-
Notifications
You must be signed in to change notification settings - Fork 71
Open
Description
AI coding assistants (Claude Code, Cursor, Copilot, etc.) working with Mellea could produce code that doesn't follow project patterns - wrong import paths, missing type hints, tests without proper markers.
There's a convention called AGENTS.md that provides structured guidance for AI assistants, similar to how .editorconfig helps editors and CONTRIBUTING.md helps human contributors.
References:
Why this would help:
- AI assistants would use @Generative instead of raw OpenAI calls
- Correct import paths (mellea.stdlib.requirements not mellea.stdlib.requirement)
- Tests get proper markers (qualitative, integration)
- Code follows project conventions (types, docstrings, ... in generative bodies)
What's needed:
- AGENTS.md in repo root for contributors working on Mellea internals
- docs/AGENTS_TEMPLATE.md that downstream projects can copy when building apps with Mellea
I have a PR I'll propose which it would be great to get feedback on.
Metadata
Metadata
Assignees
Labels
No labels