LeanLLM brings a unified LiteLLM-style LLM client to the Lean 4 ecosystem
hargup13 · x · 2026-10-14
Theoric Labs released LeanLLM — essentially LiteLLM for Lean 4 — a unified, typed LLM client library backed by LeanHttp.
- Broad coverage: Claude, OpenAI, Gemini, DeepSeek, Kimi/Moonshot, Z.AI GLM, Qwen, MiniMax, Grok and Mistral models, across direct vendor APIs, subscription OAuth sessions (Claude Code, OpenAI Codex/ChatGPT, Gemini Code Assist, Antigravity, GitHub Copilot, Grok Build), OpenRouter, AWS Bedrock, Azure OpenAI, Vertex AI, and local OpenAI-compatible servers.
- Full SDK feature set: tool calling, structured outputs, image inputs, streaming, all behind one typed interface (Messages, ModelConfig, Tools).
- Requires Lean v4.34.1 with libcurl loaded dynamically at runtime. Open source on GitHub, as the author puts it: 'piece by piece we're building the whole ecosystem in Lean.'
More from coding & agent
- One-shot flying dragon: Gemini 4 Argon with Blender MCP and Tripo 3D builds a pure .blend scene — sidahuj · 2026-10-14
- Llamaforge v0.16: a control panel for upstream llama.cpp with local agents whose quotes are code-verified — Sleepybear2611 · 2026-10-14
- hexvero: a binary-analysis MCP server with a 3k-token tool catalog vs ~510k for the incumbent — ImaginationVarious30 · 2026-10-14
- Questioning OpenAI's Decisions API: are cloud round-trip latency and privacy viable in production? — WhichCardiologist800 · 2026-10-14
- Factory engineer: third-party agent harnesses can beat Claude Code and Codex even on their own models — matanSF · 2026-10-14
- teamai unifies Skills, rules and MCP configs for teams juggling multiple AI coding tools — sujingshen · 2026-10-14