{"repo":"Productive-Superintelligence/lllm","free":true,"listed":false,"github":"https://github.com/Productive-Superintelligence/lllm","clone":"git clone https://github.com/Productive-Superintelligence/lllm.git","description":"A light-weight, modular, and high-performance framework for building llm agentic systems.","language":"Python","stars":99,"topics":["agent","agentic-ai","agents","ai","anthropic","framework","gemini","llm","multiagent","open-source"],"license":"Apache-2.0","category":"ai-agents","readme_excerpt":"Low-Level Language Models (LLLM) Low-Level Language Models (LLLM) is a small protocol and service layer for reusable agentic tactics. The center model is Tactic : a typed, runtime-agnostic unit that does one thing well and can be called locally, exposed through FastAPI, described for a PsiHub package, and composed later through refs and local config. Install The PyPI distribution is lllm-core ; the import package remains lllm : For local development: Smallest Tactic Serve It Remote clients normalize base service URLs into /run and /stream endpoints. RemoteTactic.arun() calls the JSON run endpoint, while RemoteTactic.astream() consumes the service-sent event stream and yields the same raw data items as local Tactic.astream() . Use RemoteTactic.aevents() when you need the full TacticEvent envelopes. RemoteTactic.fetch info() and RemoteTactic.afetch info() retrieve the service-advertised TacticInfo from /info without making local info() do network I/O. Pydantic AI Pydantic AI remains the runtime owner. Configure the agent normally, then wrap it: LLLM forwards request metadata where the agent run method accepts metadata . examples/pydantic ai tactic/structured agent.py shows structured input/output, streaming, and tool wrapping with an offline fake agent. examples/pydantic ai tactic/surrounding features.py shows that normal runtime-owned kwargs such as model settings, deps, eval hooks, durable IDs, graph/workflow state, and tool approval pass through the wrapper. Live provider cr","default_branch":null,"files":null,"tree":[],"storefront":"/r/Productive-Superintelligence","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/Productive-Superintelligence/lllm/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}