Phuriwaj

LiteLLM — Configuring Local LLM and MiniMax

LiteLLM — Configuring Local LLM and MiniMax

How to add both a local Ollama model and the MiniMax cloud API to a single LiteLLM config.yaml.

Why / When to Use

Use when you want LiteLLM proxy to serve multiple backends (local inference + cloud API) under one unified OpenAI-compatible endpoint.

Core Config

Ollama (local LLM)

model_list:
  - model_name: my-local-llama
    litellm_params:
      model: ollama/llama3.1:8b
      api_base: http://localhost:11434

Requires ollama serve to be running before starting the proxy.

LM Studio or other OpenAI-compatible local server

model_list:
  - model_name: my-local-model
    litellm_params:
      model: openai/your-model-name
      api_base: http://localhost:1234/v1
      api_key: none   # required field but value doesn't matter

MiniMax (cloud API)

model_list:
  - model_name: minimax-m2-5
    litellm_params:
      model: minimax/MiniMax-Text-01
      api_key: os.environ/MINIMAX_API_KEY

Set the env var before starting the proxy:

export MINIMAX_API_KEY=your-minimax-api-key
litellm --config config.yaml

Combined config (local + MiniMax)

model_list:
  - model_name: my-local-llama
    litellm_params:
      model: ollama/llama3.1:8b
      api_base: http://localhost:11434
 
  - model_name: minimax-m2-5
    litellm_params:
      model: minimax/MiniMax-Text-01
      api_key: os.environ/MINIMAX_API_KEY

Testing

# Test local model
curl http://0.0.0.0:4000/chat/completions \
  -H "Authorization: Bearer sk-your-proxy-key" \
  -d '{"model": "my-local-llama", "messages": [{"role": "user", "content": "Hello"}]}'
 
# Test MiniMax
curl http://0.0.0.0:4000/chat/completions \
  -H "Authorization: Bearer sk-your-proxy-key" \
  -d '{"model": "minimax-m2-5", "messages": [{"role": "user", "content": "Hello"}]}'

Gotchas

  • Ollama must be running (ollama serve) before LiteLLM starts; otherwise requests to that model fail immediately.
  • api_key: none is still required for OpenAI-compatible local servers — omitting it causes a validation error.
  • MiniMax model name in litellm_params uses minimax/ prefix; the model_name alias can be anything.

Source

Conversation: “Configuring local LLM and Minimax on LiteLLM” — 2026-05-29