Release v0.10.0
The context layer for production AI agents: local-first context planning, compression and runtime for LLM apps.
Status: Beta
Brought to you by:
yashvt
Originally created by: Thanatos9404
Release prep for llmslim 0.10.0, approved in chat. After merge I tag v0.10.0 on main and publish the GitHub release. release.yml then builds and uploads to PyPI with attestations, and I verify a clean pip install llmslim==0.10.0.
compress_chat_messages no longer cause HTTP 400s on Claude models with preserved thinking (#60).llmslim.memory: agent memory with no LLM in the loop, with Mem0-compatible method names (#56).llmslim mcp: memory and context tools for any MCP client (#58).llmslim serve: memory and context over HTTP, with a Dockerfile (#59).clients/typescript (#61). They're not on npm yet.pyproject.toml, llmslim.__version__, CITATION.cff, the README and tests/test_dx.py.[0.10.0] - 2026-10-17.docs/MCP.md and docs/SERVER.md install from PyPI;uv build produces llmslim-0.10.0-py3-none-any.whl and llmslim-0.10.0.tar.gz; twine check passes on both.112 -> 72 tokens (35.7% fewer));llmslim.memory recalls "I moved to Pune." for "Where did I move to?";llmslim serve --new-key works.[server,mcp]: this installs starlette 1.8.0, uvicorn 0.54.0, mcp 2.3.0 and tiktoken 0.14.0. The HTTP app and an in-process MCP client both recall the stored turn.==0.10.0 once the package is on PyPI, as for 0.9.0 (#55).server.json is ready. It needs mcp-publisher login with your GitHub account.llmslim: needs npm login on your side.
Originally posted by: vercel[bot]
The latest updates on your projects. Learn more about Vercel for GitHub.
Ticket changed by: Thanatos9404