AgenticMem - Cross-Platform Memory Layer for AI Agents

Vision

Build a user-owned, cross-platform memory layer for AI agents that seamlessly unifies personal data across applications.

Description

This project explores building an interoperable memory layer that consolidates a user's memories in a single place so different AI agents can leverage richer context and produce better results. The design is inspired by existing systems such as Mem0, Zep, and LangMem.

Memory Ingestion Methods

Memories are primarily created and updated through two complementary methods:

• Foreground (Real-Time Extension): A browser extension observes user interactions on approved websites and updates relevant information immediately.
• Background (Periodic Sync): Memories are periodically refreshed using APIs for user-specified applications.

Storage Architecture

Different types of memories are persisted in specialized stores:

• Semantic & Contextual Memories: Stored in Azure Cosmos DB and retrieved via vector search.
• Episodic & Procedural Memories: Incrementally updated in a knowledge graph using Graphiti and Neo4j.

Both stores are queried when contextualization is required, and their results can be merged or ranked for downstream agent workflows.

Rationale

Current AI assistants lack persistent, structured, and user-controlled long-term context. A unified, portable memory layer can:

• Reduce redundant prompting and repeated user input.
• Enable personalization across heterogeneous agents and tools.
• Improve retrieval quality by combining semantic embeddings with graph-linked episodic structure.
• Support explainability and traceability of agent behavior.

Project Structure

/extension - Chrome Extension for Real-Time Memory Ingestion

Browser extension that captures user interactions and conversations in real-time across supported platforms.

• manifest.json: Extension configuration and permissions
• background.js: Service worker for extension lifecycle management
• popup.html/js: Extension popup interface for user controls
• src/teams/content.js: Microsoft Teams content script for memory retrieval and suggested replies
• src/chatgpt/content.js: ChatGPT content script for conversation capture
• src/services/auth.service.js: Authentication service for secure API communication

/server - Django Backend API

RESTful API server providing memory storage, retrieval, and processing capabilities.

• memories/: Core memory management Django app
  • models.py: Memory and summary data models
  • views.py: API endpoints for memory CRUD operations and retrieval
  • azure_openai.py: Azure OpenAI integration for embeddings and LLM calls
  • cosmos_db.py: Azure Cosmos DB connection and vector search
  • graphiti_client.py: Graphiti knowledge graph integration
• authentication/: User authentication and authorization
• requirements.txt: Python dependencies
• manage.py: Django management script

/teams_app - Microsoft Teams Application

Native Teams application for enhanced integration and bot capabilities.

• src/: TypeScript source code for Teams app logic
• appPackage/: Teams app manifest and assets
• infra/: Azure infrastructure as code (Bicep templates)

Getting Started

Prerequisites

• Python 3.8+
• Node.js 16+
• Azure Cosmos DB account
• Azure OpenAI service
• Chrome browser for extension development

Installation

1. Clone the repository

   git clone https://github.com/srwshkrshnn/AgenticMem.git
   cd AgenticMem

2. Set up the backend server

   cd server
   pip install -r requirements.txt
   python manage.py migrate
   python manage.py runserver

3. Install the Chrome extension

   • Open Chrome and navigate to chrome://extensions/
   • Enable "Developer mode"
   • Click "Load unpacked" and select the /extension folder

4. Configure environment variables

   • Copy .env.example to .env in the server directory
   • Add your Azure Cosmos DB and OpenAI credentials

Usage

1. Extension Usage: Install the Chrome extension and visit supported platforms (Teams, ChatGPT)
2. Memory Retrieval: The extension automatically suggests contextual replies based on conversation history
3. API Access: Use the REST API endpoints for programmatic memory management