# WritersLogic Adaptive Retention & Memory (WARM)

Hybrid search, graph indexing, and vector-symbolic memory.

> Local, encrypted, provenance-aware memory for AI agents. WARM is an open-source Rust engine for search and memory that runs inside your app, on your machine. It makes no service calls and sends no telemetry.

WARM is made by WritersLogic, Inc. (https://writerslogic.com).

## What it is

- Embeddable memory for AI agents: hybrid search (BM25 plus the embeddings you supply), vector-symbolic fact memory, optional encryption at rest and signed provenance, in one Rust library.
- Runs in-process. Embeddings are supplied by you or computed by local models from weights you provide; it never downloads a model.
- Python applications install warm-sdk 0.1.3 from PyPI; it is an HTTP client for warm-server.
- WARM Cloud (in development) adds encrypted sync, backup and independent verification. Memory stays on your machines.

## License

Current WARM releases are Apache-2.0. Historical releases under the former names retain their original license terms.

## Links

- [Product site](https://writerslogic.com/warm/)
- [Source code](https://github.com/writerslogic/warm)
- [Documentation](https://writerslogic.com/warm/docs/)
- [Quickstart](https://writerslogic.com/warm/docs/latest/quickstart/)
- [Benchmark results](https://github.com/writerslogic/warm/tree/main/benchmarks/results)
- [crates.io](https://crates.io/crates/warm-rust)
- [Python SDK: warm-sdk 0.1.3](https://pypi.org/project/warm-sdk/0.1.3/)
- [Privacy model](https://github.com/writerslogic/warm/blob/main/docs/PRIVACY.md)

Python distribution names: warm-python (import writerslogic_warm), warm-sdk (import writerslogic_warm_sdk), and warm-models (import writerslogic_warm_models). warm-python 0.6.4, warm-sdk 0.1.3 and warm-models 0.1.0 are published on PyPI. Native Python provides vector search and phase-vector operations, not document search. Node.js: npm install writerslogic-warm@0.6.4. Rust: cargo add warm-rust.