AI Agent

AI-powered Japanese business contract generator with web interface and CLI tools. Built with OpenAI GPT-4o and integrated Langfuse observability.

Features

• Contract Generation: Japanese lease agreements and outsourcing contracts
• Dual Interface: Web UI and command-line tools
• Observability: Complete tracing with Langfuse
• Docker Deployment: Single-command setup

Prerequisites

• Docker & Docker Compose
• OpenAI API Key
• Free ports 3000 and 8000 (or modify docker-compose.yml)

Quick Start

git clone <this-repository-url>
cd <your-project-directory>
chmod +x setup.sh
./setup.sh

The setup script will:

1. Create .env from template
2. Prompt for your OpenAI API key
3. Start all services

Access:

• Web Interface: http://localhost:8000
• Langfuse Dashboard: http://localhost:3000

Running the Project

If everything is already set up, simply start the services:

# Start all services
docker compose up -d

# Check if services are running
docker compose ps

# Stop services when done
docker compose down

Configuration

Edit .env with your API keys:

# Required
OPENAI_API_KEY=your_actual_openai_api_key_here
OPENAI_MODEL=gpt-4o

# Get from Langfuse dashboard after first run
LANGFUSE_PUBLIC_KEY=pk-lf-your-keys-here
LANGFUSE_SECRET_KEY=sk-lf-your-keys-here

Getting Langfuse Keys:

1. Open http://localhost:3000
2. Sign up → Create project → Settings → API Keys
3. Copy keys to .env
4. Restart: docker compose restart ai-agent

Usage

Web Interface

1. Open http://localhost:8000
2. Select contract type and parameters
3. Generate contract

CLI

The CLI works seamlessly alongside the web interface with shared configuration and tracing.

# 1. Activate virtual environment (REQUIRED)
source .venv/bin/activate

# 2. Install dependencies if not already done
pip install -r requirements.txt

# 3. Generate contract
ai_agent \
  generate_contract \
  --contract_type lease_agreement \
  --number_of_words 1000 \
  --party_a "Company A" \
  --party_b "Company B"

Quick usage options:

# One-liner
source .venv/bin/activate && ai_agent generate_contract --contract_type lease_agreement --number_of_words 1000 --party_a "Company A" --party_b "Company B"

# Create an alias for repeated use
alias ai_agent_cli='source .venv/bin/activate && ai_agent'
ai_agent_cli generate_contract --contract_type lease_agreement --number_of_words 1000 --party_a "Company A" --party_b "Company B"

Note: If Docker services are not running, add export TELEMETRY_ENABLED=false before the command for cleaner output.

Contract Types:

• lease_agreement - Japanese lease agreements
• outsourcing_contract - Business outsourcing contracts

Benefits of unified configuration:

• ✅ CLI and web interface share the same contracts/ folder
• ✅ Both use identical AI model settings and API keys
• ✅ All operations traced in the same Langfuse dashboard
• ✅ No environment switching needed between CLI and web usage

Troubleshooting

Port Conflicts

If ports 3000/8000 are in use:

# Check what's using ports
lsof -i :3000
lsof -i :8000

# Modify docker-compose.yml ports section
langfuse-web:
  ports:
    - "3004:3000"  # Use different external port

Common Issues

• Services won't start: docker compose ps to check status
• OpenAI errors: Verify API key and billing
• Langfuse connection: Ensure keys are correct in .env

CLI Issues

• Connection timeout errors: Wait 30-60 seconds after docker compose up for Langfuse to fully start
• Command not found: Activate virtual environment with source .venv/bin/activate
• Missing dependencies: Run pip install -r requirements.txt in activated venv
• No tracing when Docker is off: Add export TELEMETRY_ENABLED=false for standalone CLI usage

Development

# View logs
docker compose logs ai-agent

# Restart services
docker compose restart

# Stop everything
docker compose down

Architecture

Docker Compose orchestrates AI Agent (FastAPI), Langfuse (observability), and supporting databases. The AI Agent connects to OpenAI's API for contract generation.