Skip to content

Quickstart

Setup

With uv (recommended, reproducible via the committed uv.lock):

git clone https://github.com/TeleAI-UAGI/telemem.git
cd telemem
uv sync --all-extras

Or with pip:

pip install -e .            # core (text memory)
pip install -e ".[all]"     # + MCP server + video pipeline

First memory

export OPENAI_API_KEY=sk-...

import telemem as mem0

memory = mem0.Memory()

messages = [
    {"role": "user", "content": "Jordan, did you take the subway to work again today?"},
    {"role": "assistant", "content": "Yes, James. The subway is much faster than driving."},
    {"role": "user", "content": "Which station is closest?"},
    {"role": "assistant", "content": "Line 2 to Civic Center Station, Exit A, 5 minutes on foot."},
]

memory.add(messages=messages, user_id="Jordan")
results = memory.search("How does Jordan get to work?", user_id="Jordan")
for hit in results["results"]:   # same result shape as mem0
    print(hit["memory"])

Character isolation

Every user_id gets an independent memory profile; conversation-level events are stored under the shared pseudo-user "events" and searched automatically. In a two-character dialogue, search(..., user_id="Jordan") only sees Jordan's profile plus shared events — never another character's private memories.

Local, no-cloud setup

Use the repository's Qwen + FAISS config, or the fully local Ollama config:

TELEMEM_CONFIG=config/config.ollama.yaml python examples/quickstart.py

See LLM Providers for all supported backends.

Framework integrations

The integration pattern is two calls: search() before answering, add() after each exchange.