Write apps/ai_agent/docs/api-transfers-analyse.md documenting POST /transfers/analyse — the rule-based short-circuit path (high-value transfers skip the LLM entirely) vs. the LLM-analysis path, and the meaning of every response field (confidence, flagged, and whatever else TransferAnalyseResponse defines).
Acceptance criteria:
- Both analysis paths (rule-based and LLM) are documented with the exact condition that routes a request to each
- Every response field's meaning and valid value range is documented
- Default/fallback behavior when the LLM omits
confidence/flagged is documented
Write
apps/ai_agent/docs/api-transfers-analyse.mddocumentingPOST /transfers/analyse— the rule-based short-circuit path (high-value transfers skip the LLM entirely) vs. the LLM-analysis path, and the meaning of every response field (confidence,flagged, and whatever elseTransferAnalyseResponsedefines).Acceptance criteria:
confidence/flaggedis documented