Backend-Kisum-RAG-KB-LLM

Related documentation: AI Addon Backend · Data ownership

Runtime status

Backend-Kisum-RAG-KB-LLM is implemented.

Purpose

This repo provides the retrieval-augmented generation layer for Kisum knowledge retrieval over artist/market directory data.

Implemented now

• ingestion scripts from PostgreSQL source data
• legacy Mongo ingest path still present
• OpenAI embeddings
• Qdrant vector storage
• chat/retrieval API server
• helper scripts for collection setup and smoke tests

Current architecture

Source flow:

PostgreSQL -> ingest scripts -> embeddings -> Qdrant

Query flow:

user question -> query embedding -> Qdrant search -> context assembly -> LLM answer

Ownership boundary

• this repo is not a system of record for the underlying business data
• PostgreSQL source tables remain authoritative
• Qdrant stores retrieval artifacts, not master truth

Not implemented as a platform-wide replacement

This repo is one AI retrieval component. It does not replace:

• Chat-Kisum-MCP-Node
• Backend-MCP-Kisum
• Auth/Core/business-domain services