Open Source Features Fresh

Production-ready capabilities for self-hosted Mem0 deployments.

Feature Overview

Feature	Description
Graph Memory	Store and recall entity relationships across multiple hops
Metadata Filtering	Query using logical operators with nested conditions
Search with Reranking	Enhanced relevance through specialized ranking models
Async Operations	Non-blocking calls for high-volume applications
Multimodal Support	Handle images, audio, and video as memory inputs
Custom Fact Extraction	Modify how text becomes structured facts
Custom Memory Updates	Apply personalized refinement instructions
REST API	HTTP-based integration for any language
OpenAI Compatibility	Drop-in replacement for OpenAI chat endpoints

Graph Memory

Pairs vector embeddings with graph structures to preserve entity relationships.

Setup

pip install "mem0ai[graph]"

Configuration

from mem0 import Memory

config = {
    "graph_store": {
        "provider": "neo4j",
        "config": {
            "url": "bolt://localhost:7687",
            "username": "neo4j",
            "password": "password"
        }
    },
    "vector_store": {
        "provider": "qdrant",
        "config": {
            "collection_name": "memories",
            "host": "localhost",
            "port": 6333
        }
    }
}

memory = Memory.from_config(config)

Supported Graph Backends

• Neo4j / Neo4j Aura (managed)
• Memgraph (Docker)
• AWS Neptune Analytics / Neptune DB
• Kuzu (embedded)
• Apache AGE (PostgreSQL extension)

Features

• Custom extraction prompts for entity/relationship extraction
• Confidence thresholds to filter low-confidence edges
• Per-request toggles for vector-only fallback
• Multi-agent organization using user_id, agent_id, run_id

Metadata Filtering

results = memory.search(
    "preferences",
    user_id="alice",
    filters={"category": {"in": ["food", "travel"]}}
)

Async Operations

from mem0 import AsyncMemory

memory = AsyncMemory.from_config(config)
await memory.add(messages, user_id="alice")
results = await memory.search("preferences", user_id="alice")

Multimodal Support

Pass images, audio, or video alongside text for memory extraction.

Custom Fact Extraction

config["custom_fact_extraction_prompt"] = """
Extract: goals, preferences, constraints
Exclude: greetings, filler, casual chat
Return JSON with key "facts" as list of strings.
"""
memory = Memory.from_config(config)

REST API

Run Mem0 as an HTTP server for any language integration.

OpenAI Compatibility

Use Mem0 as a drop-in replacement for OpenAI's chat completions endpoint, with automatic memory management.