Skip to content

Latest commit

 

History

History
252 lines (180 loc) · 8.05 KB

STARTER_KIT.md

File metadata and controls

252 lines (180 loc) · 8.05 KB

EverMemOS Competition Starter Kit

Home > Docs > Starter Kit

Welcome to the EverMemOS Competition! This starter kit will help you get up and running quickly so you can focus on building amazing AI memory applications.

Join our Discord to brainstorm ideas, find teammates, and get help from the community!

Table of Contents

Quick Start (5 Minutes)

1. Clone and Start Services

# Clone the repository
git clone https://github.com/EverMind-AI/EverMemOS.git
cd EverMemOS

# Start all services with Docker
docker compose up -d

# Verify services are running
curl http://localhost:1995/health

2. Store Your First Memory

curl -X POST "http://localhost:1995/api/v1/memories" \
  -H "Content-Type: application/json" \
  -d '{
    "message_id": "msg_001",
    "create_time": "2025-01-20T10:00:00+00:00",
    "sender": "user_001",
    "content": "I love building AI applications with memory!"
  }'

3. Search Memories

curl -X GET "http://localhost:1995/api/v1/memories/search" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "What does the user love?",
    "user_id": "user_001",
    "retrieve_method": "hybrid"
  }'

You're ready to build!

Competition Tracks

Track 1: Agent + Memory

Build intelligent agents with long-term, evolving memories.

Idea Description
Personal Digital Twin AI that learns your preferences, habits, and communication style over time
Content Creator Copilot Assistant that remembers your brand voice, past content, and audience feedback
CRM Copilot Sales assistant with full customer interaction history and relationship context
Therapy/Health Companion Supportive AI that tracks emotional patterns and wellness journey
Education AI Tutor that adapts to learning style and remembers progress across subjects
Multi-Agent Collaboration Team of specialized agents sharing knowledge through common memory
Customer Support Bot Support agent with full case history and user preference memory
Game NPC System NPCs with persistent memory of player interactions and world events

Track 2: Platform Plugins

Develop plugins or SDKs to integrate EverMemOS everywhere developers work.

Idea Description
VSCode Extension Remember coding context, past solutions, and project decisions
Chrome Extension Capture and recall browsing context, research sessions, and bookmarks
Slack/Discord Bot Team memory bot that tracks decisions, action items, and discussions
Notion Integration Sync notes and documents with conversational memory
Obsidian Plugin Connect personal knowledge base with AI memory
LangChain Memory Backend Drop-in memory class for LangChain applications
LlamaIndex Integration Memory store connector for LlamaIndex pipelines
CLI Tool Command-line interface for memory operations

Track 3: OS Infrastructure

Optimize the EverMemOS platform: core functionality, performance tuning, and system improvements.

Idea Description
Custom Retrieval Methods Graph-based, time-weighted, or multi-modal retrieval strategies
Memory Extractors Domain-specific extractors (medical, legal, technical)
Performance Optimization Query optimization, caching strategies, batch processing
New Benchmarks Evaluation datasets and metrics for specific use cases
Emotion-Aware Tagging Extract and tag emotional context from conversations
Action Item Extraction Automatically identify TODOs and follow-ups from conversations
Multi-Language Support Improve memory extraction for non-English languages
Privacy Features Memory anonymization, selective forgetting, access controls

How to Submit

Note: A submission portal will be available soon. Stay tuned for updates!

Required

Item Description
GitHub Repository Public repo with your project code
README.md Clear project introduction, setup instructions, and how EverMemOS is used
Video Demo 3-5 minute video demonstrating your project and explaining the concept

Optional (Preferred)

Item Description
Live Demo Deployed application or interactive demo link

Submission Checklist

  • GitHub repository is public and accessible
  • README includes project overview and setup instructions
  • README explains how EverMemOS powers your solution
  • Video demo is 3-5 minutes and covers functionality + concept
  • (Optional) Live demo URL included in README

Evaluation Criteria

Projects will be evaluated based on the following priorities:

Innovation

We value creative and original approaches. Does your project solve a problem in a new way? Is the use of memory capabilities imaginative? Does it stand out from existing solutions?

Technical Depth

We look for well-engineered projects with clean implementation, effective use of EverMemOS features, and appropriate technical sophistication for the problem being solved.

Consumer Value

We prioritize projects that solve real problems for real users. Is it useful? Is it intuitive? Could it scale or be commercialized?

API Cheatsheet

Core Endpoints

Method Endpoint Description
POST /api/v1/memories Store a message
GET /api/v1/memories Fetch memories by type
GET /api/v1/memories/search Search memories
DELETE /api/v1/memories Delete memories

Key Parameters

Parameter Values Description
retrieve_method keyword, vector, hybrid, rrf, agentic Search strategy
memory_types episodic_memory, profile, foresight, event_log Memory categories
role user, assistant Message sender type

Response Structure

{
  "status": "ok",
  "message": "Success description",
  "result": {
    // Response data here
  }
}

Sample Data

Pre-loaded sample conversations are available:

File Description
data/group_chat_en.json English group chat example
data/group_chat_zh.json Chinese group chat example

Load sample data:

uv run python src/bootstrap.py src/run_memorize.py \
  --input data/group_chat_en.json \
  --scene group_chat \
  --api-url http://localhost:1995/api/v1/memories

Example Projects

Project Description
Game of Thrones Demo Interactive demo comparing memory-augmented AI vs vanilla LLM responses using "A Game of Thrones" book content

Resources

Documentation

Examples

External

Tips for Success

  1. Start Simple - Get the basic flow working before adding complexity
  2. Use Hybrid Retrieval - Best balance of speed and accuracy for most use cases
  3. Group Your Memories - Use group_id for multi-user or multi-agent scenarios
  4. Check Pending Messages - Memories may not extract immediately (boundary detection)
  5. Ask for Help - Join our Discord for real-time support

Good luck and happy building!