Enterprise AI Platform

Multi-Agent Document Intelligence — 100% On-Premise

Built for fintech use case. Every line runs locally. Zero data leaves this machine.

What It Does

Processes business documents automatically using a multi-agent AI pipeline:

• Invoices — extracts vendor, amount, flags anomalies
• Support Tickets — understands issue, drafts professional reply
• Contracts — extracts clauses, identifies risks
• Code Files — explains, finds bugs, generates unit tests

Security

• Ollama runs LLM 100% locally — no API keys needed
• All telemetry disabled (LangChain, ChromaDB, Streamlit)
• Proven air-gap — works with WiFi completely off
• Full audit trail in local SQLite database

Tech Stack

Tool	Purpose
Ollama + Mistral 7B	Local LLM inference
LangChain	Agent framework
ChromaDB	Local vector database
Streamlit	Web UI
SQLite	Audit logging
Python	Core language

Quick Start

# Install Ollama
brew install ollama
ollama serve
ollama pull mistral

# Run the app
cd enterprise_ai_platform
python3 -m venv venv
source venv/bin/activate
pip install streamlit langchain langchain-community chromadb pymupdf python-docx requests
streamlit run app.py

Architecture

Document Input
      ↓
Classifier Agent
      ↓
Invoice / Support / Contract / Code Agent
      ↓
Summarizer Agent
      ↓
SQLite Audit Log + ChromaDB Vector Store

Author

Aditya Raj — AI Engineer GitHub: https://github.com/Aditraj07