LLM Provider Configuration
Configure LLM providers for OSSA agents - Anthropic, OpenAI, Google, and more
LLM Provider Configuration
OSSA supports multiple LLM providers through a unified configuration interface.
Supported Providers
| Provider | Models | Status |
|---|---|---|
| Anthropic | Claude 3.5, Claude 3 Opus/Sonnet/Haiku | Stable |
| OpenAI | GPT-4o, GPT-4 Turbo, GPT-3.5 | Stable |
| Gemini 2.0, Gemini 1.5 | Stable | |
| Mistral | Mistral Large, Mixtral | Beta |
| Groq | LLaMA, Mixtral | Beta |
Configuration
apiVersion: ossa/v0.4.9 kind: Agent spec: llm: provider: ${LLM_PROVIDER:-anthropic} model: ${LLM_MODEL:-claude-sonnet-4-20250514} temperature: ${LLM_TEMPERATURE:-0.1} maxTokens: ${LLM_MAX_TOKENS:-16000} # Fallback configuration fallback_models: - provider: openai model: gpt-4o condition: on_error - provider: google model: gemini-2.0-flash condition: on_rate_limit # Retry configuration retry_config: max_attempts: 3 backoff_strategy: exponential initial_delay_ms: 1000
Environment Variables
| Variable | Description | Default |
|---|---|---|
LLM_PROVIDER | Primary provider | anthropic |
LLM_MODEL | Model name | claude-sonnet-4-20250514 |
LLM_TEMPERATURE | Response randomness | 0.1 |
LLM_MAX_TOKENS | Max output tokens | 16000 |