Endpoint Reference

Prompt

• POST /ask: single prompt/response

Chat

• POST /chat: create/continue conversation using message (+ optional conversation_id)
• GET /chat/{conversation_id}: conversation metadata
• DELETE /chat/{conversation_id}: remove in-memory conversation

Health and Ops

• GET /health: health + readiness + runtime observability
• GET /health/history: rolling readiness history (?since= optional)
• POST /health/recheck: force readiness re-check (API key required)
• GET /metrics: Prometheus metrics
• GET /version: service version metadata

Discovery and Docs

• GET /models: locally available Claude models
• GET /openapi.yaml: OpenAPI spec
• GET /docs: runtime Swagger docs UI

Common Error Codes

• validation_error (400): request shape/value issue
• unauthorized (401): missing/invalid API key
• payload_too_large (413): body exceeds configured limit
• rate_limited (429): per-IP rate cap reached
• concurrency_limited (429): Claude queue is full
• shutting_down (503): drain mode enabled
• queue_timeout (504): queue wait exceeded timeout
• timeout (504): Claude execution timeout
• cli_error / spawn_error / internal_error (500): runtime/internal failures

For exact schemas and all statuses, use OpenAPI Explorer.