Back to Showcase
Verified Architecture
Logistics & Supply Chain

Enterprise Multi-Source RAG Support System

An enterprise-grade hybrid retrieval-augmented generation (RAG) assistant that indexes dense operational SOPs and live warehouse shipment data to resolve Tier-1 tickets autonomously with complete factual accuracy.

LangChainQdrant Vector DBFastAPIZendesk
Live System Demo
Architecture & Interface Screenshots (2)
Enterprise Multi-Source RAG Support System screenshot 1
Figure 1View Full Size ↗
Enterprise Multi-Source RAG Support System screenshot 2
Figure 2View Full Size ↗

What Was Engineered

  • Semantic vector indexing across PDF manuals, policy guides, and warehouse docs
  • Sub-400ms hybrid search utilizing Qdrant vector database and BM25 keywords
  • Direct ERP integration for real-time shipment status and tracking lookups
  • Strict fallback guardrails transferring edge cases seamlessly to human Tier-2 agents

Execution Pipeline Flow

  1. 1Incoming Zendesk ticket triggers a secure FastAPI webhook request.
  2. 2Query embedding is matched against vector embeddings in Qdrant Vector DB.
  3. 3Context chunks along with real-time ERP data are passed into LLM prompt template.
  4. 4Validated, hallucination-free response is posted back to Zendesk while updating ticket status.