Memory

Overview

Memory in FloTorch enables your AI agents to maintain persistent context, conversation history, and long-term memory across interactions. Memory providers store and retrieve information that agents can use to provide more contextual and personalized responses.

Accessing the Memory Section

To access the Memory management interface:

1. Navigate to the Agents section in the left sidebar
2. Click on Memory (highlighted with a database/server icon)
3. You’ll be presented with the Memory management page

Memory Management Interface

The Memory page displays a comprehensive table of all configured memory providers in your workspace with the following information:

• Name: The unique identifier for each memory provider
• Provider: The underlying memory service (e.g., Mem0, AgentCore Memory)
• Description: Optional description of the memory provider’s purpose
• Actions: Options to configure, edit, or delete the memory provider

Creating a New Memory Provider

Adding a Memory Provider

1. Click the ”+ Add Memory Provider” button in the top right corner

2. Fill in the required information:

   • Name*: A unique identifier for your memory provider (required)
   • Description: Optional description of the memory provider’s purpose
   • Provider*: Select the memory service provider (e.g., Mem0, AgentCore Memory) (required)

3. Provider Credentials: Configure the connection details based on your selected provider:

   For Mem0:

   • URL*: The endpoint URL for the memory service (required)
   • Project ID*: Your project identifier from the memory service (required)
   • Org ID*: Your organization identifier (required)
   • API Key*: Authentication key for the memory service (required)

   For AgentCore Memory:

   • Region*: The AWS region where your AgentCore Memory is deployed (required)
   • Access Key ID*: Your AWS access key ID with appropriate permissions (required)
   • Secret Access Key*: Your AWS secret access key for authentication (required)
   • Memory ID (Optional): Specific memory identifier if you want to use a particular memory instance

4. Click Create to add the memory provider

Memory Provider Configuration Best Practices

• Use descriptive names that clearly indicate the memory provider’s purpose
• Ensure all credential fields are properly configured for secure access
• Verify the URL is accessible from your FloTorch workspace (for Mem0)
• Keep API keys and credentials secure and regularly rotate them
• For AgentCore Memory, ensure your AWS credentials have the necessary permissions

Managing Existing Memory Providers

Configuring a Memory Provider

1. Locate the memory provider you want to configure in the memory providers table
2. Click the vertical ellipsis (⋮) icon in the Actions column
3. Select Configure from the dropdown menu
4. Modify the provider’s configuration as needed
5. Click Update to apply changes

Reconfiguring a Memory Provider

When reconfiguring an existing memory provider:

1. Click the Configure option from the Actions menu
2. Modify the desired fields:
   • Name: Update the provider name if needed
   • Description: Modify the description
   • Provider: Change the underlying memory service if required
3. Provider Credentials:
   • Leave any field blank to keep the existing configuration
   • Update only the fields you want to change
4. Click Update to save the changes

Deleting a Memory Provider

Archiving and Unarchiving a Memory Provider

Archiving hides a memory provider from the active list but keeps it available for dependencies. You can restore it later.

Archive a memory provider

1. In the left navigation, go to Memory.
2. In the memory providers table, open the Actions (three dots) menu for the provider and select Archive.
3. In the confirmation dialog, review the message: “Are you sure you want to archive this memory provider? You can restore it later if needed.”
4. Click Archive. The provider is archived and removed from the active list. Existing dependencies continue to work.
5. To view archived providers, open the top filters dropdown and select Archived.

Unarchive a memory provider

1. In the memory providers table, use the top filters dropdown and select Archived.
2. Open the Actions (three dots) menu for the provider and select Unarchive.

Deleting a Memory Provider

Deleting a memory provider permanently removes the provider and all related dependencies. This action cannot be undone.

1. In the left navigation, go to Agents > Memory.
2. In the memory providers table, open the Actions (three dots) menu for the provider and select Delete.
3. Review the dependency list in the deletion modal.
4. Click Continue. A confirmation modal opens and asks you to type the provider name.
5. Enter the provider name and click Permanently Delete.

After confirmation, the provider and all listed dependencies are deleted permanently and cannot be recovered.

Search and Filtering

Use the search bar labeled “Search for a provider” to quickly find specific memory providers by name or other attributes. The search functionality helps you locate providers in workspaces with many configured memory services.

Memory Provider Types

Mem0

Mem0 is a memory service that provides persistent storage for AI agent conversations and context. It enables:

• Conversation History: Store and retrieve previous interactions
• Context Persistence: Maintain context across multiple sessions
• Long-term Memory: Store information that agents can reference later
• Scalable Storage: Handle large amounts of conversation data

Configuration Requirements

To use Mem0 as a memory provider, you’ll need:

• API Endpoint: https://api.mem0.ai
• Project ID: Your Mem0 project identifier
• Organization ID: Your Mem0 organization identifier
• API Key: Authentication key for accessing the Mem0 service

AgentCore Memory

AgentCore Memory is Amazon Bedrock’s advanced memory management service that provides enterprise-grade memory capabilities for AI agents. It offers:

• Enterprise-Grade Storage: Built on AWS infrastructure with high availability and scalability
• Context-Aware Memory: Intelligent memory management with automatic context organization
• Advanced Retrieval: Sophisticated memory retrieval with semantic understanding
• Scalable Architecture: Handles large-scale memory operations with AWS Bedrock integration
• Security & Compliance: Enterprise-level security with AWS IAM integration

Configuration Requirements

To use AgentCore Memory as a memory provider, you’ll need:

• AWS Region: The AWS region where your AgentCore Memory is deployed (e.g., us-east-1, us-west-2)
• Access Key ID: Your AWS access key ID with appropriate permissions
• Secret Access Key: Your AWS secret access key for authentication
• Memory ID (Optional): Specific memory identifier if you want to use a particular memory instance

AWS Permissions Required

Your AWS credentials must have the following permissions for AgentCore Memory:

• bedrock:ListMemories - To list available memory instances
• bedrock:GetMemory - To retrieve memory content
• bedrock:CreateMemory - To create new memory instances (if needed)
• bedrock:UpdateMemory - To update existing memory instances (if needed)
• bedrock:DeleteMemory - To delete memory instances (if needed)

Best Practices for AgentCore Memory

• Region Selection: Choose the AWS region closest to your users for optimal performance
• IAM Roles: Use IAM roles instead of access keys when possible for better security
• Memory Organization: Use descriptive memory IDs to organize different types of memories
• Access Control: Implement proper IAM policies to control access to memory resources
• Monitoring: Enable AWS CloudWatch monitoring for memory usage and performance

Use Cases

Memory providers are particularly useful for:

• Conversational AI: Maintaining context across multiple user interactions
• Personalization: Storing user preferences and history
• Knowledge Retention: Remembering important information from previous conversations
• Session Continuity: Resuming conversations from where they left off
• Context Awareness: Providing more relevant responses based on conversation history

Security Considerations

• Ensure all memory provider URLs use HTTPS for secure communication
• Implement proper authentication through API keys and credentials
• Regularly review and audit memory provider permissions and access
• Monitor memory usage and performance metrics
• Consider data retention policies for stored information
• For AgentCore Memory, follow AWS security best practices and use IAM roles when possible

Troubleshooting

Common Issues

1. Connection Failed: Verify the URL is accessible and the service is running
2. Authentication Errors: Check that all required credentials are properly configured
3. Project/Org ID Issues: Ensure the Project ID and Organization ID are correct (for Mem0)
4. API Key Problems: Verify the API key is valid and has proper permissions
5. AWS Region Issues: Ensure the specified AWS region is correct and accessible (for AgentCore Memory)
6. AWS Permissions: Verify that your AWS credentials have the necessary Bedrock permissions (for AgentCore Memory)

Getting Help

If you encounter issues with memory provider configuration:

1. Verify your memory service credentials are correct
2. Check the memory service’s status and documentation
3. Ensure network connectivity between FloTorch and your memory service
4. For AgentCore Memory, verify AWS region and permissions
5. Contact support if issues persist

Using Memory in Agents

After configuring your memory providers, you can integrate them with your AI agents to enable persistent context and memory:

Agent Configuration

1. Navigate to the Agent Builder section in the left sidebar
2. Select or create an agent you want to configure
3. In the agent configuration, look for the Memory section
4. Assign your configured memory providers to the agent

Memory Integration Benefits

• Context Persistence: Agents maintain conversation context across sessions
• Long-term Memory: Store and retrieve important information over time
• Personalization: Remember user preferences and interaction history
• Session Continuity: Resume conversations seamlessly from where they left off

Example Use Cases

• Conversational AI: Maintain context across multiple user interactions
• Knowledge Management: Store and retrieve learned information
• User Preferences: Remember user settings and preferences
• Historical Context: Reference previous conversations and decisions

Integration with Agents

After configuring memory providers:

1. Assign to Agents: Configure agents to use specific memory providers in the Agent Builder
2. Test Memory: Verify that agents can store and retrieve information correctly
3. Monitor Usage: Track memory provider performance and usage patterns
4. Scale Up: Add more memory providers as your agent needs expand

Best Practices

• Naming Convention: Use consistent naming patterns for memory providers
• Documentation: Keep detailed descriptions of each memory provider’s purpose
• Regular Review: Periodically review and clean up unused memory providers
• Performance Monitoring: Track response times and storage usage
• Backup Strategy: Consider backup and recovery procedures for critical memory data
• Provider Selection: Choose the right memory provider based on your specific needs:
  • Mem0: Ideal for general-purpose memory with simple API integration
  • AgentCore Memory: Best for enterprise applications requiring AWS integration and advanced features