# LiteLLM Proxy Configuration — Molecule AI # ───────────────────────────────────────────────────────────────────────────── # This file is mounted into the litellm container at /app/config.yaml. # Workspace agents reach the proxy at http://litellm:4000 (Docker-internal). # # Usage: # docker compose --profile multi-provider up # # Then in a workspace config.yaml, set: # model: openai/claude-opus-4-5 # routed via LiteLLM → Anthropic # model: openai/gpt-4o # routed via LiteLLM → OpenAI # model: openai/ollama/llama3.2 # routed via LiteLLM → Ollama # # The workspace's OPENAI_BASE_URL env var must be set to http://litellm:4000 # and OPENAI_API_KEY to the value of LITELLM_MASTER_KEY below (default: sk-molecule). # ───────────────────────────────────────────────────────────────────────────── general_settings: master_key: ${LITELLM_MASTER_KEY:-sk-molecule} # used as OPENAI_API_KEY by agents model_list: # ── Anthropic ────────────────────────────────────────────────────────────── - model_name: claude-opus-4-5 litellm_params: model: anthropic/claude-opus-4-5 api_key: ${ANTHROPIC_API_KEY} - model_name: claude-sonnet-4-6 litellm_params: model: anthropic/claude-sonnet-4-6 api_key: ${ANTHROPIC_API_KEY} - model_name: claude-opus-4-7 litellm_params: model: anthropic/claude-opus-4-7 api_key: ${ANTHROPIC_API_KEY} # ── OpenAI ───────────────────────────────────────────────────────────────── - model_name: gpt-4o litellm_params: model: openai/gpt-4o api_key: ${OPENAI_API_KEY} - model_name: gpt-4o-mini litellm_params: model: openai/gpt-4o-mini api_key: ${OPENAI_API_KEY} # ── OpenRouter ───────────────────────────────────────────────────────────── - model_name: openrouter/deepseek-r1 litellm_params: model: openrouter/deepseek/deepseek-r1 api_key: ${OPENROUTER_API_KEY} - model_name: openrouter/qwen3-235b litellm_params: model: openrouter/qwen/qwen3-235b-a22b api_key: ${OPENROUTER_API_KEY} # ── Ollama (local — requires --profile local-models) ─────────────────────── - model_name: ollama/llama3.2 litellm_params: model: ollama/llama3.2 api_base: http://ollama:11434 - model_name: ollama/qwen2.5-coder litellm_params: model: ollama/qwen2.5-coder:7b api_base: http://ollama:11434 litellm_settings: # Drop unsupported params silently (e.g. some models don't support temperature) drop_params: true # Request timeout in seconds request_timeout: 120