Skip to content

06 — AI Assistant / RAG

Agentic Chat & Retrieval

Ask a question in plain language. The right agent answers it from your own documents and data.

A conversational layer over a document and records platform. A router decides what a question actually needs — retrieval from documents, a query against the database, or a specialist agent — and dispatches accordingly.

Retrieval-augmented generation over a vector store, natural language to SQL against live tables, and graph traversal for relationships between entities.

What I built

  • 01Agent orchestration with a routing layer that picks the right specialist per query
  • 02Retrieval-augmented generation over an embedded document corpus
  • 03Natural-language-to-SQL against live operational tables
  • 04Graph-backed relationship queries across entities
  • 05Conversation memory, streaming responses and citation back to source documents
  • 06Tenant-scoped retrieval so no answer can cross a customer boundary

Outcomes

  • —One conversational entry point over several very different data sources
  • —Answers traceable back to the document they came from
  • —Model versions swapped without rewriting agent logic