Hayley Features
Four intelligent capabilities — orchestration-first tour scheduling, intent-driven apartment discovery, specialized fee reasoning, and a unified multi-modal API — all centralizing intelligence in Hayley Engine while keeping every adapter thin, traceable, and consistent.
Schedule a Tour
SAT is implemented as an orchestration handoff — not an in-service booking engine. The adapter accepts a tourScheduleEnabled flag, forwards schedule_a_tour_flag to Hayley Engine, and returns the engine's answer with full context and latency. Scheduling intelligence stays centralized; the adapter stays as a reliable routing layer.
Find an Apartment
FAA is intent-level support — not a local inventory module. The adapter packages user query and history, sends it to Hayley Engine, and surfaces apartment outputs when the engine classifies intent as find_an_apartment (mapped to "unit availability"). One stable conversational API contract; all business logic in the engine.
Fee Sub-agent
The fee sub-agent is recognized via retrieval_type="fee_sub_agent" (a Constants.CONCISE_RETRIEVAL_TYPES member). Hayley Engine routes fee queries to this dedicated reasoning track; the adapter selects concise output, attaches logs and latency, and persists both sides via GraphQL — with zero fee logic duplicated inside the microservice.
Multi-modality
Multi-modality here means multi-response-mode + multilingual orchestration — not native image/audio parsing. The adapter handles retrieval modes (node, rag, fixed, adapter, fee_sub_agent, spf), supports english/spanish toggles, and passes through context, follow_up, and memory. Clients get one stable endpoint that serves varied AI behaviors without any client-side complexity.