A2A CR Box - Enterprise Content Discovery Agent

A professional enterprise content discovery assistant powered by Gemini, specializing in Box content search, AI analysis, document generation, and MCP integration.

🚀 Features

Core Capabilities

• Box Content Search: Find files, documents, and folders in Box
• Box AI Ask: Ask intelligent questions about specific file content
• Box Hub Ask: Automatically discover and use the most relevant Box Hub
• Box Doc Gen: Generate professional capital call notices and other documents using templates
• Box MCP Integration: Leverage Box MCP Remote Server for enhanced AI capabilities and cross-platform integration
• Enterprise Content Discovery: Locate specific documents, regulatory files, reports, and business content
• Professional Communication: Business-appropriate responses with enterprise focus

Tools Available

1. Box Search - Find documents and files across your Box enterprise
2. Box AI Ask - Ask questions about specific file content using Box AI
3. Box Hub Ask - Automatically discover and use the most relevant knowledge hub
4. Box Doc Gen - Generate capital call notices and other documents using Box Doc Gen API
5. Box MCP Tools - Enhanced AI capabilities through Box MCP Remote Server
6. Capital Call Workflow - Complete workflow for capital call notice creation

🏗️ Architecture

• Google ADK (Agent Development Kit): Powers the GeminiAgent
• Box Python SDK: Handles Box API authentication and operations
• JWT Authentication: Secure server-to-server authentication with Box
• Cloud Run: Deployed as a scalable containerized service
• AgentSpace Integration: Ready for Discovery Engine registration

🔧 Setup

Prerequisites

• Python 3.9+
• Google Cloud Platform account
• Box Developer account with JWT app credentials

Installation

1. Clone the repository

   git clone <your-repo-url>
   cd A2A_CR_Box

2. Install dependencies

   pip install -r requirements.txt

3. Configure Box JWT Authentication

   Option A: Local JWT Config (Development)

   • Copy box_jwt_config.example.json to box_jwt_config.json
   • Fill in your Box JWT app credentials:
     • clientID
     • clientSecret
     • enterpriseID
     • publicKeyID
     • privateKey
     • passphrase

   Option B: Google Secret Manager (Production - Recommended)

   This is the secure, production-ready approach that stores your JWT credentials in Google Secret Manager instead of local files.

   Step 1: Enable Secret Manager API

   gcloud services enable secretmanager.googleapis.com --project=your-project-id

   Step 2: Create Box JWT Secrets

   # Create secret for Box Client ID
   echo -n "your_client_id_here" | gcloud secrets create box-client-id --data-file=- --project=your-project-id
   
   # Create secret for Box Client Secret  
   echo -n "your_client_secret_here" | gcloud secrets create box-client-secret --data-file=- --project=your-project-id
   
   # Create secret for Box Public Key ID
   echo -n "your_public_key_id_here" | gcloud secrets create box-public-key-id --data-file=- --project=your-project-id
   
   # Create secret for Box Private Key (the entire PEM content)
   gcloud secrets create box-private-key --data-file=path/to/your/private_key.pem --project=your-project-id
   
   # Create secret for Box Private Key Passphrase
   echo -n "your_passphrase_here" | gcloud secrets create box-private-key-passphrase --data-file=- --project=your-project-id
   
   # Create secret for Box Enterprise ID
   echo -n "your_enterprise_id_here" | gcloud secrets create box-enterprise-id --data-file=- --project=your-project-id

   Step 3: Use the Setup Script (Easiest Method)

   Save this script as box_jwt_secrets_setup.sh:

   #!/bin/bash
   # box_jwt_secrets_setup.sh
   
   echo "🔐 Setting up Box JWT secrets in Google Secret Manager..."
   echo ""
   
   # Prompt for credentials
   read -p "Enter Box Client ID: " BOX_CLIENT_ID
   read -p "Enter Box Client Secret: " BOX_CLIENT_SECRET
   read -p "Enter Box Public Key ID: " BOX_PUBLIC_KEY_ID
   read -p "Enter Box Enterprise ID: " BOX_ENTERPRISE_ID
   read -s -p "Enter Box Private Key Passphrase: " BOX_PASSPHRASE
   echo ""
   read -p "Enter path to Box Private Key PEM file: " BOX_PRIVATE_KEY_PATH
   
   # Set project
   PROJECT_ID="your-project-id"
   
   echo ""
   echo "📝 Creating secrets in project: $PROJECT_ID"
   echo ""
   
   # Create secrets
   echo "Creating box-client-id..."
   echo -n "$BOX_CLIENT_ID" | gcloud secrets create box-client-id --data-file=- --project=$PROJECT_ID
   
   echo "Creating box-client-secret..."
   echo -n "$BOX_CLIENT_SECRET" | gcloud secrets create box-client-secret --data-file=- --project=$PROJECT_ID
   
   echo "Creating box-public-key-id..."
   echo -n "$BOX_PUBLIC_KEY_ID" | gcloud secrets create box-public-key-id --data-file=- --project=$PROJECT_ID
   
   echo "Creating box-enterprise-id..."
   echo -n "$BOX_ENTERPRISE_ID" | gcloud secrets create box-enterprise-id --data-file=- --project=$PROJECT_ID
   
   echo "Creating box-private-key-passphrase..."
   echo -n "$BOX_PASSPHRASE" | gcloud secrets create box-private-key-passphrase --data-file=- --project=$PROJECT_ID
   
   echo "Creating box-private-key..."
   gcloud secrets create box-private-key --data-file="$BOX_PRIVATE_KEY_PATH" --project=$PROJECT_ID
   
   echo ""
   echo "✅ All Box JWT secrets created successfully!"
   echo "🔒 Your credentials are now securely stored in Google Secret Manager"

   Make it executable and run:

   chmod +x box_jwt_secrets_setup.sh
   ./box_jwt_secrets_setup.sh

   Step 4: Update Requirements.txt Add the Secret Manager client library:

   google-cloud-secret-manager>=2.0.0

   Step 5: Update .gitignore Now you can safely ignore the JWT config file:

   # Box Configuration (Sensitive)
   box_jwt_config.json
   *.jwt
   *.pem
   *.key
   *.p12
   *.pfx

4. Set environment variables

   export GOOGLE_CLOUD_PROJECT="your-project-id"
   export GOOGLE_CLOUD_LOCATION="us-central1"
   export GOOGLE_GENAI_USE_VERTEXAI="TRUE"
   export MODEL="gemini-2.5-flash"

🚀 Deployment

Local Development

python -m uvicorn agent_executor:app --reload

Cloud Run Deployment

./deploy.sh <project-id> <service-name>

AgentSpace Registration

After deployment, register your agent in Google Cloud Discovery Engine:

curl -X POST \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
  "https://discoveryengine.googleapis.com/v1alpha/projects/<PROJECT_ID>/locations/global/collections/default_collection/engines/<ENGINE_ID>/assistants/default_assistant/agents" \
  -d '{
    "name": "Box_Search_Agent",
    "displayName": "Box Search Agent",
    "description": "Enterprise content discovery agent for Box",
    "a2aAgentDefinition": {
      "jsonAgentCard": "{\"provider\": {\"url\": \"<YOUR_CLOUD_RUN_URL>\"},\"name\": \"box_search_agent\",\"description\": \"A Box search assistant that helps you find content in your Box enterprise\"}"
    }
  }'

📖 Usage Examples

Box Content Search

User: "Find regulatory documents"
Agent: [Searches Box and returns organized results with file counts and details]

Box AI Ask

User: "What are the key points in Capital Call Notice.pdf?"
Agent: [Uses Box AI to analyze the specific file and provide insights]

Box Hub Ask

User: "What are our company policies?"
Agent: [Automatically discovers relevant hubs and provides answers from the best knowledge base]

Box MCP Integration

User: "Search for Project Phoenix files using MCP"
Agent: [Uses Box MCP Remote Server for enhanced search with advanced filtering]

User: "Analyze this document with Box AI via MCP"
Agent: [Leverages Box AI through MCP for deeper document analysis]

User: "Extract structured data from this file via MCP"
Agent: [Uses Box MCP to convert unstructured documents into structured data]

User: "Get project status from Box Hub via MCP"
Agent: [Provides project-level insights using Box Hub analysis via MCP]

Capital Call Notice Creation

User: "Help me create capital call notices"
Agent: [Guides through complete workflow from template selection to document generation]

User: "Find templates for capital call notices"
Agent: [Discovers available templates and provides guidance]

User: "Show me sample LP data structure"
Agent: [Provides comprehensive sample data structure for capital call notices]

🔒 Security

Authentication Methods

• JWT Authentication: Secure server-to-server authentication with Box
• Environment Variables: Sensitive configuration stored securely
• Box API Permissions: Minimal required permissions for enterprise access
• Cloud Run Security: No unauthenticated access, secure by default

Google Secret Manager Benefits

Using Google Secret Manager for JWT credentials provides several security advantages:

1. 🔒 Enhanced Security: Credentials stored in Google's secure Secret Manager instead of local files
2. 🚀 Easier Deployment: No need to include sensitive files in deployments
3. 📝 Version Control: Secrets can be versioned and rotated without code changes
4. 👥 Access Control: IAM controls who can access secrets
5. 🔄 Easy Updates: Update secrets without redeploying code
6. 🔍 Audit Logging: Track who accessed secrets and when
7. 🌍 Multi-Region: Secrets can be replicated across regions for availability

Best Practices

• Use Google Secret Manager for production deployments
• Keep local JWT config only for development/testing
• Regularly rotate JWT credentials
• Use least-privilege access for secret permissions
• Monitor secret access through Cloud Logging

🛠️ Development

Project Structure

A2A_CR_Box/
├── agent_executor.py      # Main A2A executor
├── gemini_agent.py        # Gemini agent with tools
├── box_auth.py           # Box JWT authentication
├── box_search.py         # Box content search
├── box_ai_ask.py         # Box AI file analysis
├── box_hub_ask.py        # Box Hub discovery and querying
├── box_doc_gen.py        # Box Doc Gen for capital call notices
├── box_mcp_client.py     # Box MCP Remote Server integration
├── requirements.txt      # Python dependencies
├── deploy.sh            # Cloud Run deployment script
├── box_jwt_config.example.json # JWT configuration template
├── box_jwt_secrets_setup.sh   # Secret Manager setup script
└── README.md            # This file

Adding New Tools

1. Create your tool function in a new Python file
2. Import it in gemini_agent.py
3. Add it to the tools list
4. Update the agent instructions
5. Add corresponding skills

📝 License

This project is licensed under the Box TOS. This project is delivered as is with no commitments to maintanence.

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📞 Support

For issues and questions:

• Check the logs in Google Cloud Console
• Review Box API documentation
• Check Google ADK documentation

🔄 Version History

• v1.0.0: Initial release with Box search and AI capabilities
• v1.1.0: Added Box Hub Ask functionality
• v1.2.0: Enhanced enterprise focus and professional communication
• v1.3.0: Added Box Doc Gen integration for capital call notices
• v1.4.0: Added Box MCP Remote Server integration for enhanced AI capabilities
• v1.5.0: Enhanced workflow assistants and cross-platform MCP support