Memori

The memory fabric for enterprise AI.

Memori adds persistent memory to your LLM applications without changing your architecture. It is model, framework, and datastore agnostic.

---

Why Memori

Memori captures LLM interactions, enriches them, and makes them retrievable as high-quality context for future generations.

• Low integration overhead: wrap your existing LLM client and keep your current stack.
• Attribution-aware memory: organize memory by entity, process, and session.
• Asynchronous augmentation: extract structured memory without adding user-facing latency.
• Flexible infrastructure: supports multiple models, frameworks, and databases.

Install

pip install memori

Optional one-time optimization:

python -m memori setup

Quickstart (OpenAI + SQLite)

import os
import sqlite3

from memori import Memori
from openai import OpenAI


def get_sqlite_connection():
    return sqlite3.connect("memori.db")


client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
memori = Memori(conn=get_sqlite_connection).llm.register(client)

# Required so Memori can store contextual memory for this actor/workflow.
memori.attribution(entity_id="user_123", process_id="assistant_demo")
memori.config.storage.build()

client.chat.completions.create(
    model="gpt-4.1-mini",
    messages=[{"role": "user", "content": "My favorite color is blue."}],
)

# Wait for async augmentation in short-lived scripts.
memori.augmentation.wait()

response = client.chat.completions.create(
    model="gpt-4.1-mini",
    messages=[{"role": "user", "content": "What is my favorite color?"}],
)
print(response.choices[0].message.content)

Core Concepts

Attribution

Attribution tells Memori who the interaction belongs to and what workflow produced it.

memori.attribution(entity_id="user_123", process_id="support_agent")

If attribution is not set, memory cannot be built correctly.

Session Management

Sessions group related interactions together.

memori.new_session()

Reuse an existing session:

session_id = memori.config.session_id
memori.set_session(session_id)

Schema Setup

Build or migrate storage schema during deployment or after upgrades:

Memori(conn=db_session_factory).config.storage.build()

Supported Integrations

LLM Providers

• OpenAI
• Anthropic
• Google Gemini
• xAI (Grok)
• Bedrock (via LangChain)

Supports sync, async, streamed, and unstreamed interaction modes.

Frameworks and Platforms

• LangChain
• Agno
• Pydantic AI
• Nebius AI Studio

Database Connection Methods

• SQLAlchemy
• DB API 2.0 (PEP 249 drivers such as psycopg, pymysql, sqlite3, and others)
• Django ORM integration

Datastores

• SQLite
• PostgreSQL
• MySQL
• MariaDB
• Oracle
• MongoDB
• Neon
• Supabase
• CockroachDB

Advanced Augmentation

Memori can enrich captured conversations into structured memory such as:

• attributes
• events
• facts
• people
• preferences
• relationships
• rules
• skills

Augmentation runs asynchronously and is available without an account (rate limited).

Get higher limits:

python -m memori sign-up <email_address>

Set your API key:

export MEMORI_API_KEY=<api_key>

Check usage quota:

python -m memori quota

CLI

python -m memori

See full CLI docs in docs/cli.md.

Contributing

Contributions are welcome. See CONTRIBUTING.md for setup, standards, and PR guidance.

License

Apache 2.0. See LICENSE.