Skip to content

LingTai kernel v0.14.1

Choose a tag to compare

View all tags
@huangzesen huangzesen released this 23 Jun 08:45
v0.14.1
d47473b
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

LingTai kernel v0.14.1

Patch release v0.14.1 is a Codex/runtime stability release on top of v0.14.0. It keeps the kernel package version moving after the Responses/WebSocket cache work landed on main and before the release blogs/docs follow-up.

Highlights

  • Codex Responses WebSocket stability: persistent WebSocket reuse now supports incremental previous_response_id continuation, tool-output freezing, baseline restore, and explicit fresh-epoch resets.
  • Codex cache hygiene: LingTai can reset stale Codex remote state on schedule or after local summarize, while the runtime comment now warns that repeated summarize calls force ws_full and should be batched for Codex.
  • Honest Codex identity: default ChatGPT Codex requests identify as LingTai (originator=lingtai, User-Agent=LingTai/<version>); the official CLI-shaped identity remains only an explicit local comparison switch.
  • Daemon/runtime notification durability: daemon terminal-state notifications and tool-result/runtime metadata paths were tightened after the v0.14.0 release line.
  • ANATOMY freshness: Codex identity comments and kernel ANATOMY citations were refreshed, with a full local ANATOMY citation pass at 599 citations / 0 issues.

Compatibility notes

  • No user-facing install command changes are required for existing LingTai TUI-managed projects; this is the kernel/runtime package version used by project venv dependency resolution.
  • Codex users may see ws_full after a summarize-triggered fresh epoch. That is expected: the next request is rebuilt from local history instead of continuing the previous remote chain.
  • For Codex specifically, avoid several one-by-one summarize calls in quick succession; group ordinary long-result summaries when possible.

Validation

  • python -m pytest -q → 2715 passed, 4 skipped in 304.41s
  • python -m py_compile src/lingtai/llm/openai/adapter.py
  • python -m pytest -q tests/test_codex_ws_session.py::test_codex_adapter_comment_explains_epoch_reset_and_summarize_delay tests/test_codex_prompt_cache_key.py → 42 passed
  • custom all-ANATOMY citation checker → checked=599, issues=0
  • python -m build
  • python -m twine check dist/* → passed for wheel and sdist
  • artifact inspection: __pycache__ / .pyc entries = 0 for both wheel and sdist
  • git diff --check

Artifact SHA-256

  • 952be6499f75df3f83624276c3b9adb0ade8a8862f3f1f57575e0d9a148f7321 dist/lingtai-0.14.1-cp312-cp312-macosx_11_0_arm64.whl
  • 9574b9a81f0deb673e71fa090ba925aa3f05f84c2dbf9ccafa040e1381e756a3 dist/lingtai-0.14.1.tar.gz
Assets 4
Loading