An enterprise-grade multi-agent AI system that automates procurement operations using LangGraph, FastAPI, PostgreSQL, ChromaDB, and Large Language Models.
The platform combines invoice intelligence, procurement analytics, vendor intelligence, approval workflows, procurement risk assessment, and Retrieval-Augmented Generation (RAG) into a single conversational interface.
- Multi-Agent Architecture using LangGraph
- Retrieval-Augmented Generation (RAG) over 100+ procurement documents
- Natural Language Invoice Search and Analysis
- Procurement Spend Analytics
- Vendor Intelligence and Spend Tracking
- Procurement Risk Assessment
- Approval Workflow Automation
- Enterprise-Style Conversational Interface
- Fully Deployed Frontend and Backend
- Query procurement invoices using natural language
- Semantic search across procurement documents
- Business justification extraction
- Invoice summary generation
- Metadata-aware document retrieval
Example:
What is the business justification for invoice INV-1047?
Summarize invoice INV-1001.
Which invoices involve software licenses?
- Total procurement spend analysis
- Approved invoice reporting
- Pending payment tracking
- Executive procurement summaries
Example:
What is the total procurement spend?
Generate a procurement summary.
- Top vendor identification
- Vendor spend analysis
- Spend concentration insights
- Supplier dependency analysis
Example:
Who is the top vendor by spend?
How much have we spent with MedSupply Inc?
- Vendor concentration risk detection
- Outstanding payment analysis
- Procurement exposure monitoring
- Actionable risk recommendations
Example:
Generate a procurement risk report.
- Approve invoices using natural language
- Reject invoices using natural language
- Database-backed workflow execution
- Automated status updates
Example:
Approve invoice INV-1002.
Reject invoice INV-1003.
- Procurement-only assistant behavior
- Out-of-scope query detection
- Missing invoice handling
- Structured business responses
Example:
Tell me about cricket.
Response:
I am an Enterprise Procurement Agent and can only assist with procurement-related questions.
User
│
▼
React Frontend (Vercel)
│
▼
FastAPI Backend (Railway)
│
▼
LangGraph Workflow
│
├── Intent Agent
├── Database Agent
├── Analytics Agent
├── Vendor Analytics Agent
├── Risk Agent
├── Approval Agent
├── Summary Agent
├── RAG Agent
├── Out-of-Scope Agent
└── Response Agent
│
▼
Data Layer
│
├── PostgreSQL (Neon)
├── ChromaDB
└── Procurement PDF Knowledge Base
│
▼
OpenRouter + DeepSeek LLM
- React
- Vite
- Tailwind CSS
- Axios
- FastAPI
- LangGraph
- SQLAlchemy
- OpenRouter
- DeepSeek Chat V3
- Sentence Transformers
- ChromaDB
- PostgreSQL (Neon)
- Railway
- Vercel
- Docker
agents/
api/
database/
docs/
frontend/
models/
rag/
scripts/
workflows/
Dockerfile
docker-compose.yml
requirements.txt
What is the status of invoice INV-1001?
Summarize invoice INV-1047.
What is the business justification for invoice INV-1047?
What is the total procurement spend?
Show unpaid invoices.
Generate a procurement summary.
Who is the top vendor by spend?
How much have we spent with MedSupply Inc?
Generate a procurement risk report.
Approve invoice INV-1002.
Reject invoice INV-1003.
pip install -r requirements.txt
uvicorn api.app:app --reloadcd frontend
npm install
npm run dev
- User Authentication
- Role-Based Access Control
- Supplier Performance Scoring
- Procurement Forecasting
- Contract Intelligence
- Document Upload Interface
- Real-Time Approval Notifications
- Multi-Tenant Procurement Workspaces
Ajay Armugam
B.Tech Graduate | AI/ML Engineer
Built using LangGraph, FastAPI, PostgreSQL, ChromaDB, OpenRouter, and DeepSeek.