InMemory - Enhanced Memory Management for AI

🧠 Long-term memory for AI Agents with zero-setup simplicity

⚡ Zero Setup • 🚀 Instant Library • 💼 REST API Ready

🔥 Key Features

• 🚀 Zero Setup: pip install inmemory and start using immediately
• 🏗️ Dual Architecture: Local Memory class + Managed InmemoryClient
• 🔍 Advanced Search: Semantic similarity with ChromaDB embeddings
• 🌐 Two Usage Modes: Direct library usage OR REST API server
• 💼 Dashboard Ready: MongoDB authentication + clean REST endpoints

🚀 Quick Start

Zero-Setup Library Usage

pip install inmemory

from inmemory import Memory

# Works immediately - no setup required!
memory = Memory()

# Add memories with metadata
memory.add(
    "I love pizza but hate broccoli",
    tags="food,preferences"
)

memory.add(
    "Meeting with Bob and Carol about Q4 planning tomorrow at 3pm",
    tags="work,meeting",
    people_mentioned="Bob,Carol",
    topic_category="planning"
)

# Search memories
results = memory.search("pizza")
for result in results["results"]:
    print(f"Memory: {result['content']}")
    print(f"Score: {result['score']}")

# Health check
health = memory.health_check()
print(f"Status: {health['status']}")

Managed Client Usage (Dashboard Integration)

from inmemory import InmemoryClient

# Connect to managed service
client = InmemoryClient(
    api_key="your_api_key",
    host="http://localhost:8081"
)

# Same API as Memory, but with authentication
client.add("Meeting notes from dashboard", tags="dashboard")
results = client.search("meeting notes")

REST API Server Mode

# Start the server (from inmemory-core directory)
cd server/
python main.py

# Or with custom configuration
MONGODB_URI=mongodb://localhost:27017/inmemory python main.py

Server runs on http://localhost:8081 with endpoints:

• POST /v1/memories - Add memory
• GET /v1/memories - Get all memories
• POST /v1/search - Search memories
• DELETE /v1/memories/{id} - Delete memory

📦 Installation Options

Mode	Command	Dependencies	Use Case
Basic SDK	pip install inmemory	Zero external deps	Development, testing, simple apps
API Server	pip install inmemory[server]	FastAPI, Uvicorn	Integration, dashboards
Enterprise	pip install inmemory[enterprise]	MongoDB, OAuth	Production, multi-user
Full	pip install inmemory[full]	Everything + MCP	Complete installation

🏗️ Architecture

┌─────────────────────────────────────────────────────────────┐
│                     InMemory Package                         │
├─────────────────────────────────────────────────────────────┤
│  SDK Layer    │ Memory Class (Primary Interface)            │
│  API Layer    │ FastAPI Server (Optional)                   │
│  Storage Layer│ File (Default) │ MongoDB (Enterprise)       │
│  Search Layer │ Enhanced Search Engine + Qdrant            │
└─────────────────────────────────────────────────────────────┘

💡 Core API Reference

Memory Class

from inmemory import Memory

# Initialize with different backends
memory = Memory()                        # Auto-detect (file by default)
memory = Memory(storage_type="file")     # Force file storage
memory = Memory(storage_type="mongodb")  # Force MongoDB (requires deps)

# Memory operations
result = memory.add(content, user_id, tags=None, people_mentioned=None, topic_category=None)
results = memory.search(query, user_id, limit=10, tags=None, temporal_filter=None)
memories = memory.get_all(user_id, limit=100)
result = memory.delete(memory_id, user_id)

# Advanced search
results = memory.search_by_tags(["work", "important"], user_id, match_all=True)
results = memory.search_by_people(["Alice", "Bob"], user_id)
results = memory.temporal_search("yesterday", user_id, semantic_query="meetings")

# User management
result = memory.create_user(user_id, email="user@example.com")
api_key = memory.generate_api_key(user_id, name="my-app")
keys = memory.list_api_keys(user_id)
stats = memory.get_user_stats(user_id)

Configuration

from inmemory import InMemoryConfig, Memory

# Custom configuration
config = InMemoryConfig(
    storage={
        "type": "file",           # or "mongodb"
        "path": "~/my-memories"   # for file storage
    },
    auth={
        "type": "simple",         # or "oauth", "api_key"
        "default_user": "my_user"
    },
    qdrant={
        "host": "localhost",
        "port": 6333
    }
)

memory = Memory(config=config)

🌐 REST API Endpoints

When running in server mode (inmemory serve), these endpoints are available:

Method	Endpoint	Description
POST	/v1/memories	Add new memory
GET	/v1/memories	Get user's memories
DELETE	/v1/memories/{id}	Delete specific memory
POST	/v1/search	Search memories
POST	/v1/temporal-search	Temporal search
POST	/v1/search-by-tags	Tag-based search
POST	/v1/search-by-people	People-based search
GET	/v1/health	Health check

🔧 Configuration Options

Environment Variables

# Storage backend
export INMEMORY_STORAGE_TYPE="file"           # or "mongodb"
export INMEMORY_DATA_DIR="~/.inmemory"        # for file storage
export MONGODB_URI="mongodb://localhost:27017/inmemory" # for mongodb

# Server settings
export INMEMORY_HOST="0.0.0.0"
export INMEMORY_PORT="8081"

# Qdrant settings
export QDRANT_HOST="localhost"
export QDRANT_PORT="6333"

YAML Configuration

Create ~/.inmemory/config.yaml:

storage:
  type: "file"              # or "mongodb"
  path: "~/.inmemory/data"

auth:
  type: "simple"            # or "oauth", "api_key"
  default_user: "user123"

qdrant:
  host: "localhost"
  port: 6333

embedding:
  provider: "ollama"
  model: "nomic-embed-text"
  ollama_host: "http://localhost:11434"

🚀 Deployment

Single File Deployment

# Just run the server - file storage included
inmemory serve --port 8080

Docker Deployment

# Simple mode (file storage)
docker run -p 8080:8080 -v inmemory-data:/root/.inmemory inmemory:latest

# Enterprise mode (MongoDB)
docker-compose up  # Uses provided docker-compose.yml

Production Deployment

# Enterprise mode with MongoDB
export MONGODB_URI="mongodb://prod-mongo:27017/inmemory"
export GOOGLE_CLIENT_ID="your-prod-client-id"
export GOOGLE_CLIENT_SECRET="your-prod-client-secret"

inmemory serve --host 0.0.0.0 --port 8080

🔄 Migration Between Modes

Easily migrate from simple file storage to enterprise MongoDB:

from inmemory.stores import FileBasedStore, MongoDBStore

# Initialize both backends
file_store = FileBasedStore()
mongo_store = MongoDBStore(mongodb_uri="mongodb://localhost:27017")

# Migrate all data
success = mongo_store.migrate_from_file_store(file_store)
print(f"Migration {'successful' if success else 'failed'}!")

🧪 Development & Testing

# Install with development tools
pip install inmemory[dev]

# Run tests
inmemory test

# Check configuration
inmemory config

# View storage statistics
inmemory stats

# Initialize with sample data
inmemory init

🤝 Integration Examples

Personal AI Assistant

from inmemory import Memory
from openai import OpenAI

class PersonalAssistant:
    def __init__(self):
        self.memory = Memory()
        self.llm = OpenAI()

    def chat(self, user_input: str, user_id: str) -> str:
        # Get relevant memories
        memories = self.memory.search(user_input, user_id=user_id, limit=5)
        context = "\n".join([m['memory'] for m in memories['results']])

        # Generate response with context
        response = self.llm.chat.completions.create(
            model="gpt-4o-mini",
            messages=[
                {"role": "system", "content": f"Context: {context}"},
                {"role": "user", "content": user_input}
            ]
        )

        # Store conversation
        self.memory.add(f"User: {user_input}", user_id=user_id)
        self.memory.add(f"Assistant: {response.choices[0].message.content}", user_id=user_id)

        return response.choices[0].message.content

Customer Support Bot

from inmemory import Memory

class SupportBot:
    def __init__(self):
        self.memory = Memory()

    def handle_ticket(self, customer_id: str, issue: str):
        # Check customer history
        history = self.memory.search_by_people([customer_id], user_id="support")
        similar_issues = self.memory.search(issue, user_id="support", limit=3)

        # Generate contextual response based on history
        response = self.generate_response(issue, history, similar_issues)

        # Store interaction
        self.memory.add(
            f"Customer {customer_id} reported: {issue}",
            user_id="support",
            tags="ticket,customer_support",
            people_mentioned=customer_id,
            topic_category="support"
        )

        return response

📚 Documentation

• Installation Guide: Detailed installation and usage
• Architecture Plan: Technical architecture details
• API Reference: Interactive API documentation (when server running)

🏢 Enterprise Features

For enterprise deployments, InMemory provides:

• Multi-user Support: MongoDB backend with user isolation
• OAuth Integration: Google OAuth for dashboard authentication
• Scalable Storage: MongoDB collections per user
• API Key Management: Secure key generation and management
• Dashboard Ready: REST API for your private dashboard integration

🤖 MCP Server Integration

InMemory works seamlessly with MCP (Model Context Protocol) for AI agent integration:

# Separate repository for MCP server
git clone https://github.com/you/inmemory-mcp
cd inmemory-mcp
pip install -e .

# Configure to connect to any InMemory API
export INMEMORY_API_URL="http://localhost:8080"
python src/server.py

🛠️ Requirements

Minimal Installation

• Python: 3.10+ (supports Python 3.10, 3.11, 3.12, 3.13)
• Qdrant: Vector database for embeddings
• Ollama: Local embeddings (or OpenAI API key)

Enterprise Installation

• MongoDB: User management and authentication
• Google OAuth: Dashboard authentication

🎯 Roadmap

• Storage Abstraction: File-based and MongoDB backends
• CLI Tools: Easy server management
• PostgreSQL Backend: Alternative to MongoDB
• TypeScript SDK: Cross-language support
• More Vector DBs: Chroma, Pinecone integration
• Cloud Storage: S3, GCS backends

🤝 Contributing

We welcome contributions! Please see:

• Issues: Report bugs and request features
• Pull Requests: Follow our coding standards (ruff, pre-commit)
• Documentation: Help improve our guides

# Development setup
git clone https://github.com/you/inmemory
cd inmemory
pip install -e .[dev]
pre-commit install

# Run tests
inmemory test
pytest

📄 License

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.

🙏 Acknowledgments

• FastAPI: Excellent API framework
• Qdrant: High-performance vector database
• Pydantic: Data validation and configuration

---

Start simple. Scale seamlessly. 🚀