Skip to content

We're open source. If actrone-memory has been useful to you, a star on GitHub means a lot to us.

Installation

Install the open-source, self-hosted memory library for your language: actrone-memory, from PyPI for Python or from npm for TypeScript.

Python: actrone-memory (self-hosted)

The actrone-memory package is the open-source, self-hosted two-tier memory library. MemoryManager.create() works out of the box with zero services (an in-memory store and a local embedder), with no Actrone account, Redis, Qdrant, or API key required to start. The bare install recalls by shared keywords; add the [onnx] extra to recall by meaning with a small local model (about 130 MB, downloaded once).

bash
pip install actrone-memory

# With uv
uv add actrone-memory

Optional extras: recall by meaning, infrastructure for production, plus 16 framework integrations (see Framework integrations for the full list). The quotes matter in zsh, the macOS default shell:

bash
pip install "actrone-memory[onnx]"        # recall by meaning, runs locally (recommended)
pip install "actrone-memory[production]"  # redis + qdrant + openai, bundled
pip install "actrone-memory[redis]"       # Redis L1 store only
pip install "actrone-memory[qdrant]"      # Qdrant L2 store only
pip install "actrone-memory[openai]"      # OpenAI embeddings
pip install "actrone-memory[local]"       # sentence-transformers embedder (heavier: pulls in torch)
pip install "actrone-memory[tiktoken]"    # exact token counts instead of the ~4-characters heuristic
pip install "actrone-memory[langchain]"   # one framework extra, of 16 available
pip install "actrone-memory[all]"         # every extra, including all 16 integrations

Tip

actrone-memory add <framework> scaffolds an integration snippet for any of the 16 supported frameworks straight from the terminal, see Framework integrations.

TypeScript: actrone-memory (self-hosted)

The actrone-memory package is the open-source, self-hosted two-tier memory library for JavaScript and TypeScript, the counterpart to actrone-memory, with the same two-tier model and result shapes. It has one runtime dependency (zod): an in-memory store and a local embedder work out of the box; Redis and Qdrant adapters plug in for production. Install fastembed alongside it to recall by meaning with a local model (about 130 MB, downloaded once); without it, recall is by shared keywords.

bash
npm install actrone-memory
npm install fastembed     # optional: recall by meaning, runs locally
# or
pnpm add actrone-memory
# or
yarn add actrone-memory

Every framework integration (the framework-agnostic memoryFor core plus 11 named adapters, from vercelMemory to LangChain.js, LangGraph.js, Mastra, LlamaIndex.TS, and more) lives in the actrone-memory/adapters entry point (see Framework integrations). They're structural (duck-typing the framework's own primitives) against optional peer dependencies, so there are no separate install extras, bring whichever framework you already use.

Tip

Requires Node.js 22 or later. Redis + Qdrant are optional, only needed for the production stores.

Requirements

PackageRuntimeRequired infrastructure
actrone-memory (PyPI)Python ≥ 3.11None to start (in-memory + local embedder); Redis + Qdrant + OpenAI optional for production
actrone-memory (npm)Node ≥ 22None to start (in-memory + local embedder); Redis + Qdrant optional for production

What changed in each release of actrone-memory and create-actrone-app is in the changelog.

Verifying the install (Python)

bash
# Memory library (Python)
python -c "from actrone_memory import MemoryManager; print('ok')"

Verifying the install (TypeScript)

bash
# Memory library (TypeScript)
node --input-type=module -e "import { MemoryManager } from 'actrone-memory'; console.log('ok')"
Was this page helpful?