v0.16.2
Highlights
- Added richer summary support and metadata: daemon/glob summaries, a-priori compression effect reporting, and prompt/docs guidance that tells agents when to summarize before vs after bulky results.
- Split info/manual signpost surfaces and updated progressive-disclosure contracts so tools can expose runtime health separately from bundled manuals.
- Improved prompt/runtime hygiene: adapt-from-evidence learning principle, timely transient
_metafiltering, prompt section contracts, and retention-footprint reporting. - Expanded daemon/MCP capability paths: Kimi native MCP config, parent MCP propagation into CLI backends, daemon architecture capability contracts, Codex pool provider/preset support, and external skill intake docs.
- Hardened IM/notification context: persistent Telegram notification/reply context, LICC notification contract docs, and cleaner transient hook behavior.
- Release-prep cleanup aligned stale tests with current
.gitignoresecret/transient exclusions and the knowledgeinfo/manualaction schema.
Validation
- ✅
git diff --check v0.16.1...HEAD - ✅
python -m compileall -q src tests - ✅
PYTHONPATH=src python -m pytest -q→3798 passed, 4 skippedin322.83s - ✅
python -m build - ✅
python -m twine check dist/*
PyPI artifacts prepared
| Artifact | Size | SHA256 |
|---|---|---|
lingtai-0.16.2-cp312-cp312-macosx_11_0_arm64.whl |
2,469,350 bytes | c650935ecbc018b381759af7fbbec38068bf64d7e305442298fcb5da3750f7b2 |
lingtai-0.16.2.tar.gz |
1,777,985 bytes | dc9ee90e7fd947f835aac693e940224b637a50e25a6be4a8633d0e415d16cb34 |
Contributors
Compare
Contributors
TZZheng and huangzesen