AI Memory Service - Quick Start Guide

🚀 Getting Started with AI Memory Service & Streamlit Playground

This guide will help you quickly set up and start playing with the AI Memory Service using the interactive Streamlit interface.

Memory Service

This project utelise a form of the ai-memory project.

📋 Prerequisites

• Python 3.10+
• MongoDB Atlas account (free tier works)
• OpenAI API key
• AWS account with Bedrock access (optional, for advanced features)

⚙️ Environment Setup

1. Clone the Repository

git clone https://github.com/mongodb-partners/ai-memory.git
cd ai-memory

2. Install Dependencies

pip install -r requirements.txt

3. Configure Environment Variables

Create a .env file in the project root with the following configuration:

# MongoDB Atlas Configuration
MONGODB_URI=mongodb+srv://username:password@cluster.mongodb.net/?retryWrites=true&w=majority&appName=YourCluster

# OpenAI Configuration (Required for Streamlit app)
OPENAI_API_KEY=sk-proj-your-openai-api-key-here

# AWS Configuration (Optional - for advanced memory features)
AWS_ACCESS_KEY_ID=your-aws-access-key
AWS_SECRET_ACCESS_KEY=your-aws-secret-key
AWS_REGION=us-east-1
LLM_MODEL_ID=us.anthropic.claude-3-7-sonnet-20250219-v1:0
EMBEDDING_MODEL_ID=amazon.titan-embed-text-v1

# Memory Service Configuration
SERVICE_HOST=0.0.0.0
SERVICE_PORT=8182
DEBUG=false

# Memory System Parameters (Optional - has defaults)
MAX_DEPTH=5
SIMILARITY_THRESHOLD=0.7
DECAY_FACTOR=0.99
REINFORCEMENT_FACTOR=1.1

4. MongoDB Atlas Setup

1. Create a MongoDB Atlas Account: Go to mongodb.com/atlas
2. Create a Cluster: Use the free M0 tier
3. Enable Vector Search: In your cluster, go to "Search" → "Create Search Index"
4. Create Database: Name it ai_memory
5. Get Connection String: Replace username, password, and cluster details in MONGODB_URI

5. OpenAI API Key Setup

1. Get API Key: Go to platform.openai.com
2. Create New Key: Click "Create new secret key"
3. Copy Key: Add it to your .env file as OPENAI_API_KEY

🏃‍♂️ Running the Services

Start the Memory Service (Backend)

python main.py

The service will start on http://localhost:8182

Start the Streamlit App (Frontend)

streamlit run streamlit_app_fixed.py

The app will open in your browser at http://localhost:8501

🎮 What You Can Play Around With

1. System Prompt Templates 📝

• Helpful Assistant: General-purpose AI
• Business Advisor: Strategic business consultant
• Learning Tutor: Educational assistant
• Code Mentor: Programming expert
• Custom: Write your own personality

Try This: Switch between templates and see how the AI's responses change!

2. PydanticAI Tools Playground 🛠️

Pre-built Sample Tools:

• User Preferences: "Get my user preferences for user_001"
• Weather Lookup: "What's the weather in New York?"
• Investment Calculator: "Calculate returns for $1000 at 5% for 10 years"

Create Custom Tools:

{
  "tool_name": "task_manager",
  "input": {
    "user_id": "string",
    "action": "string"
  },
  "output": {
    "tasks": ["array of strings"],
    "status": "string"
  }
}

Try This: Create a tool for your specific use case and watch the AI use it!

3. Memory System Experiments 🧠

Generate Sample Data:

1. Click "🎯 Random Username" → Gets: amazing_developer_123
2. Click "💬 Random Conversation ID" → Gets: project_planning_20250131_abc123
3. Click "🚀 Generate Sample Conversations" → Creates realistic conversation history

Test Memory Retrieval:

After generating sample data, try these queries:

• "What did we discuss about AI projects?"
• "Tell me about the code review conversation"
• "What MongoDB advice was given?"
• "Remind me about the tool development discussion"

4. Memory Retrieval Modes 🔍

• Automatic: Always retrieves memories (best for ongoing conversations)
• Query-based: Only when you ask for context (keywords: "remember", "what did we")
• Disabled: No memory retrieval (for testing current context only)

Try This: Switch modes and see how the AI's awareness changes!

5. Configuration Management 💾

Save Your Perfect Setup:

1. Configure your ideal system prompt
2. Set up your favorite tools
3. Click "💾 Save Config"
4. Download the JSON file

Load Previous Configurations:

1. Click "📤 Load Config"
2. Upload your saved JSON
3. Everything restores automatically!

6. Context Window Debugging 🔍

At the bottom of the app, explore:

• Memory Context: See what memories the AI has access to
• System Prompt: View the current AI instructions
• Tool Status: Check which tools are available
• Tool Execution History: See detailed tool calls with inputs/outputs
• Configuration: Review all current settings

🎯 Fun Experiments to Try

Experiment 1: Memory Building

1. Generate sample conversations
2. Start a new conversation thread
3. Ask: "What do you know about my preferences?"
4. Watch it pull relevant memories!

Experiment 2: Tool Chaining

1. Enable PydanticAI tools
2. Load sample tools
3. Ask: "Check my user preferences and then get the weather for my location"
4. Watch multiple tools work together!

Experiment 3: Personality Testing

1. Switch to "Business Advisor" template
2. Ask: "Should I invest in this project?"
3. Switch to "Learning Tutor" template
4. Ask the same question
5. Compare the different approaches!

Experiment 4: Memory Persistence

1. Have a conversation about your interests
2. Start a new conversation thread (different ID)
3. Ask: "What do you remember about me?"
4. Test cross-conversation memory!

🔧 Troubleshooting

Service Not Starting?

• Check MongoDB URI is correct
• Verify AWS credentials (if using Bedrock)
• Check port 8182 isn't in use

Streamlit Errors?

• Ensure OpenAI API key is valid
• Check memory service is running on localhost:8182
• Verify all Python dependencies are installed

Memory Not Working?

• Check MongoDB Atlas connection
• Verify vector search indexes are created
• Try the health check button in the app

🌟 Pro Tips

1. Start Simple: Begin with OpenAI mode, then add PydanticAI tools
2. Generate Sample Data: Use sample conversations to test memory features
3. Save Configurations: Save different setups for different use cases
4. Monitor Context Window: Use it to understand what the AI "sees"
5. Experiment with Prompts: Custom system prompts can create specialized agents

📚 Next Steps

Once you're comfortable with the basics:

• Create custom tools for your business logic
• Experiment with different memory retrieval strategies
• Build complex multi-step workflows using tool chaining
• Integrate with your own data sources

Enjoy exploring the AI Memory Service! 🚀