multi-rag

One agent. Structured data, documents, and memory — answered together.

An ops copilot that compiles governed metrics, retrieves cited policies, and keeps context across chats.

What it brings together

Three retrieval surfaces, one streaming tool-use loop — built for e-commerce operations questions that span systems.

  • Governed SQL metrics

    Ask about sales, inventory, finance, and HR through a curated semantic layer — not free-form SQL guessing.

  • Document RAG

    Search policies, FAQs, and manuals in Pinecone with citations, so every policy claim points back to a chunk.

  • Durable agent memory

    Remember preferences and facts across conversations, then recall or forget them when the context changes.

  • Tool-use agent loop

    Anthropic tool_use streams SQL tables, charts, and doc hits in one answer — with live progress you can audit.

Why teams use it

Less context-switching, fewer hallucinated numbers, and a clear trail from question to evidence.

  • Answers you can trust

    Metrics are compiled from a registry. Documents are cited. Raw SQL stays a guarded escape hatch.

  • One place for ops questions

    Hybrid prompts span warehouse numbers and return policy in a single turn — no tab-hopping between tools.

  • Built for demos and daily ops

    Upload corpora, inspect memory, watch live KPI insights, and chat with charts beside the transcript.

Ask the warehouse and the policy in one breath.

Start a chat, or ingest a document and watch retrieval light up.

Start chatting