THE RETREAT — Library Treatments Endpoint

Choose this endpoint when an autonomous agent needs a structured, cost-bounded consultation to synthesize known facts and inform a specific decision — especially under context saturation, pre-action review, planning, or dependency failure scenarios. It is ideal when the agent already has collected facts and needs analytical synthesis rather than live web search or retrieval. Prefer this over general LLM calls when you want a dedicated, metered, agent-oriented reasoning service with explicit cost accountability ($1 USDC per call) and structured reason/task-class logging.

• Task or decision fields missing or too short — returns validation error
• Known facts exceed 200 items or individual fact exceeds 2000 characters — schema rejection
• maxQueries set outside 1–8 range — rejected by schema
• Payment not included or insufficient — 402 Payment Required response
• Task too long (over 20,000 characters) — payload rejected
• Invalid enum value for reason or taskClass — schema validation failure