LLM Provider Switching
Nova can run with different LLM providers without changing business logic.
What LLMs are used for
- understanding and drafting language
- extracting structured information
- summarization and response generation
What LLMs are not used for
LLMs do not directly decide unsafe side effects. Deterministic services still control:
- target resolution
- policy checks
- delivery execution
Supported providers
openaianthropicollamaopenai_compatible
Typical setup example
LLM_PROVIDER=ollama
OLLAMA_BASE_URL=https://<your-endpoint>
OLLAMA_API_KEY=<token>
OLLAMA_MODEL=<model>
Provider selection model
- global default from env
- optional per-assistant override
- controlled fallback to compatible mode when configured