LLM API

The LLM surface wraps the rust engine’s retrieval primitives in four agent-friendly endpoints: /llm/answer (RAG over page transcriptions), /llm/expand (natural-language to KWS query expansion), /llm/page-context (structured per-page context bundle), and /llm/status (poll a running answer job). The configured model and budget are controlled by the LLM profile in cfg.

Method	Path	Summary	Key params	Response
POST	/llm/answer	RAG answer over selected page transcriptions	body: LlmAnswerRequest	LlmAnswerResponse
POST	/llm/expand	Translate a natural-language question into a KWS query	body: LlmExpandRequest	LlmExpandResponse
POST	/llm/page-context	Structured transcription context for the requested pages	body: LlmAnswerRequest	(no body)
POST	/llm/status	Poll a previous /llm/answer call	body: LlmAnswerRequest	(no body)

Examples

Ask a question (RAG)

curl -X POST -H "X-API-Key: $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"question": "who attended the 1789 session?"}' \
  "$ARCHAIC/llm/answer"

See also

• /docs/deployment/llm-setup/