Supply Chain Optimizer Agent

This project implements an AI-powered agent to optimize and analyze supply chain data using a Neo4j graph database, Google Cloud, and the Agent Development Kit (ADK).

Features

• Intent-Driven Agent: Understands natural language queries about the supply chain.
• Neo4j Knowledge Graph: Models suppliers, products, locations, and their relationships.
• AI-Powered Tools:
  • CypherQueryTool: Runs queries against the Neo4j database.
  • VectorSearchTool: Finds similar past incidents.
  • GeminiPlannerTool: Suggests alternative sourcing routes.
  • SimulatorTool: Simulates disruptions like port closures.
• Cloud Native: Designed for deployment on Google Cloud Run.
• Optional UI: Includes a Streamlit app for visualization.

Project Structure

supply-optimizer-agent/
├── app/
│   ├── __init__.py
│   ├── main.py         # FastAPI application
│   ├── agent.py        # Agent logic (LangGraph/ADK)
│   ├── tools/
│   │   ├── __init__.py
│   │   ├── cypher.py
│   │   ├── search.py
│   │   ├── planner.py
│   │   └── simulator.py
│   └── models/
│       ├── __init__.py
│       └── schema.py       # Pydantic models for data
├── data/
│   └── ingest.py       # Script to generate and ingest demo data
├── ui/
│   └── app.py          # Streamlit UI
├── Dockerfile
├── requirements.txt
└── .env.example

Setup Instructions

1. Prerequisites

• Python 3.9+
• Docker
• Google Cloud SDK
• Access to a Neo4j instance (local or Aura)
• Google Cloud Project with Vertex AI API enabled

2. Local Setup

1. Clone the repository:

   git clone https://github.com/jitu028/supply-chain-optimizer-agent.git
   cd supply-optimizer-agent

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Set up environment variables:

   • Copy .env.example to .env: cp .env.example .env
   • Fill in the values in your .env file.

5. Ingest demo data into Neo4j:

   python data/ingest.py

6. Run the agent locally:

   uvicorn app.main:app --reload

7. Run the UI (optional):

   streamlit run ui/app.py

3. Cloud Deployment (Google Cloud Run)

3.1 api - Agent Backend

gcloud run deploy supply-chain-api \
  --source ./api \
  --region us-central1 \
  --platform managed \
  --allow-unauthenticated \
  --project YOUR_PROJECT_ID

Test the endpoint:

curl -X POST https://supply-chain-api-<hash>-<region>.a.run.app/invoke \
  -H "Content-Type: application/json" \
  -d '{"query": "Simulate the impact if Shanghai suppliers are delayed by 5 days."}'

---

3.2 data-ingestion - Neo4j Seeder (Cloud Run Job)

# Step 1: Build image
gcloud builds submit --tag gcr.io/YOUR_PROJECT_ID/data-ingestion ./data-ingestion

# Step 2: Create Cloud Run Job
gcloud run jobs create data-ingestion-job \
  --image=gcr.io/YOUR_PROJECT_ID/data-ingestion \
  --region=us-central1 \
  --project=YOUR_PROJECT_ID

# Step 3: Run the Job
gcloud run jobs execute data-ingestion-job \
  --region=us-central1 \
  --project=YOUR_PROJECT_ID

---

3.3 ui - Streamlit Frontend

Ensure PORT=8080 is set in .streamlit/config.toml or via os.environ.

gcloud run deploy supply-chain-ui \
  --source ./ui \
  --region us-central1 \
  --platform managed \
  --allow-unauthenticated \
  --project=YOUR_PROJECT_ID