Skip to content

changelog.md

Latest commit

 

History

History
80 lines (60 loc) · 4.14 KB

changelog.md

File metadata and controls

80 lines (60 loc) · 4.14 KB

Changelog

v0.2.0

Highlights

  • Windows arm64 binary now available (in addition to Windows x64, macOS x64/arm64, Linux x64/arm64).
  • Docs polish: all badges use flat-square with clear, consistent labels.

Added

  • Windows arm64 build artifact and README download button.
  • Detailed release notes at docs/release-notes/v0.2.0.md.

Changed

  • Consistent Shields.io badge styling and labels across README and development docs.
  • Bumped version to 0.2.0 in pyproject.toml.

Fixed

  • Minor inconsistencies in badge labels and links.

Notes

  • No breaking changes; config schema unchanged.

v0.1.3

Added

  • Release notes for this iteration (docs/release-notes/v0.1.3.md) highlighting the logging shim fix and broadened QA surface.
  • High-signal regression tests that exercise the doctor feature probe, asynchronous HTTP logging queue, keychain pass backend flows, update-check entrypoints, and prompt detector helper signatures.

Changed

  • The compatibility shim now mirrors src into PYTHONPATH before spawning subprocesses so doctor logging checks and other child processes load the in-repo modules deterministically.
  • Homebrew formula and Scoop manifest now point to the v0.1.3 artifacts; Scoop URL updated pending the freshly built Windows binary hash.
  • Documentation ordering in the changelog cleaned up to keep the latest version at the top for easier scanning.

Fixed

  • Remote logging subprocesses no longer crash with ModuleNotFoundError, restoring the doctor feature probe when PYTEST_CURRENT_TEST is unset.
  • Update-check integration tests reuse the CLI shim module so type exports stay in sync with release packaging.

Maintenance

  • Bumped project metadata to 0.1.3, regenerated sdist/wheel via python -m build, and verified formatting (ruff, black) plus the full test suite.
  • Ensured release packaging artefacts (sdist, wheel) incorporate the extended tests and shim adjustments.

v0.1.2

Highlights

  • Add ARM builds for Linux and macOS in Release workflow
  • Optional keychain backend via --keychain (macOS Keychain, Windows DPAPI, Linux Secret Service when available)
  • Structured logging fields + --log-level/--level, --log-remote buffered async sink with drop-oldest policy
  • Safer I/O: atomic writes with fsync + os.replace; .bak only when target existed
  • Dry-run diffs (--dry-run --diff) and post-run summary (target, backup, profile, provider, model, context/max tokens)
  • Extended local server probes (vLLM, Text-Gen-WebUI OpenAI plugin, TGI shim, OpenRouter local)
  • Model selection accepts substrings with deterministic tie-break; --yes non-interactive option
  • Multi-provider profiles via --providers (lmstudio, ollama, vllm, tgwui, tgi, openrouter, jan, llamafile, gpt4all, local)
  • Windows: clear_screen disabled by default; --clear to enable
  • Docs: badges with links, coverage badge, multiple Quick Starts, Dockerfile + compose, full short-flag map; --keychain documented
  • Minimal refactor: add cli.py, detect.py, render.py, io_safe.py, spec.py facades

Fixes and maintenance

  • Lint and format clean-up (ruff, black)
  • CI guardrails and versioning remain in place

All notable changes to this project are documented here. The project follows Semantic Versioning and a lightweight variant of Keep a Changelog.

[0.1.1] - 2025-09-07

Added

  • GitHub Actions workflows for tagged releases: build PyInstaller binaries and publish to PyPI via Trusted Publishing.
  • Contributor guidance in agents.md.

Changed

  • Code style cleanups (ruff + black) and test suite lint cleanup.
  • Relaxed coverage gate in CI (removed --cov-fail-under=100), retained overall timeout.

Fixed

  • Logging reliability improvements: prevent duplicate handlers, escape JSON log messages, and ensure HTTPHandler import for remote logging.

[0.1.0] - 2025-08-31

Added

  • Initial release of codex-cli-linker: single-file, no-deps tool to auto-detect local servers (LM Studio, Ollama), pick a model, and emit Codex config in TOML (optionally JSON/YAML). Includes safe backups and a small persisted state file.