RAG Tool Quick Reference Card

Tool Name

manage_rag_documents(operation, params)

Location

/root/qubinode_navigator/airflow/plugins/qubinode/mcp_server_fastmcp.py

Status

✅ Running & Deployed (MCP Server Port 8889)

Operations at a Glance

Operation Parameters Purpose Output
scan {'doc_dir': '/opt/documents/incoming'} Find documents ready for processing List of files, sizes, count
ingest {'doc_dir': '/opt/documents/incoming'} Start document processing pipeline DAG run ID, status
status None Check ingestion progress DAG info, task pipeline
list {'limit': 10} View processed documents Document metadata, chunks
estimate {'doc_dir': '/opt/documents/incoming'} Pre-flight check before ingestion Storage needs, timing

Quick Examples 

# Basic scan
manage_rag_documents('scan')

# Estimate requirements
manage_rag_documents('estimate')

# Trigger ingestion
manage_rag_documents('ingest')

# Check progress
manage_rag_documents('status')

# View results (last 5 documents)
manage_rag_documents('list', {'limit': 5})

# Custom directory
manage_rag_documents('scan', {'doc_dir': '/my/docs'})
manage_rag_documents('ingest', {'doc_dir': '/my/docs'})

Supported File Types

  • .md - Markdown (chunked by headers)
  • .markdown - Markdown (alternative extension)
  • .yml - YAML (entire file as chunk)
  • .yaml - YAML (alternative extension)
  • .txt - Plain text (chunked by paragraphs)

Workflow Example 

1. manage_rag_documents('scan')
   ↓
   → Find 3 documents (256 KB)
   
2. manage_rag_documents('estimate')
   ↓
   → ~180 chunks, 0.41 MB storage, 2 sec processing
   
3. manage_rag_documents('ingest')
   ↓
   → DAG triggered, run ID: scheduled__2025-11-26...
   
4. Wait 10-30 seconds...

5. manage_rag_documents('status')
   ↓
   → Pipeline shows 3 tasks complete
   
6. manage_rag_documents('list')
   ↓
   → 3 documents processed, 25 total chunks

Features

✅ Read-only operations: scan, status, list, estimate
✅ Write operation (requires access): ingest
✅ LLM-optimized responses (markdown formatted)
✅ Clear next-step guidance in every response
✅ Comprehensive error messages
✅ Works with custom directories

Integration

  • Triggers: rag_document_ingestion Airflow DAG
  • Input: /opt/documents/incoming/
  • Output: /opt/documents/processed/metadata.json
  • Vector DB: Qdrant, ChromaDB, or FAISS
  • Embedding Model: all-MiniLM-L6-v2 (384-d)

Access Points

Component URL Port
MCP Server http://localhost:8889/sse 8889
Airflow UI http://localhost:8888/ 8888
RAG DAG http://localhost:8888/dags/rag_document_ingestion 8888
PostgreSQL localhost:5432 5432

Total MCP Tools

11 Total:

  • 3 DAG Management
  • 5 VM Operations
  • 1 Status
  • 1 Information
  • 1 RAG Operations ⭐

Documentation

  • Full Guide: /root/qubinode_navigator/docs/RAG-TOOL-GUIDE.md
  • Implementation: /root/qubinode_navigator/RAG-TOOL-IMPLEMENTATION-SUMMARY.md
  • MCP Server Logs: podman logs $(podman ps | grep mcp-server | awk '{print $1}')
  • Airflow Logs: podman logs $(podman ps | grep airflow-scheduler | awk '{print $1}')
  • Check Tool: podman logs ... | grep "Tools: 11"
  • Verify RAG: podman logs ... | grep "RAG: 1"

Key Tip

Always call manage_rag_documents('estimate') before manage_rag_documents('ingest') to:

  • Verify documents are found
  • Check storage requirements
  • Estimate processing time
  • Confirm system readiness

Status: 🟢 Production Ready
Last Updated: 2025-11-26
Tool Version: 1.0