Expand description
The LlmBackend trait — one complete call maps a chat_request
payload to a chat_response payload.
Keeping the trait payload-in / payload-out (rather than record-in / record-out) lets the engine own the envelope (ts, parent, id) while the backend only implements the LLM call itself. See SPEC.md §10 lifecycle.
Enums§
- LlmError
- Errors a backend may return.
Traits§
- LlmBackend
- Pluggable LLM backend for replay.