---

Enterprise RAG Connector Kit

A modular Python framework for ingesting enterprise content, validating and indexing documents, retrieving grounded context, and exposing a local MCP-compatible chatbot workflow.

The current implementation includes:

• composable source adapters
• batching, validation, retry, and structured logging
• a retrieval + grounded-answer workflow
• an MCP tool for local client integration
• test coverage for core indexing and retrieval paths

Overview

This project is structured as a reusable connector and retrieval framework rather than a one-off script. It is designed to support repeated onboarding scenarios where content sources, retrieval workflows, and client integrations may evolve over time.

At a high level, the project supports:

• content ingestion from local JSON or REST-based sources
• document validation and batch-oriented indexing
• retrieval of relevant content for a user question
• grounded answer generation with source attribution
• local MCP exposure for compatible developer tools

Architecture

Source Data
   ↓
Adapter
   ↓
Validator
   ↓
Indexing Engine
   ↓
Provider Indexing API

User Question
   ↓
RAG Service
   ↓
Search API → Retrieve Documents
   ↓
Chat API → Generate Answer
   ↓
Return Answer + Sources

---

Quick Start

# Activate the cirtual environment
.\.venv\Scripts\Activate.ps1

# Sanity-check the codebase
pytest
ruff check .

# Indexing flow
python .\main.py

# Search API verification script
python .\verify_indexing.py

# Chatbot from CLI
python .\ask_chatbot.py "Which documents discuss observability?" 5

# MCP server
python -m glean_indexing_connector.mcp.server

# MCP Tool (Web Interface)
mcp dev src\glean_indexing_connector\mcp\server.py
or
python -m mcp dev src\glean_indexing_connector\mcp\server.py