Project status
GraphBuff is at version 0.2.0. It is free and open source, it is new, and this page says plainly what you can rely on today.
License
Licensed under the GNU Affero General Public License v3.0 (AGPL-3.0-only). Free and open source: use it, change it, use it commercially. If you distribute GraphBuff, or a program built on it, or let users reach a modified version over a network, you must make the corresponding source available under the same license.
Version and releases
The engine is at version 0.2.0. The latest release is v0.2.0, from 2026-11-15. There is no release schedule; work lands on the main branch. Releases appear in the changelog and its feed.
What is done and tested
- Storage
- In-memory, hybrid (write-ahead log and snapshots) and disk-resident paged storage, with recovery and format-upgrade tests.
- Cypher
- The clauses, expressions, functions and procedures listed in the documentation. Anything else fails with an Unsupported error, never with a wrong answer.
- Transactions
- Snapshot reads and explicit transactions in every storage mode.
- Server
- HTTP/JSON and Bolt 4.4 and 5.0 to 5.4, tested with the official Python and Java drivers.
- Operations
- Users and roles, TLS, bulk CSV import, online backup, restore and consistency checks.
- Platforms
- Windows, Linux and macOS.
What is experimental or missing
- GPU backend
- An optional build feature. Multi-hop walks run 7.6 to 9.6 times faster than on the CPU, short of the project's own goal of 10 times; filters and aggregates still run on the CPU.
- Pure-Rust TLS
- The tls-rustcrypto build uses a pre-release crypto provider. The default tls build uses ring.
- Packaging
- The Docker image and the prebuilt binaries are plain builds: the GPU backend and TLS need a build from source. The image is built for x86-64 and arm64; the Linux binaries for x86-64 only.
- Storage format
- New format versions keep reading the previous one, but the project is at version 0.2.0 and makes no long-term compatibility promise yet.
Cypher that is not supported yet is listed in the documentation.
What is planned
The engine's roadmap holds the plan and the exit criterion of each stage; its progress log tracks every open item. Open today: GPU follow-ups (filters and aggregates on the device, a GPU-aware cost model) and ongoing storage, benchmark and tooling work.
Where the claims come from
Performance numbers: Benchmarks. Features: the documentation, generated from the engine README. Query outputs on the use-case pages: produced by the engine. Reporting a vulnerability: Security.