Improve Windmill’s AI chat by reviewing its tools, prompts, result shapes, and context usage. The guidance helps place shared instructions where chat and CLI can use them, reduce repeated token costs, and compare behavior before and after changes with A/B evaluations.
Guidance for improving the Windmill AI chat (copilot), especially global mode — tools, prompts, and context-window discipline. Use when editing chat tools, system prompts, or tool-result shapes under frontend/src/lib/components/copilot/chat, or when changing how the chat manages its context window.
Run score Not tested
See how to installReview the instructions the agent will receive before you install.
No context or behavior change ships without an ai_evals A/B on the affected mode.
Add or adjust cases for exactly what you changed — see the ai-evals skill for
authoring and the full run reference.
Run the affected mode before your change and after, same model(s), same cases.
What the model should know about Windmill itself (flow shapes, groups, data tables, app
access, secrets…) belongs in system_prompts/base/ or system_prompts/languages/, which also
feed the CLI's skills — not in a */core.ts prompt builder, where only this chat would see it.
The builders add tool plumbing and runtime values only. Scope a sentence to one consumer with
<!-- chat-only --> / <!-- cli-only -->, keep chat tool names out of the shared files, and
rerun python system_prompts/generate.py. system_prompts/README.md has the mechanics.
Optimize finalContextTokens (window occupancy — what drives overflow and
compaction), then cumulative prompt tokens.
The dominant fixed cost is per-iteration overhead: the system prompt plus every tool schema is re-sent on every loop iteration. So:
{ success, message } instead. When you touch a
shared write helper (e.g. finishAppDraftWrite in global/core.ts), re-check
this invariant for all the write tools routing through it — the echo has
regressed before via a shared refactor.The system prompt and tool descriptions steer behavior as much as the tools themselves, and are benchmarkable the same way. A description that advertises truncation makes the model self-limit; the path-conventions block changes where drafts land. Treat prompt/description edits as real changes and A/B them — a pure-prompt change is a legitimate, measurable improvement.
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Plan the code changes needed to add a TriggerCrud-based trigger to Windmill. The checklist covers backend, database, API, CLI, frontend, capture, and validation work so the new trigger can be managed across the product.
Create and run black-box benchmarks for Windmill AI generation modes such as flows, apps, scripts, CLI, and global requests. Each case exercises the production path, checks hard requirements, and uses an LLM judge to assess user-facing behavior before and after changes.
Use this reference to work with Windmill resources from the command line, including scripts, flows, datatables, and workspace settings. It also covers listing jobs, inspecting details and logs, retrieving results, and rerunning or restarting executions when investigating failures.
Use this skill to shape modules with small interfaces and substantial behavior behind them. It helps you decide where a seam belongs, reduce complexity exposed to callers, and make code easier to test and maintain.
Use this skill to review pending changes, select only the relevant files for staging, and create a focused Git commit. It formats the single-line message with a Conventional Commits type such as feat, fix, or docs, and checks the resulting status.
Manage schema changes to Windmill data tables through versioned SQL migrations, including applying and rolling them back. The guide also explains role-based access, migration safety, and how changes affect workspaces, apps, and scripts.