You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
No more hardcoded default model — on first run, oy prompts you to pick and save a model from available backends. In non-interactive mode it aborts with clear instructions.
Recommended model — glm-5 (via Bedrock) offers the best balance of intelligence, cost, and tool-use ability from testing. kimi-k2.5 is another strong option.
Single version source — version is now read from pyproject.toml via importlib.metadata, no more duplication.
More tests — 34 tests (up from 15), covering path traversal, text replacement, token clipping, model selection, JSON loading, model spec parsing.
Docstrings — added to key public and internal functions across both modules.
Dev tooling — pytest added to dev deps, mise run test task, pytest config in pyproject.toml.
Docs updated — README describes first-run model selection and recommended models; CONTRIBUTING.md documents PEP 758 except syntax and test commands; ISSUES.md updated with resolved items.