v0.5.0

Changed

• --help now prints usage: log-ai (argparse prog="log-ai"), matching the installed
  console script name; pinned by a new test (104 total).

v0.4.0

Changed

• PyPI distribution renamed to agent-log-ai1 (PyPI's name-similarity guard rejects agent-log-ai; the trusted-publisher pending project is agent-log-ai1). GitHub repo name, the log-ai console script, and single-file adoption are unchanged.

v0.3.0

Added

• Professional packaging: pyproject.toml (version derived from the git tag via
  setuptools-scm - no drift), a global log-ai console script, zero runtime
  dependencies.
• Installed-mode defaults guard (_default_base): pip-installed runs resolve
  default paths against the current directory; in-place copies keep resolving
  against the file's folder.
• CI packaging job: builds the wheel, installs it into a fresh venv, and
  smoke-tests the console script + module import.
• publish.yml: trusted publishing to PyPI, gated behind the PUBLISH_TO_PYPI
  repository variable (skipped until enabled).

Fixed

• Typed entries: parse_entries() now returns LogEntry dataclasses (dict-compatible via __getitem__/get), full type hints on all functions, and an exception vocabulary (AgentLogError / ValidationError) — same behavior, same exit codes.
• stdin reconfigured to UTF-8 on Windows: piped unicode prompts no longer double-encode (stdout-only reconfigure bug).

v0.2.0

Fixed

• L10: load() no longer crashes on a locked/unreadable log file (graceful OSError fallback; regression tests added).

v0.1.0

Added

• Project scaffold: README, AGENTS.md, LICENSE, community files
  (CONTRIBUTING, SECURITY, CODE_OF_CONDUCT), issue/PR templates, and the
  release workflow.
• Repository created on GitHub (vartiainen1/agent-log-ai).
• check_logs_ai.py — the LLM reasoning layer: --lessons (error-log
  root-cause drafts), --review (decision-log reversal analysis),
  --notes (session-note drafts), --check (endpoint ping), --dry-run
  (prompt preview, sends nothing), and --check-commit (CI gate). Stdlib
  only (urllib), local-first (Ollama default), prompt-injection guard.
• _test_logs_ai.py — 82 offline tests (HTTP mocked, no network, no key).
• _check_readme_count.py — README test-count drift guard.
• .github/workflows/ci.yml — CI: tests + linter matrix (Ubuntu + Windows,
  Python 3.9/3.11/3.12, offline HTTP-mocked suite), README drift guard, and
  the commit-message gate (AREA marker must name a logged decision).
• start.py — boot briefing: sibling-log health (errors + decisions, if the
  sibling repos live next door), a 2-second endpoint reachability probe, the
  command cheat-sheet, and the latest session note. Uses the canonical
  check_logs_ai helpers (never reimplements parsing).
• Removed the broken start.bat launcher (it referenced a start.py that
  never existed).
• --init [DIR] — one-command adoption: scaffolds errors.txt,
  decisions.txt, rules.txt (with the §7 LESSONS header), and notes.txt
  from built-in templates; never overwrites existing files; runs the offline
  self-test suite when present.
• The reasoning commands are now mutually exclusive (--lessons / --review /
  --notes / --check / --check-commit / --init) — previously two could
  be passed and one silently won.
• chat() now explicitly re-raises KeyboardInterrupt — Ctrl-C always
  escapes and is never folded into an error string.
• _test_logs_ai.py — 93 tests (+11): _token_warn threshold, --timeout
  pass-through, KeyboardInterrupt propagation, --max-entries truncation,
  --apply through the command path, mutual exclusion, and four --init
  cases (scaffold, no-overwrite, self-test run, failing self-test fails
  adoption).