what a senior Swift dev would flag in review. Bullets only.
• Performance (llama3.2:3b) — Report ONLY performance concerns for this Swift code — hot loops, allocations, complexity. Bullets.

Three specialists audit Swift code: bugs, style/idioms, and performance.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Bug Hunter, Idiom Checker, Performance. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Bug Hunter (llama3.2:3b) — List ONLY Swift syntax errors, logic flaws, likely bugs as short bullets. No praise, no rewrite.
• Idiom Checker (gemma2:2b) — Point out non-idiomatic Swift — what a senior Swift dev would flag in review. Bullets only.
• Performance (qwen2.5:1.5b) — Report ONLY performance concerns for this Swift code — hot loops, allocations, complexity. Bullets.

TypeScript diff/notes → summary → Conventional Commit message.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Summarizer, Commit Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Summarizer (deepseek-r1:1.5b) — Summarize what this TypeScript diff/change actually does, in plain English.
• Commit Writer (smollm2:1.7b) — Write a Conventional Commits message (type, scope, subject, body) for this change.

TypeScript diff/notes → summary → Conventional Commit message.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Summarizer, Commit Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Summarizer (qwen2.5-coder:1.5b) — Summarize what this TypeScript diff/change actually does, in plain English.
• Commit Writer (qwen2.5-coder:3b) — Write a Conventional Commits message (type, scope, subject, body) for this change.

TypeScript diff/notes → summary → Conventional Commit message.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Summarizer, Commit Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Summarizer (llama3.2:1b) — Summarize what this TypeScript diff/change actually does, in plain English.
• Commit Writer (llama3.2:3b) — Write a Conventional Commits message (type, scope, subject, body) for this change.

Premise → cozy mystery outline → next chapter → cliffhanger.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outliner, Writer, Hook Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Outliner (llama3.2:3b) — Outline the next 3 beats of this cozy mystery story premise.
• Writer (gemma2:2b) — Write the next chapter/scene in cozy mystery style, following the outline.
• Hook Writer (qwen2.5:1.5b) — Write a one-paragraph cliffhanger ending for this cozy mystery chapter.

Premise → cyberpunk heist outline → next chapter → cliffhanger.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outliner, Writer, Hook Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Outliner (deepseek-r1:1.5b) — Outline the next 3 beats of this cyberpunk heist story premise.
• Writer (smollm2:1.7b) — Write the next chapter/scene in cyberpunk heist style, following the outline.
• Hook Writer (qwen2.5-coder:1.5b) — Write a one-paragraph cliffhanger ending for this cyberpunk heist chapter.

One prompt → fairy-tale retelling setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect (smollm2:1.7b) — Describe a fairy-tale retelling setting for this prompt — geography, mood, era.
• Faction Writer (qwen2.5-coder:1.5b) — Invent 2-3 factions/groups that fit this fairy-tale retelling setting.
• NPC Writer (qwen2.5-coder:3b) — Invent 2 memorable NPCs that fit this fairy-tale retelling setting.

One prompt → gothic horror setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect (qwen2.5:3b) — Describe a gothic horror setting for this prompt — geography, mood, era.
• Faction Writer (deepseek-r1:1.5b) — Invent 2-3 factions/groups that fit this gothic horror setting.
• NPC Writer (smollm2:1.7b) — Invent 2 memorable NPCs that fit this gothic horror setting.

One prompt → hard sci-fi setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect (qwen2.5-coder:1.5b) — Describe a hard sci-fi setting for this prompt — geography, mood, era.
• Faction Writer (qwen2.5-coder:3b) — Invent 2-3 factions/groups that fit this hard sci-fi setting.
• NPC Writer (llama3.2:1b) — Invent 2 memorable NPCs that fit this hard sci-fi setting.

Premise → high fantasy outline → next chapter → cliffhanger.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outliner, Writer, Hook Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Outliner (qwen2.5-coder:3b) — Outline the next 3 beats of this high fantasy story premise.
• Writer (llama3.2:1b) — Write the next chapter/scene in high fantasy style, following the outline.
• Hook Writer (llama3.2:3b) — Write a one-paragraph cliffhanger ending for this high fantasy chapter.

Premise → historical fiction outline → next chapter → cliffhanger.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outliner, Writer, Hook Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Outliner (qwen2.5-coder:1.5b) — Outline the next 3 beats of this historical fiction story premise.
• Writer (qwen2.5-coder:3b) — Write the next chapter/scene in historical fiction style, following the outline.
• Hook Writer (llama3.2:1b) — Write a one-paragraph cliffhanger ending for this historical fiction chapter.

One prompt → noir detective setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect (llama3.2:1b) — Describe a noir detective setting for this prompt — geography, mood, era.
• Faction Writer (llama3.2:3b) — Invent 2-3 factions/groups that fit this noir detective setting.
• NPC Writer (gemma2:2b) — Invent 2 memorable NPCs that fit this noir detective setting.

One prompt → post-apocalyptic survival setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect (gemma2:2b) — Describe a post-apocalyptic survival setting for this prompt — geography, mood, era.
• Faction Writer (qwen2.5:1.5b) — Invent 2-3 factions/groups that fit this post-apocalyptic survival setting.
• NPC Writer (qwen2.5:3b) — Invent 2 memorable NPCs that fit this post-apocalyptic survival setting.

One prompt → romantic comedy setting + factions + tech/magic + NPCs, in parallel.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: World Architect, Faction Writer, NPC Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• World Architect (qwen2.5-coder:3b) — Describe a romantic comedy setting for this prompt — geography, mood, era.
• Faction Writer (llama3.2:1b) — Invent 2-3 factions/groups that fit this romantic comedy setting.
• NPC Writer (llama3.2:3b) — Invent 2 memorable NPCs that fit this romantic comedy setting.

Premise → space opera outline → next chapter → cliffhanger.

Good for: fiction writing — worldbuilding, plotting, or continuing a story.

Example: Say you paste in a story idea, character, or scene you're building. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Outliner, Writer, Hook Writer. Use it when you want fiction writing — worldbuilding, plotting, or continuing a story.

Steps (3):

• Outliner (qwen2.5:1.5b) — Outline the next 3 beats of this space opera story premise.
• Writer (qwen2.5:3b) — Write the next chapter/scene in space opera style, following the outline.
• Hook Writer (deepseek-r1:1.5b) — Write a one-paragraph cliffhanger ending for this space opera chapter.

clinical trial data notes → summary + notable claims + what to verify, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Claim Spotter, Verifier. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Summarizer (smollm2:1.7b) — Summarize the key findings visible in this clinical trial data notes sample.
• Claim Spotter (qwen2.5-coder:1.5b) — List any strong/surprising claims implied by this clinical trial data notes sample.
• Verifier (qwen2.5-coder:3b) — List what in this clinical trial data notes sample would need independent verification before acting on it.

customer support transcripts → summary + notable claims + what to verify, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Claim Spotter, Verifier. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Summarizer (llama3.2:1b) — Summarize the key findings visible in this customer support transcripts sample.
• Claim Spotter (llama3.2:3b) — List any strong/surprising claims implied by this customer support transcripts sample.
• Verifier (gemma2:2b) — List what in this customer support transcripts sample would need independent verification before acting on it.

customer support transcripts → summary + notable claims + what to verify, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Claim Spotter, Verifier. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Summarizer (qwen2.5:1.5b) — Summarize the key findings visible in this customer support transcripts sample.
• Claim Spotter (qwen2.5:3b) — List any strong/surprising claims implied by this customer support transcripts sample.
• Verifier (deepseek-r1:1.5b) — List what in this customer support transcripts sample would need independent verification before acting on it.

Messy e-commerce order exports → schema guess + cleaned rows + anomalies, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Schema Guesser, Cleaner, Anomaly Spotter. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Schema Guesser (gemma2:2b) — Guess a reasonable schema (column names + types) for this e-commerce order exports sample.
• Cleaner (qwen2.5:1.5b) — Reformat this e-commerce order exports sample into clean, consistent rows.
• Anomaly Spotter (qwen2.5:3b) — Flag rows/values in this e-commerce order exports sample that look wrong or inconsistent.

Messy e-commerce order exports → schema guess + cleaned rows + anomalies, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Schema Guesser, Cleaner, Anomaly Spotter. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Schema Guesser (qwen2.5-coder:3b) — Guess a reasonable schema (column names + types) for this e-commerce order exports sample.
• Cleaner (llama3.2:1b) — Reformat this e-commerce order exports sample into clean, consistent rows.
• Anomaly Spotter (llama3.2:3b) — Flag rows/values in this e-commerce order exports sample that look wrong or inconsistent.

Messy financial transaction records → schema guess + cleaned rows + anomalies, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Schema Guesser, Cleaner, Anomaly Spotter. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Schema Guesser (gemma2:2b) — Guess a reasonable schema (column names + types) for this financial transaction records sample.
• Cleaner (qwen2.5:1.5b) — Reformat this financial transaction records sample into clean, consistent rows.
• Anomaly Spotter (qwen2.5:3b) — Flag rows/values in this financial transaction records sample that look wrong or inconsistent.

Messy inventory stock reports → schema guess + cleaned rows + anomalies, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Schema Guesser, Cleaner, Anomaly Spotter. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Schema Guesser (qwen2.5-coder:3b) — Guess a reasonable schema (column names + types) for this inventory stock reports sample.
• Cleaner (llama3.2:1b) — Reformat this inventory stock reports sample into clean, consistent rows.
• Anomaly Spotter (llama3.2:3b) — Flag rows/values in this inventory stock reports sample that look wrong or inconsistent.

Messy IoT device logs → schema guess + cleaned rows + anomalies, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Schema Guesser, Cleaner, Anomaly Spotter. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Schema Guesser (deepseek-r1:1.5b) — Guess a reasonable schema (column names + types) for this IoT device logs sample.
• Cleaner (smollm2:1.7b) — Reformat this IoT device logs sample into clean, consistent rows.
• Anomaly Spotter (qwen2.5-coder:1.5b) — Flag rows/values in this IoT device logs sample that look wrong or inconsistent.

Messy sensor telemetry readings → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor (llama3.2:3b) — Pull structured fields out of this sensor telemetry readings sample, as a table.
• Summarizer (gemma2:2b) — Summarize what's happening in this sensor telemetry readings sample, in plain English.
• Timestamp Finder (qwen2.5:1.5b) — List every timestamp/date found in this sensor telemetry readings sample.

Messy server access logs → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor (qwen2.5-coder:1.5b) — Pull structured fields out of this server access logs sample, as a table.
• Summarizer (qwen2.5-coder:3b) — Summarize what's happening in this server access logs sample, in plain English.
• Timestamp Finder (llama3.2:1b) — List every timestamp/date found in this server access logs sample.

Messy server access logs → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor (llama3.2:3b) — Pull structured fields out of this server access logs sample, as a table.
• Summarizer (gemma2:2b) — Summarize what's happening in this server access logs sample, in plain English.
• Timestamp Finder (qwen2.5:1.5b) — List every timestamp/date found in this server access logs sample.

shipment tracking logs → summary + notable claims + what to verify, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Claim Spotter, Verifier. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Summarizer (llama3.2:1b) — Summarize the key findings visible in this shipment tracking logs sample.
• Claim Spotter (llama3.2:3b) — List any strong/surprising claims implied by this shipment tracking logs sample.
• Verifier (gemma2:2b) — List what in this shipment tracking logs sample would need independent verification before acting on it.

Messy social media engagement exports → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor (qwen2.5:3b) — Pull structured fields out of this social media engagement exports sample, as a table.
• Summarizer (deepseek-r1:1.5b) — Summarize what's happening in this social media engagement exports sample, in plain English.
• Timestamp Finder (smollm2:1.7b) — List every timestamp/date found in this social media engagement exports sample.

survey response exports → summary + notable claims + what to verify, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Claim Spotter, Verifier. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Summarizer (qwen2.5:1.5b) — Summarize the key findings visible in this survey response exports sample.
• Claim Spotter (qwen2.5:3b) — List any strong/surprising claims implied by this survey response exports sample.
• Verifier (deepseek-r1:1.5b) — List what in this survey response exports sample would need independent verification before acting on it.

Messy website analytics exports → structured tokens + human summary + timestamps, in parallel.

Good for: checking claims, cleaning messy data, or digesting long research material.

Example: Say you paste in a claim, a messy CSV snippet, or a stack of research notes. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Summarizer, Timestamp Finder. Use it when you want checking claims, cleaning messy data, or digesting long research material.

Steps (3):

• Extractor (qwen2.5-coder:1.5b) — Pull structured fields out of this website analytics exports sample, as a table.
• Summarizer (qwen2.5-coder:3b) — Summarize what's happening in this website analytics exports sample, in plain English.
• Timestamp Finder (llama3.2:1b) — List every timestamp/date found in this website analytics exports sample.

Idea → 3 distinct anime/manga prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A (deepseek-r1:1.5b) — Write a anime/manga-style image prompt for this idea, direction A (bold/dramatic).
• Concept B (smollm2:1.7b) — Write a anime/manga-style image prompt for this idea, direction B (subtle/minimal).
• Concept C (qwen2.5-coder:1.5b) — Write a anime/manga-style image prompt for this idea, direction C (unexpected angle).

Idea → detailed art nouveau image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer (qwen2.5-coder:1.5b) — Turn this idea into a detailed art nouveau-style image generation prompt.
• Negative Prompt Writer (qwen2.5-coder:3b) — Write a negative prompt to pair with the art nouveau prompt above — what to avoid.

Concept → scenes → charcoal sketch-style shot prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scene Breakdown, Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Scene Breakdown (llama3.2:1b) — Break this concept into 4-6 numbered scenes.
• Prompt Writer (llama3.2:3b) — Write a charcoal sketch-style image prompt for each scene above.

Idea → detailed claymation image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer (llama3.2:3b) — Turn this idea into a detailed claymation-style image generation prompt.
• Negative Prompt Writer (gemma2:2b) — Write a negative prompt to pair with the claymation prompt above — what to avoid.

Concept → scenes → concept art-style shot prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scene Breakdown, Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Scene Breakdown (smollm2:1.7b) — Break this concept into 4-6 numbered scenes.
• Prompt Writer (qwen2.5-coder:1.5b) — Write a concept art-style image prompt for each scene above.

Concept → scenes → cyberpunk neon-style shot prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scene Breakdown, Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Scene Breakdown (llama3.2:1b) — Break this concept into 4-6 numbered scenes.
• Prompt Writer (llama3.2:3b) — Write a cyberpunk neon-style image prompt for each scene above.

Concept → scenes → film noir-style shot prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scene Breakdown, Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Scene Breakdown (smollm2:1.7b) — Break this concept into 4-6 numbered scenes.
• Prompt Writer (qwen2.5-coder:1.5b) — Write a film noir-style image prompt for each scene above.

Idea → detailed flat vector illustration image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer (llama3.2:3b) — Turn this idea into a detailed flat vector illustration-style image generation prompt.
• Negative Prompt Writer (gemma2:2b) — Write a negative prompt to pair with the flat vector illustration prompt above — what to avoid.

Concept → scenes → isometric 3D render-style shot prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scene Breakdown, Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Scene Breakdown (qwen2.5:1.5b) — Break this concept into 4-6 numbered scenes.
• Prompt Writer (qwen2.5:3b) — Write a isometric 3D render-style image prompt for each scene above.

Idea → 3 distinct low-poly 3D prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A (qwen2.5-coder:3b) — Write a low-poly 3D-style image prompt for this idea, direction A (bold/dramatic).
• Concept B (llama3.2:1b) — Write a low-poly 3D-style image prompt for this idea, direction B (subtle/minimal).
• Concept C (llama3.2:3b) — Write a low-poly 3D-style image prompt for this idea, direction C (unexpected angle).

Idea → detailed minimalist line art image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer (qwen2.5:3b) — Turn this idea into a detailed minimalist line art-style image generation prompt.
• Negative Prompt Writer (deepseek-r1:1.5b) — Write a negative prompt to pair with the minimalist line art prompt above — what to avoid.

Idea → 3 distinct oil painting prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A (gemma2:2b) — Write a oil painting-style image prompt for this idea, direction A (bold/dramatic).
• Concept B (qwen2.5:1.5b) — Write a oil painting-style image prompt for this idea, direction B (subtle/minimal).
• Concept C (qwen2.5:3b) — Write a oil painting-style image prompt for this idea, direction C (unexpected angle).

Idea → detailed photorealistic image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer (qwen2.5-coder:1.5b) — Turn this idea into a detailed photorealistic-style image generation prompt.
• Negative Prompt Writer (qwen2.5-coder:3b) — Write a negative prompt to pair with the photorealistic prompt above — what to avoid.

Idea → detailed pixel art image-generation prompt → negative prompt.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Prompt Writer, Negative Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Prompt Writer (qwen2.5:3b) — Turn this idea into a detailed pixel art-style image generation prompt.
• Negative Prompt Writer (deepseek-r1:1.5b) — Write a negative prompt to pair with the pixel art prompt above — what to avoid.

Idea → 3 distinct pop art prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A (deepseek-r1:1.5b) — Write a pop art-style image prompt for this idea, direction A (bold/dramatic).
• Concept B (smollm2:1.7b) — Write a pop art-style image prompt for this idea, direction B (subtle/minimal).
• Concept C (qwen2.5-coder:1.5b) — Write a pop art-style image prompt for this idea, direction C (unexpected angle).

Concept → scenes → steampunk-style shot prompts.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Scene Breakdown, Prompt Writer. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (2):

• Scene Breakdown (qwen2.5:1.5b) — Break this concept into 4-6 numbered scenes.
• Prompt Writer (qwen2.5:3b) — Write a steampunk-style image prompt for each scene above.

Idea → 3 distinct vaporwave prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A (gemma2:2b) — Write a vaporwave-style image prompt for this idea, direction A (bold/dramatic).
• Concept B (qwen2.5:1.5b) — Write a vaporwave-style image prompt for this idea, direction B (subtle/minimal).
• Concept C (qwen2.5:3b) — Write a vaporwave-style image prompt for this idea, direction C (unexpected angle).

Idea → 3 distinct watercolor painting prompt directions, in parallel.

Good for: turning a rough idea into a detailed prompt for an image generator.

Example: Say you paste in one line describing what you want to see (e.g. "a lighthouse at sunset"). This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Concept A, Concept B, Concept C. Use it when you want turning a rough idea into a detailed prompt for an image generator.

Steps (3):

• Concept A (qwen2.5-coder:3b) — Write a watercolor painting-style image prompt for this idea, direction A (bold/dramatic).
• Concept B (llama3.2:1b) — Write a watercolor painting-style image prompt for this idea, direction B (subtle/minimal).
• Concept C (llama3.2:3b) — Write a watercolor painting-style image prompt for this idea, direction C (unexpected angle).

Text in Arabic → grammar fix → clarity tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar Fixer, Clarity Editor. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar Fixer (gemma2:2b) — Fix grammar/spelling errors in this Arabic text. Output corrected text only.
• Clarity Editor (qwen2.5:1.5b) — Tighten this Arabic text for clarity — cut redundancy, keep meaning.

Text in Arabic → grammar fix → clarity tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar Fixer, Clarity Editor. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar Fixer (qwen2.5:3b) — Fix grammar/spelling errors in this Arabic text. Output corrected text only.
• Clarity Editor (deepseek-r1:1.5b) — Tighten this Arabic text for clarity — cut redundancy, keep meaning.

Translate to Dutch → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator (qwen2.5-coder:3b) — Translate the given text into Dutch. Natural, not literal.
• Back-Translator (llama3.2:1b) — Translate the Dutch translation back into English.
• Drift Fixer (llama3.2:3b) — Compare original vs back-translation; fix the Dutch version if meaning drifted.

UI text → French + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer (qwen2.5-coder:3b) — Translate this UI text into French, concise and natural for a UI label.
• Tooltip Writer (llama3.2:1b) — Write a short French tooltip/help text version of this UI text.

UI text → French + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer (llama3.2:3b) — Translate this UI text into French, concise and natural for a UI label.
• Tooltip Writer (gemma2:2b) — Write a short French tooltip/help text version of this UI text.

Text → formal + casual German versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice (gemma2:2b) — Translate/rewrite this text into formal, professional German.
• Casual Voice (qwen2.5:1.5b) — Translate/rewrite this text into casual, everyday German.

Text → formal + casual German versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice (llama3.2:1b) — Translate/rewrite this text into formal, professional German.
• Casual Voice (llama3.2:3b) — Translate/rewrite this text into casual, everyday German.

UI text → Greek + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer (deepseek-r1:1.5b) — Translate this UI text into Greek, concise and natural for a UI label.
• Tooltip Writer (smollm2:1.7b) — Write a short Greek tooltip/help text version of this UI text.

Transcript lines → Hebrew translation → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Format Checker. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator (qwen2.5:1.5b) — Translate these subtitle lines into Hebrew, keeping line breaks intact.
• Format Checker (qwen2.5:3b) — Verify the Hebrew lines still match the original line count/timing markers.

Transcript lines → Hebrew translation → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Format Checker. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator (llama3.2:3b) — Translate these subtitle lines into Hebrew, keeping line breaks intact.
• Format Checker (gemma2:2b) — Verify the Hebrew lines still match the original line count/timing markers.

Transcript lines → Hindi translation → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Format Checker. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator (gemma2:2b) — Translate these subtitle lines into Hindi, keeping line breaks intact.
• Format Checker (qwen2.5:1.5b) — Verify the Hindi lines still match the original line count/timing markers.

UI text → Italian + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer (smollm2:1.7b) — Translate this UI text into Italian, concise and natural for a UI label.
• Tooltip Writer (qwen2.5-coder:1.5b) — Write a short Italian tooltip/help text version of this UI text.

UI text → Italian + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer (qwen2.5:3b) — Translate this UI text into Italian, concise and natural for a UI label.
• Tooltip Writer (deepseek-r1:1.5b) — Write a short Italian tooltip/help text version of this UI text.

Translate to Japanese → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator (deepseek-r1:1.5b) — Translate the given text into Japanese. Natural, not literal.
• Back-Translator (smollm2:1.7b) — Translate the Japanese translation back into English.
• Drift Fixer (qwen2.5-coder:1.5b) — Compare original vs back-translation; fix the Japanese version if meaning drifted.

Translate to Japanese → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator (qwen2.5:1.5b) — Translate the given text into Japanese. Natural, not literal.
• Back-Translator (qwen2.5:3b) — Translate the Japanese translation back into English.
• Drift Fixer (deepseek-r1:1.5b) — Compare original vs back-translation; fix the Japanese version if meaning drifted.

Text in Korean → grammar fix → clarity tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar Fixer, Clarity Editor. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar Fixer (llama3.2:1b) — Fix grammar/spelling errors in this Korean text. Output corrected text only.
• Clarity Editor (llama3.2:3b) — Tighten this Korean text for clarity — cut redundancy, keep meaning.

Text in Korean → grammar fix → clarity tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar Fixer, Clarity Editor. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar Fixer (qwen2.5-coder:1.5b) — Fix grammar/spelling errors in this Korean text. Output corrected text only.
• Clarity Editor (qwen2.5-coder:3b) — Tighten this Korean text for clarity — cut redundancy, keep meaning.

Text in Mandarin Chinese → grammar fix → clarity tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar Fixer, Clarity Editor. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar Fixer (qwen2.5:1.5b) — Fix grammar/spelling errors in this Mandarin Chinese text. Output corrected text only.
• Clarity Editor (qwen2.5:3b) — Tighten this Mandarin Chinese text for clarity — cut redundancy, keep meaning.

UI text → Polish + tooltip variants, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Localizer, Tooltip Writer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Localizer (llama3.2:1b) — Translate this UI text into Polish, concise and natural for a UI label.
• Tooltip Writer (llama3.2:3b) — Write a short Polish tooltip/help text version of this UI text.

Text → formal + casual Portuguese versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice (deepseek-r1:1.5b) — Translate/rewrite this text into formal, professional Portuguese.
• Casual Voice (smollm2:1.7b) — Translate/rewrite this text into casual, everyday Portuguese.

Text → formal + casual Portuguese versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice (qwen2.5-coder:1.5b) — Translate/rewrite this text into formal, professional Portuguese.
• Casual Voice (qwen2.5-coder:3b) — Translate/rewrite this text into casual, everyday Portuguese.

Transcript lines → Russian translation → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Format Checker. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator (smollm2:1.7b) — Translate these subtitle lines into Russian, keeping line breaks intact.
• Format Checker (qwen2.5-coder:1.5b) — Verify the Russian lines still match the original line count/timing markers.

Transcript lines → Russian translation → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Format Checker. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator (qwen2.5-coder:3b) — Translate these subtitle lines into Russian, keeping line breaks intact.
• Format Checker (llama3.2:1b) — Verify the Russian lines still match the original line count/timing markers.

Translate to Spanish → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator (qwen2.5-coder:1.5b) — Translate the given text into Spanish. Natural, not literal.
• Back-Translator (qwen2.5-coder:3b) — Translate the Spanish translation back into English.
• Drift Fixer (llama3.2:1b) — Compare original vs back-translation; fix the Spanish version if meaning drifted.

Translate to Spanish → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator (llama3.2:1b) — Translate the given text into Spanish. Natural, not literal.
• Back-Translator (llama3.2:3b) — Translate the Spanish translation back into English.
• Drift Fixer (gemma2:2b) — Compare original vs back-translation; fix the Spanish version if meaning drifted.

Translate to Swedish → back-translate to verify → fix drift.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Back-Translator, Drift Fixer. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (3):

• Translator (qwen2.5:3b) — Translate the given text into Swedish. Natural, not literal.
• Back-Translator (deepseek-r1:1.5b) — Translate the Swedish translation back into English.
• Drift Fixer (smollm2:1.7b) — Compare original vs back-translation; fix the Swedish version if meaning drifted.

Transcript lines → Thai translation → keep timing/format.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Translator, Format Checker. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Translator (qwen2.5-coder:1.5b) — Translate these subtitle lines into Thai, keeping line breaks intact.
• Format Checker (qwen2.5-coder:3b) — Verify the Thai lines still match the original line count/timing markers.

Text → formal + casual Turkish versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice (llama3.2:3b) — Translate/rewrite this text into formal, professional Turkish.
• Casual Voice (gemma2:2b) — Translate/rewrite this text into casual, everyday Turkish.

Text in Ukrainian → grammar fix → clarity tighten.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Grammar Fixer, Clarity Editor. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Grammar Fixer (qwen2.5-coder:3b) — Fix grammar/spelling errors in this Ukrainian text. Output corrected text only.
• Clarity Editor (llama3.2:1b) — Tighten this Ukrainian text for clarity — cut redundancy, keep meaning.

Text → formal + casual Vietnamese versions, in parallel.

Good for: translating or adapting text between languages, tones, or reading levels.

Example: Say you paste in a sentence or paragraph you want translated or rewritten. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Formal Voice, Casual Voice. Use it when you want translating or adapting text between languages, tones, or reading levels.

Steps (2):

• Formal Voice (smollm2:1.7b) — Translate/rewrite this text into formal, professional Vietnamese.
• Casual Voice (qwen2.5-coder:1.5b) — Translate/rewrite this text into casual, everyday Vietnamese.

3 agents argue about Bayesian inference; a judge synthesizes the strongest take.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Advocate A, Advocate B. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Advocate A (qwen2.5:1.5b) — Give the strongest argument/explanation of Bayesian inference, angle A.
• Advocate B (qwen2.5:3b) — Give the strongest argument/explanation of Bayesian inference, angle B, different framing.

Explain blockchain consensus at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher (llama3.2:3b) — Explain blockchain consensus like the reader is 10 years old. No jargon.
• Intermediate Teacher (gemma2:2b) — Explain blockchain consensus for someone with basic background — some technical terms OK.
• Expert (qwen2.5:1.5b) — Explain blockchain consensus at expert depth, precise terminology, no hand-holding.

3 agents argue about compound interest; a judge synthesizes the strongest take.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Advocate A, Advocate B. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Advocate A (llama3.2:1b) — Give the strongest argument/explanation of compound interest, angle A.
• Advocate B (llama3.2:3b) — Give the strongest argument/explanation of compound interest, angle B, different framing.

3 agents argue about DNA transcription; a judge synthesizes the strongest take.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Advocate A, Advocate B. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Advocate A (llama3.2:1b) — Give the strongest argument/explanation of DNA transcription, angle A.
• Advocate B (llama3.2:3b) — Give the strongest argument/explanation of DNA transcription, angle B, different framing.

Text about game theory Nash equilibrium → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor (qwen2.5-coder:3b) — Extract the key facts about game theory Nash equilibrium from the given text, as bullets.
• Flashcard Writer (llama3.2:1b) — Turn those facts into Q/A flashcards about game theory Nash equilibrium (JSON array of {q,a}).

Text about greenhouse effect → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor (deepseek-r1:1.5b) — Extract the key facts about greenhouse effect from the given text, as bullets.
• Flashcard Writer (smollm2:1.7b) — Turn those facts into Q/A flashcards about greenhouse effect (JSON array of {q,a}).

Explain inflation and monetary policy at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher (qwen2.5-coder:1.5b) — Explain inflation and monetary policy like the reader is 10 years old. No jargon.
• Intermediate Teacher (qwen2.5-coder:3b) — Explain inflation and monetary policy for someone with basic background — some technical terms OK.
• Expert (llama3.2:1b) — Explain inflation and monetary policy at expert depth, precise terminology, no hand-holding.

Text about neural network backpropagation → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor (deepseek-r1:1.5b) — Extract the key facts about neural network backpropagation from the given text, as bullets.
• Flashcard Writer (smollm2:1.7b) — Turn those facts into Q/A flashcards about neural network backpropagation (JSON array of {q,a}).

Explain photosynthesis at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher (qwen2.5-coder:1.5b) — Explain photosynthesis like the reader is 10 years old. No jargon.
• Intermediate Teacher (qwen2.5-coder:3b) — Explain photosynthesis for someone with basic background — some technical terms OK.
• Expert (llama3.2:1b) — Explain photosynthesis at expert depth, precise terminology, no hand-holding.

Text about plate tectonics → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor (gemma2:2b) — Extract the key facts about plate tectonics from the given text, as bullets.
• Flashcard Writer (qwen2.5:1.5b) — Turn those facts into Q/A flashcards about plate tectonics (JSON array of {q,a}).

Text about quantum entanglement → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor (gemma2:2b) — Extract the key facts about quantum entanglement from the given text, as bullets.
• Flashcard Writer (qwen2.5:1.5b) — Turn those facts into Q/A flashcards about quantum entanglement (JSON array of {q,a}).

Text about recursion → key facts → Q/A flashcards.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Fact Extractor, Flashcard Writer. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Fact Extractor (qwen2.5-coder:3b) — Extract the key facts about recursion from the given text, as bullets.
• Flashcard Writer (llama3.2:1b) — Turn those facts into Q/A flashcards about recursion (JSON array of {q,a}).

3 agents argue about supply and demand; a judge synthesizes the strongest take.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Advocate A, Advocate B. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Advocate A (qwen2.5:1.5b) — Give the strongest argument/explanation of supply and demand, angle A.
• Advocate B (qwen2.5:3b) — Give the strongest argument/explanation of supply and demand, angle B, different framing.

3 agents argue about TCP/IP handshake; a judge synthesizes the strongest take.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Advocate A, Advocate B. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Advocate A (smollm2:1.7b) — Give the strongest argument/explanation of TCP/IP handshake, angle A.
• Advocate B (qwen2.5-coder:1.5b) — Give the strongest argument/explanation of TCP/IP handshake, angle B, different framing.

Explain the CAP theorem at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher (llama3.2:3b) — Explain the CAP theorem like the reader is 10 years old. No jargon.
• Intermediate Teacher (gemma2:2b) — Explain the CAP theorem for someone with basic background — some technical terms OK.
• Expert (qwen2.5:1.5b) — Explain the CAP theorem at expert depth, precise terminology, no hand-holding.

Explain the French Revolution causes at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher (qwen2.5:3b) — Explain the French Revolution causes like the reader is 10 years old. No jargon.
• Intermediate Teacher (deepseek-r1:1.5b) — Explain the French Revolution causes for someone with basic background — some technical terms OK.
• Expert (smollm2:1.7b) — Explain the French Revolution causes at expert depth, precise terminology, no hand-holding.

3 agents argue about the Krebs cycle; a judge synthesizes the strongest take.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Advocate A, Advocate B. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (2):

• Advocate A (smollm2:1.7b) — Give the strongest argument/explanation of the Krebs cycle, angle A.
• Advocate B (qwen2.5-coder:1.5b) — Give the strongest argument/explanation of the Krebs cycle, angle B, different framing.

Explain the water cycle at ELI5 + intermediate + expert level, in parallel.

Good for: studying a topic at your own pace — explanations, quizzes, or practice material.

Example: Say you paste in a topic or question you want to understand better. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: ELI5 Teacher, Intermediate Teacher, Expert. Use it when you want studying a topic at your own pace — explanations, quizzes, or practice material.

Steps (3):

• ELI5 Teacher (qwen2.5:3b) — Explain the water cycle like the reader is 10 years old. No jargon.
• Intermediate Teacher (deepseek-r1:1.5b) — Explain the water cycle for someone with basic background — some technical terms OK.
• Expert (smollm2:1.7b) — Explain the water cycle at expert depth, precise terminology, no hand-holding.

Messy a board update text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (qwen2.5:3b) — Extract structured line items from this a board update as a table.
• Date Finder (deepseek-r1:1.5b) — List every date/deadline mentioned in this a board update.
• Anomaly Spotter (smollm2:1.7b) — Flag anything in this a board update that looks inconsistent or wrong.

Messy a board update text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (gemma2:2b) — Extract structured line items from this a board update as a table.
• Date Finder (qwen2.5:1.5b) — List every date/deadline mentioned in this a board update.
• Anomaly Spotter (qwen2.5:3b) — Flag anything in this a board update that looks inconsistent or wrong.

Messy a client proposal text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (qwen2.5-coder:1.5b) — Summarize this a client proposal in exactly 3 bullets.
• Action Extractor (qwen2.5-coder:3b) — List only concrete action items found in this a client proposal, as bullets.
• Tone Reader (llama3.2:1b) — State the tone of this a client proposal in one sentence (e.g. urgent, neutral, frustrated).

Messy a client proposal text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (llama3.2:1b) — Summarize this a client proposal in exactly 3 bullets.
• Action Extractor (llama3.2:3b) — List only concrete action items found in this a client proposal, as bullets.
• Tone Reader (gemma2:2b) — State the tone of this a client proposal in one sentence (e.g. urgent, neutral, frustrated).

Messy a compliance checklist text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (qwen2.5-coder:1.5b) — Extract structured line items from this a compliance checklist as a table.
• Date Finder (qwen2.5-coder:3b) — List every date/deadline mentioned in this a compliance checklist.
• Anomaly Spotter (llama3.2:1b) — Flag anything in this a compliance checklist that looks inconsistent or wrong.

Rough a customer complaint response → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (deepseek-r1:1.5b) — Polish this rough a customer complaint response draft for grammar and flow.
• Keyword Pass (smollm2:1.7b) — Add/verify relevant keywords a reader or search would expect in this a customer complaint response.

Rough a customer complaint response → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (qwen2.5:1.5b) — Polish this rough a customer complaint response draft for grammar and flow.
• Keyword Pass (qwen2.5:3b) — Add/verify relevant keywords a reader or search would expect in this a customer complaint response.

Rough an executive summary → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (qwen2.5-coder:3b) — Polish this rough an executive summary draft for grammar and flow.
• Keyword Pass (llama3.2:1b) — Add/verify relevant keywords a reader or search would expect in this an executive summary.

Messy a grant application text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (llama3.2:3b) — Summarize this a grant application in exactly 3 bullets.
• Action Extractor (gemma2:2b) — List only concrete action items found in this a grant application, as bullets.
• Tone Reader (qwen2.5:1.5b) — State the tone of this a grant application in one sentence (e.g. urgent, neutral, frustrated).

Messy a grant application text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (qwen2.5-coder:3b) — Summarize this a grant application in exactly 3 bullets.
• Action Extractor (llama3.2:1b) — List only concrete action items found in this a grant application, as bullets.
• Tone Reader (llama3.2:3b) — State the tone of this a grant application in one sentence (e.g. urgent, neutral, frustrated).

Messy an internal memo text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (qwen2.5:1.5b) — Extract structured line items from this an internal memo as a table.
• Date Finder (qwen2.5:3b) — List every date/deadline mentioned in this an internal memo.
• Anomaly Spotter (deepseek-r1:1.5b) — Flag anything in this an internal memo that looks inconsistent or wrong.

Messy an internal memo text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (llama3.2:3b) — Extract structured line items from this an internal memo as a table.
• Date Finder (gemma2:2b) — List every date/deadline mentioned in this an internal memo.
• Anomaly Spotter (qwen2.5:1.5b) — Flag anything in this an internal memo that looks inconsistent or wrong.

Messy a job posting text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (deepseek-r1:1.5b) — Summarize this a job posting in exactly 3 bullets.
• Action Extractor (smollm2:1.7b) — List only concrete action items found in this a job posting, as bullets.
• Tone Reader (qwen2.5-coder:1.5b) — State the tone of this a job posting in one sentence (e.g. urgent, neutral, frustrated).

Messy a job posting text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (qwen2.5:1.5b) — Summarize this a job posting in exactly 3 bullets.
• Action Extractor (qwen2.5:3b) — List only concrete action items found in this a job posting, as bullets.
• Tone Reader (deepseek-r1:1.5b) — State the tone of this a job posting in one sentence (e.g. urgent, neutral, frustrated).

Intent → draft an onboarding checklist → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (llama3.2:1b) — Write a first draft of an onboarding checklist given the stated intent.
• Polisher (llama3.2:3b) — Polish the tone of this an onboarding checklist draft — professional, clear, no filler.
• Summarizer (gemma2:2b) — Write a one-line subject and a 2-sentence TL;DR for this an onboarding checklist.

Intent → draft an onboarding checklist → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (gemma2:2b) — Write a first draft of an onboarding checklist given the stated intent.
• Polisher (qwen2.5:1.5b) — Polish the tone of this an onboarding checklist draft — professional, clear, no filler.
• Summarizer (qwen2.5:3b) — Write a one-line subject and a 2-sentence TL;DR for this an onboarding checklist.

Rough a partnership agreement summary → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (llama3.2:1b) — Polish this rough a partnership agreement summary draft for grammar and flow.
• Keyword Pass (llama3.2:3b) — Add/verify relevant keywords a reader or search would expect in this a partnership agreement summary.

Rough a partnership agreement summary → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (qwen2.5-coder:1.5b) — Polish this rough a partnership agreement summary draft for grammar and flow.
• Keyword Pass (qwen2.5-coder:3b) — Add/verify relevant keywords a reader or search would expect in this a partnership agreement summary.

Intent → draft a performance review → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (qwen2.5:3b) — Write a first draft of a performance review given the stated intent.
• Polisher (deepseek-r1:1.5b) — Polish the tone of this a performance review draft — professional, clear, no filler.
• Summarizer (smollm2:1.7b) — Write a one-line subject and a 2-sentence TL;DR for this a performance review.

Intent → draft a performance review → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (smollm2:1.7b) — Write a first draft of a performance review given the stated intent.
• Polisher (qwen2.5-coder:1.5b) — Polish the tone of this a performance review draft — professional, clear, no filler.
• Summarizer (qwen2.5-coder:3b) — Write a one-line subject and a 2-sentence TL;DR for this a performance review.

Long a policy document → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (deepseek-r1:1.5b) — Write a 2-sentence TL;DR of this a policy document.
• Outliner (smollm2:1.7b) — Extract a section-by-section outline of this a policy document.
• Term Extractor (qwen2.5-coder:1.5b) — List key terms/jargon used in this a policy document, with one-line definitions.

Long a policy document → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (qwen2.5-coder:1.5b) — Write a 2-sentence TL;DR of this a policy document.
• Outliner (qwen2.5-coder:3b) — Extract a section-by-section outline of this a policy document.
• Term Extractor (llama3.2:1b) — List key terms/jargon used in this a policy document, with one-line definitions.

Rough a press release → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (gemma2:2b) — Polish this rough a press release draft for grammar and flow.
• Keyword Pass (qwen2.5:1.5b) — Add/verify relevant keywords a reader or search would expect in this a press release.

Rough a press release → polish → keyword pass.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Polisher, Keyword Pass. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (2):

• Polisher (qwen2.5:3b) — Polish this rough a press release draft for grammar and flow.
• Keyword Pass (deepseek-r1:1.5b) — Add/verify relevant keywords a reader or search would expect in this a press release.

Intent → draft a project charter → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (deepseek-r1:1.5b) — Write a first draft of a project charter given the stated intent.
• Polisher (smollm2:1.7b) — Polish the tone of this a project charter draft — professional, clear, no filler.
• Summarizer (qwen2.5-coder:1.5b) — Write a one-line subject and a 2-sentence TL;DR for this a project charter.

Intent → draft a project charter → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (qwen2.5-coder:1.5b) — Write a first draft of a project charter given the stated intent.
• Polisher (qwen2.5-coder:3b) — Polish the tone of this a project charter draft — professional, clear, no filler.
• Summarizer (llama3.2:1b) — Write a one-line subject and a 2-sentence TL;DR for this a project charter.

Long a risk assessment → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (smollm2:1.7b) — Write a 2-sentence TL;DR of this a risk assessment.
• Outliner (qwen2.5-coder:1.5b) — Extract a section-by-section outline of this a risk assessment.
• Term Extractor (qwen2.5-coder:3b) — List key terms/jargon used in this a risk assessment, with one-line definitions.

Intent → draft a status report → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (llama3.2:3b) — Write a first draft of a status report given the stated intent.
• Polisher (gemma2:2b) — Polish the tone of this a status report draft — professional, clear, no filler.
• Summarizer (qwen2.5:1.5b) — Write a one-line subject and a 2-sentence TL;DR for this a status report.

Intent → draft a status report → tone polish → subject/TL;DR.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Drafter, Polisher, Summarizer. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Drafter (qwen2.5-coder:3b) — Write a first draft of a status report given the stated intent.
• Polisher (llama3.2:1b) — Polish the tone of this a status report draft — professional, clear, no filler.
• Summarizer (llama3.2:3b) — Write a one-line subject and a 2-sentence TL;DR for this a status report.

Long a support ticket reply → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (llama3.2:1b) — Write a 2-sentence TL;DR of this a support ticket reply.
• Outliner (llama3.2:3b) — Extract a section-by-section outline of this a support ticket reply.
• Term Extractor (gemma2:2b) — List key terms/jargon used in this a support ticket reply, with one-line definitions.

Long a support ticket reply → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (gemma2:2b) — Write a 2-sentence TL;DR of this a support ticket reply.
• Outliner (qwen2.5:1.5b) — Extract a section-by-section outline of this a support ticket reply.
• Term Extractor (qwen2.5:3b) — List key terms/jargon used in this a support ticket reply, with one-line definitions.

Messy a training outline text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (qwen2.5:3b) — Summarize this a training outline in exactly 3 bullets.
• Action Extractor (deepseek-r1:1.5b) — List only concrete action items found in this a training outline, as bullets.
• Tone Reader (smollm2:1.7b) — State the tone of this a training outline in one sentence (e.g. urgent, neutral, frustrated).

Messy a training outline text → 3-bullet summary + action items + tone read, in parallel.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Summarizer, Action Extractor, Tone Reader. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Summarizer (smollm2:1.7b) — Summarize this a training outline in exactly 3 bullets.
• Action Extractor (qwen2.5-coder:1.5b) — List only concrete action items found in this a training outline, as bullets.
• Tone Reader (qwen2.5-coder:3b) — State the tone of this a training outline in one sentence (e.g. urgent, neutral, frustrated).

Messy a user manual section text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (smollm2:1.7b) — Extract structured line items from this a user manual section as a table.
• Date Finder (qwen2.5-coder:1.5b) — List every date/deadline mentioned in this a user manual section.
• Anomaly Spotter (qwen2.5-coder:3b) — Flag anything in this a user manual section that looks inconsistent or wrong.

Messy a user manual section text → structured line items + totals/dates + anomalies.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Extractor, Date Finder, Anomaly Spotter. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• Extractor (qwen2.5-coder:3b) — Extract structured line items from this a user manual section as a table.
• Date Finder (llama3.2:1b) — List every date/deadline mentioned in this a user manual section.
• Anomaly Spotter (llama3.2:3b) — Flag anything in this a user manual section that looks inconsistent or wrong.

Long a vendor contract summary → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (qwen2.5:1.5b) — Write a 2-sentence TL;DR of this a vendor contract summary.
• Outliner (qwen2.5:3b) — Extract a section-by-section outline of this a vendor contract summary.
• Term Extractor (deepseek-r1:1.5b) — List key terms/jargon used in this a vendor contract summary, with one-line definitions.

Long a vendor contract summary → TL;DR + outline + key terms + open questions.

Good for: turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Example: Say you paste in a paragraph of meeting notes or a rough draft. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: TL;DR Writer, Outliner, Term Extractor. Use it when you want turning rough notes, meeting audio, or spreadsheets into clean, usable documents.

Steps (3):

• TL;DR Writer (llama3.2:3b) — Write a 2-sentence TL;DR of this a vendor contract summary.
• Outliner (gemma2:2b) — Extract a section-by-section outline of this a vendor contract summary.
• Term Extractor (qwen2.5:1.5b) — List key terms/jargon used in this a vendor contract summary, with one-line definitions.

HTML/UI → color contrast issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (gemma-4-e2b-it-qat-mtp) — Scan this HTML/UI text ONLY for color contrast issues. List concrete problems as bullets, cite the element.
• Fix Suggester (lfm2.5-1.2b-instruct) — For each color contrast issue found, suggest a concrete fix (markup/attribute/CSS change).

HTML/UI → focus indicators issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (lfm2.5-1.2b-instruct) — Scan this HTML/UI text ONLY for focus indicators issues. List concrete problems as bullets, cite the element.
• Fix Suggester (qwen_qwen3.5-0.8b) — For each focus indicators issue found, suggest a concrete fix (markup/attribute/CSS change).

HTML/UI → keyboard navigation issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (gemma-3-1b-it) — Scan this HTML/UI text ONLY for keyboard navigation issues. List concrete problems as bullets, cite the element.
• Fix Suggester (gemma-3-4b-it) — For each keyboard navigation issue found, suggest a concrete fix (markup/attribute/CSS change).

HTML/UI → screen-reader labeling issues found + concrete fixes.

Good for: making content easier to read, navigate, or use for people with different needs.

Example: Say you paste in a page of text or a UI flow you want checked. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: A11y Scanner, Fix Suggester. Use it when you want making content easier to read, navigate, or use for people with different needs.

Steps (2):

• A11y Scanner (gemma-3-4b-it) — Scan this HTML/UI text ONLY for screen-reader labeling issues. List concrete problems as bullets, cite the element.
• Fix Suggester (gemma-4-e2b-it-qat-mtp) — For each screen-reader labeling issue found, suggest a concrete fix (markup/attribute/CSS change).

Generate Bash unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (gemma-3-1b-it) — Write Bash unit tests for the normal/expected inputs only.
• Edge Cases (gemma-3-4b-it) — Write Bash unit tests for boundary/edge-case inputs only.
• Failure Modes (gemma-4-e2b-it-qat-mtp) — Write Bash unit tests for invalid input / error conditions only.

Generate Bash unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (gemma-4-e2b-it-qat-mtp) — Write Bash unit tests for the normal/expected inputs only.
• Edge Cases (lfm2.5-1.2b-instruct) — Write Bash unit tests for boundary/edge-case inputs only.
• Failure Modes (qwen_qwen3.5-0.8b) — Write Bash unit tests for invalid input / error conditions only.

Two C# coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (gemma-3-1b-it) — Solve the given task in C#, straightforward/imperative style. Code only.
• Coder B (gemma-3-4b-it) — Solve the given task in C#, a more functional/idiomatic style. Code only.

Two C# coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (qwen_qwen3.5-0.8b) — Solve the given task in C#, straightforward/imperative style. Code only.
• Coder B (qwen_qwen3.5-2b) — Solve the given task in C#, a more functional/idiomatic style. Code only.

Two C# coders solve the same task differently; a judge picks the cleaner one.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs several steps in parallel, then a final judge step reads all of them and picks or merges the best answer. The specialists involved are: Coder A, Coder B. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Coder A (gemma-4-e2b-it-qat-mtp) — Solve the given task in C#, straightforward/imperative style. Code only.
• Coder B (lfm2.5-1.2b-instruct) — Solve the given task in C#, a more functional/idiomatic style. Code only.

Refactor C++ code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (qwen_qwen3.5-2b) — Rewrite the given C++ code for clarity and idiom. Output only the revised code.
• Verifier (gemma-3-1b-it) — Sanity-check the refactored C++ code against the original — same behavior? List any risk in 3 bullets max.

Refactor C++ code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (lfm2.5-1.2b-instruct) — Rewrite the given C++ code for clarity and idiom. Output only the revised code.
• Verifier (qwen_qwen3.5-0.8b) — Sanity-check the refactored C++ code against the original — same behavior? List any risk in 3 bullets max.

Refactor C++ code for clarity, then verify nothing broke.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Refactorer, Verifier. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Refactorer (gemma-3-4b-it) — Rewrite the given C++ code for clarity and idiom. Output only the revised code.
• Verifier (gemma-4-e2b-it-qat-mtp) — Sanity-check the refactored C++ code against the original — same behavior? List any risk in 3 bullets max.

App description (Dart) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author (gemma-3-1b-it) — Write a production-sane Dockerfile for this Dart app description.
• Notes Writer (gemma-3-4b-it) — Write 3-5 short run/build notes for the Dockerfile above.

App description (Dart) → Dockerfile → run notes.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Dockerfile Author, Notes Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (2):

• Dockerfile Author (gemma-4-e2b-it-qat-mtp) — Write a production-sane Dockerfile for this Dart app description.
• Notes Writer (lfm2.5-1.2b-instruct) — Write 3-5 short run/build notes for the Dockerfile above.

Error/log in Elixir → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (gemma-4-e2b-it-qat-mtp) — Given a Elixir error/log, state the most likely root cause in 2-3 sentences.
• Fixer (lfm2.5-1.2b-instruct) — Propose a concrete Elixir code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (qwen_qwen3.5-0.8b) — Write one Elixir regression test that would have caught this bug.

Error/log in Elixir → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (qwen_qwen3.5-0.8b) — Given a Elixir error/log, state the most likely root cause in 2-3 sentences.
• Fixer (qwen_qwen3.5-2b) — Propose a concrete Elixir code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (gemma-3-1b-it) — Write one Elixir regression test that would have caught this bug.

Generate Go unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (gemma-4-e2b-it-qat-mtp) — Write Go unit tests for the normal/expected inputs only.
• Edge Cases (lfm2.5-1.2b-instruct) — Write Go unit tests for boundary/edge-case inputs only.
• Failure Modes (qwen_qwen3.5-0.8b) — Write Go unit tests for invalid input / error conditions only.

Generate Go unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (gemma-3-1b-it) — Write Go unit tests for the normal/expected inputs only.
• Edge Cases (gemma-3-4b-it) — Write Go unit tests for boundary/edge-case inputs only.
• Failure Modes (gemma-4-e2b-it-qat-mtp) — Write Go unit tests for invalid input / error conditions only.

Generate Go unit tests: happy path, edge cases, and failure modes, in parallel.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs every step at the same time on the same input, then shows you all the answers side by side so you can compare them. The specialists involved are: Happy Path, Edge Cases, Failure Modes. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Happy Path (qwen_qwen3.5-0.8b) — Write Go unit tests for the normal/expected inputs only.
• Edge Cases (qwen_qwen3.5-2b) — Write Go unit tests for boundary/edge-case inputs only.
• Failure Modes (gemma-3-1b-it) — Write Go unit tests for invalid input / error conditions only.

Error/log in Java → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (qwen_qwen3.5-0.8b) — Given a Java error/log, state the most likely root cause in 2-3 sentences.
• Fixer (qwen_qwen3.5-2b) — Propose a concrete Java code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (gemma-3-1b-it) — Write one Java regression test that would have caught this bug.

Error/log in Java → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (gemma-4-e2b-it-qat-mtp) — Given a Java error/log, state the most likely root cause in 2-3 sentences.
• Fixer (lfm2.5-1.2b-instruct) — Propose a concrete Java code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (qwen_qwen3.5-0.8b) — Write one Java regression test that would have caught this bug.

Error/log in Java → root cause → fix → a test to prevent regression.

Good for: reviewing, debugging, or documenting code without sending it to a cloud API.

Example: Say you paste in a function or file you're working on. This combo runs each step in order, feeding the previous step's output into the next — like an assembly line. The specialists involved are: Diagnoser, Fixer, Test Writer. Use it when you want reviewing, debugging, or documenting code without sending it to a cloud API.

Steps (3):

• Diagnoser (gemma-3-1b-it) — Given a Java error/log, state the most likely root cause in 2-3 sentences.
• Fixer (gemma-3-4b-it) — Propose a concrete Java code fix for the diagnosed cause. Code only, minimal diff.
• Test Writer (gemma-4-e2b-it-qat-mtp) — Write one Java regression test that would have caught this bug.