AI Gateway & Hybrid Search API

[bkkothari2255]

Goal: Build the "Intelligence Layer" that AEM will call via REST.

⚡ Sprint 3: AI Gateway & Hybrid Search

• DEV-301: FastAPI Intelligence Backend
  • Create a FastAPI server to host the AI Search service.
  • Implement an endpoint that accepts natural language queries.
• DEV-302: Hybrid Query Implementation
  • Combine Vector Similarity search (meaning) with Lexical filtering (exact terms).
  • Use MAX score mode for result ranking.
• DEV-303: RAG Logic with LangChain
  • Connect retrieved chunks to an LLM (via OpenAI or local Ollama).
  • Implement grounding prompts so responses are based ONLY on crawled AEM data.
• DEV-304: Citation & Guardrails Engine
  • Add logic to return source citations (JCR paths) with every AI response.
  • Implement safety filters for harmful or off-topic queries.