sandbox-notebookllamalm/ENTERPRISE_SETUP.md

NotebookLlaMa Enterprise Setup

This document describes the enterprise features and setup for NotebookLlaMa.

New Features

🔐 User Authentication

• User registration and login system
• Secure password hashing with bcrypt
• Session-based authentication
• Per-user document isolation

📚 Document Management

• All processed documents are saved to database
• View document library with all your processed notebooks
• Persistent storage of summaries, Q&A, highlights, and mind maps
• Document access control

💬 Persistent Chat History

• Chat sessions are saved per document
• Load previous conversations
• Multiple chat sessions per document
• Search through chat history

🤝 Document Sharing

• Share documents with other users
• Granular permissions (Read/Write/Admin)
• View documents shared with you
• Manage who has access to your documents

🎯 Multi-tenancy

• Complete data isolation per user
• User-specific pipelines (future enhancement)
• Secure access controls

Setup Instructions

1. Database Setup

The application now requires PostgreSQL. Make sure Docker is running:

```bash

Start PostgreSQL and other services

docker compose up -d

Initialize the database schema

uv run src/notebookllama/init_database.py ```

2. Environment Variables

Update your .env file with the correct database credentials:

```bash

Database Configuration

pgql_db=postgres pgql_user=postgres pgql_psw=admin

API Keys (existing)

OPENAI_API_KEY="your-key" LLAMACLOUD_API_KEY="your-key" ELEVENLABS_API_KEY="your-key" EXTRACT_AGENT_ID="your-id" LLAMACLOUD_PIPELINE_ID="your-id" ```

3. Install Dependencies

```bash uv sync ```

4. Run the Application

Start the MCP server (in one terminal):

```bash uv run src/notebookllama/server.py ```

Start the Streamlit app (in another terminal):

```bash streamlit run src/notebookllama/App.py ```

5. First Time Setup

1. Navigate to http://localhost:8501
2. Click on "Sign Up" tab
3. Create your account
4. Start using the application!

Architecture

Database Schema

• users: User accounts
• documents: Uploaded documents
• notebooks: Processed notebook data (summaries, Q&A, highlights)
• chat_sessions: Chat conversation sessions
• chat_messages: Individual chat messages
• document_shares: Document sharing permissions

Application Structure

• App.py: Main entry point with authentication
• pages/1_Process_Document.py: Document upload and processing
• pages/2_Document_Chat.py: Chat interface with persistence
• pages/3_My_Documents.py: User's document library
• pages/4_Shared_Documents.py: Documents shared with user
• auth.py: Authentication and user management
• database.py: Database models and ORM
• document_manager.py: Document CRUD operations
• instrumentation_init.py: Global instrumentation (fixes TracerProvider warning)

Bug Fixes

MCP Server 500 Error

• Added proper error handling in server.py
• Checks for required environment variables
• Graceful degradation when services unavailable
• Detailed logging for debugging

TracerProvider Override Warning

• Moved instrumentation to global singleton
• Initializes once on application start
• Prevents multiple TracerProvider registrations

PostgreSQL Connection Issues

• Fixed environment variable configuration
• Proper database connection pooling
• Connection cleanup

Performance Improvements

Implemented

• Database connection pooling (pool_size=10, max_overflow=20)
• Proper async handling for MCP client
• Efficient query patterns with SQLAlchemy
• Session state management

Future Enhancements

• Redis caching layer for frequent queries
• Background job queue for long-running tasks
• Rate limiting per user
• CDN for static assets
• Horizontal scaling with load balancer

Security

• Passwords hashed with bcrypt
• SQL injection protection via SQLAlchemy ORM
• Session-based authentication
• Document access controls
• Per-user data isolation

Troubleshooting

Database Connection Issues

If you see "role does not exist" errors:

1. Stop any local PostgreSQL instances: ```bash brew services stop postgresql@14 killall postgres ```

2. Recreate Docker volumes: ```bash docker compose down -v docker compose up -d ```

3. Re-initialize database: ```bash uv run src/notebookllama/init_database.py ```

MCP Server Not Responding

1. Check if server is running: ```bash lsof -i :8000 ```

2. Check server logs: ```bash tail -f server.log ```

3. Restart server: ```bash killall -9 python uv run src/notebookllama/server.py ```

Migration from Old Version

If you were using the old version without authentication:

1. All existing data will be lost (no migration path)
2. You'll need to re-process documents
3. Create a new user account
4. Re-upload and process your PDF files

Next Steps

Future enhancements planned:

• Per-document LlamaCloud pipelines
• Advanced search across all documents
• Team/organization support
• API access with tokens
• Webhook notifications
• Document versioning
• Export functionality
• Advanced analytics dashboard