cortex-memory-client

Python client SDK for the Cortex graph memory engine. The distribution is named cortex-memory-client; the import package remains cortex_memory.

Installation

pip install cortex-memory-client

Quick start

from cortex_memory import Cortex

# Connect to a running server
with Cortex("localhost:9090") as cx:
    # Store knowledge
    node_id = cx.store(
        "decision",
        "Use FastAPI for the backend",
        body="Chose FastAPI over Flask for async support and type hints",
        tags=["backend", "python"],
        importance=0.8,
    )

    # Semantic search
    results = cx.search("backend technology choices", limit=5)
    for r in results:
        print(f"{r.score:.2f} — {r.title}")

    # Get a briefing
    briefing = cx.briefing("my-agent")
    print(briefing)  # Ready-to-inject markdown

Library mode (embedded server)

cx = Cortex.open("./memory.redb")
# Works identically — starts a local server subprocess

Testing

from cortex_memory.testing import mock_cortex

def test_my_agent():
    with mock_cortex() as cx:
        cx.store("fact", "test data")
        results = cx.search("test")
        assert len(results) == 1
        cx.assert_stored("fact", "test data")

pytest fixture

import pytest
from cortex_memory.testing import mock_cortex

@pytest.fixture
def cortex():
    with mock_cortex() as cx:
        yield cx

def test_store_and_search(cortex):
    cortex.store("decision", "Use FastAPI", body="Async support", importance=0.8)
    results = cortex.search("FastAPI")
    assert results[0].title == "Use FastAPI"

Proto generation

The cortex_pb2.py and cortex_pb2_grpc.py stubs are pre-generated from crates/cortex-proto/proto/cortex.proto. To regenerate after a proto change:

pip install grpcio-tools
./scripts/generate_protos.sh