Repolode
Submit
nowledge-co

nowledge-co/hawdb

An embedded Rust database for AI agents that stores graph, relational, text and vector data in one crash-safe engine and returns bounded context.

View on GitHubWebsite
Stars
203
Forks
13
Open issues
138
Last commit
7 hours ago

What is it?

HawDB is an embedded database for agents. It keeps entities, relationships, documents and events in one crash-safe process, and it is designed to hand back bounded context rather than unbounded result sets. Graph structure, relational rows, lexical search and vector search share a single engine, and a retrieval request runs through a fixed pipeline of search, filtering, authorized graph expansion, reranking and hydration from a pinned snapshot.

Agents query it with parameterized Cypher and PostgreSQL-dialect SQL. Every read has an explicit row and payload budget, and a request that exceeds it fails rather than returning a silently truncated answer. Copy-on-write branches let an agent work on its own durable copy of the data without copying the database, so a plan can be tried without disturbing the main branch or other agents.

The library is written in Rust under Apache-2.0, with Python bindings published as the hawdb package on PyPI. It is built for Nowledge Mem and is intended to be linked into any host that needs a durable context layer.

Key features

  • 01Graph, relational, full-text and vector search share one embedded engine.
  • 02Parameterized Cypher and PostgreSQL-dialect SQL on the same handle.
  • 03Explicit row and payload budgets; over-budget reads fail instead of truncating.
  • 04Copy-on-write branches for agents, created without copying the parent data.
  • 05Crash recovery through a write-ahead log, with one writer process per database directory.
  • 06Python bindings available on PyPI as hawdb.

Use cases

  • Giving an AI agent a durable, queryable memory of entities and relationships.
  • Letting an agent test a multi-step plan on an isolated branch before committing results.
  • Embedding a context store with bounded retrieval into a Rust or Python application.

Quick start

  1. 1

    Add the hawdb crate to Cargo.toml

  2. 2

    Build the default library from a checkout

    cargo build --locked -p hawdb

Things to know

  • Rust library only; applications must integrate it through Rust or the in-tree Python bindings.
  • Early-stage project (version 0.25) with 197 stars, and the README says it is in nightly integration for Nowledge Mem.
  • A database directory can be owned by only one process at a time.

In collections

Open-source AI agent frameworks & apps

Topics

#graph-database#embedded-database#vector-search#rust#agents#knowledge-graph#cypher#graph#graph-algorithms#memory-database

FAQ

What is hawdb?
An embedded Rust database for AI agents that stores graph, relational, text and vector data in one crash-safe engine and returns bounded context. HawDB is an embedded database for agents. It keeps entities, relationships, documents and events in one crash-safe process, and it is designed to hand back bounded context rather than unbounded result sets. Graph structure, relational rows, lexical search and vector search share a single engine, and a retrieval request runs through a fixed pipeline of search, filtering, authorized graph expansion, reranking and hydration from a pinned snapshot.
Is hawdb free to use? What license does it have?
hawdb is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is hawdb still maintained?
Actively maintained. The last commit to hawdb was on Nov 7, 2026.
What can I use hawdb for?
Giving an AI agent a durable, queryable memory of entities and relationships. Letting an agent test a multi-step plan on an isolated branch before committing results. Embedding a context store with bounded retrieval into a Rust or Python application.
How do I install hawdb?
Add the hawdb crate to Cargo.toml Build the default library from a checkout: cargo build --locked -p hawdb
Where can I get hawdb?
The source code and installation instructions are on GitHub: https://github.com/nowledge-co/hawdb

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects