BigHugger
GH Repository · algorithmicsuperintelligence

optillm

Optimizing inference proxy for LLMs

stars
4,263
30-day movement
+21/day
Related entries
60
Connections
2
dockeragent-appprompt-engineeringchain-of-thoughtagentic-frameworkPythonoptimizationopenai-apillm-inferencemoaopenaiagentic-aiagentic-workflowproxy-serveragentspythonllmapi-gatewaymixture-of-expertsmonte-carlo-tree-searchgenailarge-language-modelsagentllmapi

optillm is an optimizing inference proxy for LLMs that sits between clients and OpenAI-compatible APIs, applying inference-time techniques to requests. Its topics indicate support for approaches like chain-of-thought, Monte Carlo tree search, mixture-of-agents, and mixture-of-experts, and it ships with a Dockerfile and Claude project context file.

You can point existing OpenAI-compatible clients at it to get optimization techniques applied to your LLM calls without changing your application code.

Use it to

  • Proxy OpenAI-compatible LLM requests through inference optimization techniques
  • Apply chain-of-thought or Monte Carlo tree search at inference time
  • Experiment with mixture-of-agents and mixture-of-experts inference
  • Route LLM traffic through a single API gateway
  • Compare optimization strategies against a standard inference baseline

For Developers running LLM applications who want inference-time optimization

Role
agent-app
Language
Python
Licence
Apache-2.0
Forks
383
Open issues
19
Last push
2026-07-18
Latest release
v0.0.1 · 2024-10-24
topicsllm-inferenceproxy-serveroptimizationchain-of-thoughtopenai-apimixture-of-experts