README.md

metadata

title: CedroPM Bot
emoji: 🤖
colorFrom: blue
colorTo: indigo
sdk: docker
pinned: false
app_port: 7860

PM-RAG-ChatBot (CedroPM)

A specialized RAG (Retrieval Augmented Generation) chatbot designed for Project Managers and Software Architects. It processes Software Requirements Documents (SRDs) and provides expert-level analysis by combining project-specific data with a pre-seeded "Gold Standard" knowledge base of industry best practices.

🚀 Key Features

• 📄 Intelligent SRD Analysis: Upload PDF requirements and get answers grounded strictly in your document.
• 🧠 Global Expert Knowledge: Pre-seeded with 30+ industry standards (ISO, OWASP, NIST, Agile) to provide "Senior Architect" advice even when your PDF is silent.
• 🎨 Diagram Vision: Understands and explains architecture diagrams using Claude Vision or Qwen2-VL.
• 🔍 Hybrid Search: Uses a dual-scope retrieval engine to search your Project PDF and Global Best Practices simultaneously.
• 💬 Interactive Chat: Built with Chainlit for a clean, chat-like interface with history retention.
• 🔐 Secure & Private: Multi-user isolation ensures one project's data never leaks to another.

🏗️ System Architecture

1. Ingestion Layer:
   • PDFs: Processed with pdfplumber and SmartSRDSplitter (Section-aware chunking).
   • Tables: Extracted via camelot and converted to Markdown.
   • Diagrams: Processed via OCR (Tesseract) or Vision LLM (Claude 3.5 Sonnet).
2. Knowledge Store:
   • ChromaDB: vector storage for semantic search.
   • Dual-Scope: Queries filter by (User + Project) OR (Global_Expert_Knowledge).
3. Reasoning Engine:
   • Claude 3.5 Sonnet: Generates detailed, structured responses.
   • Seed Data: A Python script injects "Golden Rules" for Security, DevOps, and PM methodologies.

📋 Prerequisites

• Python 3.10+
• System Tools:
  • Poppler (for PDF rendering)
  • Tesseract OCR (for image text extraction)
• API Key: Anthropic API Key (for Claude).

🔧 Installation & Setup

1. Clone the Repository

git clone <your-repo-url>
cd PM-RAG-ChatBot