mirror of
https://github.com/meta-llama/llama-stack.git
synced 2025-12-03 18:00:36 +00:00
Address all feedback from PR #3962: **Code Quality Improvements:** - Rename `_uvicorn_run` → `_run_server` for accurate method naming - Refactor error handling: move Gunicorn fallback logic from `_run_with_gunicorn` to caller - Update comments to reflect both Uvicorn and Gunicorn behavior - Update test mock from `_uvicorn_run` to `_run_server` **Environment Variable:** - Change `LLAMA_STACK_DISABLE_GUNICORN` → `LLAMA_STACK_ENABLE_GUNICORN` - More intuitive positive logic (no double negatives) - Defaults to `true` on Unix systems - Clearer log messages distinguishing platform limitations vs explicit disable **Documentation:** - Remove unnecessary `uv sync --group unit --group test` from user docs - Clarify SQLite limitations: "SQLite only allows one writer at a time" - Accurate explanation: WAL mode enables concurrent reads but writes are serialized - Strong recommendation for PostgreSQL in production with high traffic **Architecture:** - Better separation of concerns: `_run_with_gunicorn` just executes, caller handles fallback - Exceptions propagate to caller for centralized decision making 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| test_stack_config.py | ||