mrkeyoor.com_
Tue 01 Sept 17:44 UTC
Dataevaluationupdated 30 Aug 2026

FalkorDB review

FalkorDB is a graph database that stores connected data as nodes and relationships and queries it with OpenCypher. It runs as a Redis module, uses sparse matrices for graph operations, and targets knowledge graphs, fraud analysis, security data, and application memory.

+49 / 2dstars / 7d
Verdict

Our FalkorDB build failed after 116 seconds and its tests failed after 7 seconds because deps/RediSearch/bin was missing. Use the published image for a product trial, and use the recommended development container if you plan to contribute. FalkorDB deserves a shortlist for Redis-friendly OpenCypher work, but SSPL-1.0, the large native build, and open correctness reports require an engineering evaluation before production.

We ran it

Lab card: what happened when we ran FalkorDBScreenshot of FalkorDB (www.falkordb.com)
Install✓ · 16s150 packages
Build✗ · 116s
Tests✗ · 7sran, no count parsed
Repo4612 files~1,114,343 lines of source · 70.6 MB · 20 CI workflows · tests dir

Answers from our run

Does FalkorDB build from source?

Dependencies installed in 16 seconds (150 packages), and the build failed. We cloned commit dd99ae9 into a clean Debian container with 3 CPUs and no project-specific setup.

Do FalkorDB's tests pass?

The test command failed in our container, and its output did not report a pass or fail count.

Who should not use FalkorDB?

Contributors expecting cargo build to be the whole setup: the README requires GraphBLAS, LAGraph, RediSearch, recursive submodules, clang-22, and libomp-22-dev on Linux.

What are the alternatives to FalkorDB?

Neo4j, Memgraph, JanusGraph. Our FalkorDB build failed after 116 seconds and its tests failed after 7 seconds because deps/RediSearch/bin was missing.

Setup2/5Docker trial is short; source build needs several native layers
Docs4/5Good client and contributor paths with explicit native requirements
Community5/55,700 stars, an August push, and active issue triage
Maturity4/5Frequent v4 releases, with migration and correctness work ongoing

Discussed on

  1. hnShow HN: FalkorDB fork from RedisGraph bringing it back to life11 points
  2. hnFalkorDB is rewriting its graph database engine in Rust8 points
  3. hnShow HN: FalkorDB – Open-source graph database major update (C/Rust)7 points
  4. hnShow HN: Text2SQL with a Graph Semantic Layer7 points
  5. hnShow HN: We built a type-safe Python ORM for RedisGraph/FalkorDB5 points

Who it’s for

Teams that want OpenCypher queries through Redis commands and can operate a stateful database service.
Knowledge-graph builders who need vector search beside graph traversal.
Rust and database engineers prepared to use the supplied development container for source work.
Organizations that have reviewed SSPL-1.0 and accept its terms for their deployment.

Who it’s NOT for

Contributors expecting cargo build to be the whole setup: the README requires GraphBLAS, LAGraph, RediSearch, recursive submodules, clang-22, and libomp-22-dev on Linux.
Teams that require a permissive Apache-2.0 or MIT database license: FalkorDB uses SSPL-1.0.
Workloads that depend on HNSW results during ongoing ingestion without their own regression test: open issue 1710 reports empty or wrong results as its graph grows and is labeled needs-info.
Operators seeking a small dependency surface: our checkout had about 1.11 million source lines, and the build stopped when a generated RediSearch directory was absent.

Setup reality

Our sandbox installed 150 packages in 16 seconds. The build failed with exit 101 after 116 seconds, and tests failed with exit 101 after 7 seconds. Both logs ended at graph/build.rs:114: deps/RediSearch/bin missing - run ./redisearch.sh first.

The one-command user route is the published Docker image on ports 6379 and 3000. Source contributors need recursive submodules plus GraphBLAS, LAGraph, RediSearch, clang-22, and libomp-22-dev on Linux; the README recommends a VS Code development container.

This is a stateful Redis module, so production also needs persistent storage, backup and restore procedures, client authentication, resource sizing, and license review. The repository had 20 CI workflow files and a tests directory, but our stated fresh container did not reach compilation or test execution.

OpenCypher rides on Redis commands and sparse matrices

FalkorDB models data as property-bearing nodes and relationships, then queries that graph with OpenCypher plus project-specific extensions. Underneath, GraphBLAS sparse matrices represent adjacency and linear-algebra operations execute graph work. The server is a Redis module, so clients can send GRAPH.QUERY through redis-cli or raw Redis commands. Official client libraries cover Python, Java, Node.js, Rust, Go, and C#.

The quick start exposes Redis on port 6379 and a browser interface on port 3000 from one published container. That is an approachable way to test a graph, and the README's motorcycle example creates riders, teams, and relationships with a few Cypher statements. Production evaluation still needs your own queries and data shape. The repository links separate performance results, but our lab did not run a database benchmark, so we make no latency or throughput claim here.

The 116-second source build stopped before Rust compilation finished

FalkorDB's manual guide puts cargo build first, then explains the required native layers below it. GraphBLAS and LAGraph come from one project script. RediSearch lives in a recursive Git submodule and has its own build script. Linux contributors need clang-22 and libomp-22-dev, while macOS uses Homebrew LLVM and libomp. The recommended route is a VS Code development container whose first build is described as taking about 10 to 15 minutes.

That dependency chain is part of the product, not optional contributor polish. The Rust graph build links static LAGraph, GraphBLAS, and RediSearch archives, plus OpenMP, C++, OpenSSL, and crypto libraries. graph/build.rs explicitly searches for a generated RediSearch binary directory and aborts when it is missing. A source checkout therefore needs recursive submodules and the documented preprocessing scripts before Cargo can link the module.

What happened when we ran it

Our sandbox installed 150 packages in 16 seconds. The build ran for 116 seconds, then exited with code 101. Its final error came from graph/build.rs:114 and said deps/RediSearch/bin missing - run ./redisearch.sh first. The log also listed the native libraries Cargo intended to link, including LAGraph, GraphBLAS, OpenMP, C++, SSL, and crypto. We do not infer any other cause from that output.

The test command failed with the same exit code 101 after 7 seconds and the same missing RediSearch directory. No test cases ran to a reported summary. The checkout contained 4,612 files, about 1,114,343 source lines, and 70.6 MB before installed dependencies. It had 20 CI workflow files and a tests directory. Those are healthy engineering signals, but they do not turn our failed build and test attempts into passes.

Six official clients reduce application-side friction

FalkorDB's Redis command surface is practical for existing service stacks. A team can start with raw commands, then use one of the 6 official language clients when it wants typed graph results and connection management. The property-graph model and OpenCypher also make query intent readable to people who already know graph databases. Graph storage, traversal, and vector indexing live in the same database instead of being joined across unrelated services.

The combined design also widens the failure surface. Open issue 1710 reports HNSW vector queries returning empty or incorrect results during sequential ingestion and includes a reproduction based on a growing graph. Maintainers marked it needs-info, so it is an unresolved report rather than a confirmed universal defect. If retrieval depends on vectors while writes continue, replay that workload against the exact FalkorDB image you intend to ship and compare results with a known set.

SSPL-1.0 rules out some otherwise good fits

The repository license is Server Side Public License 1.0. Teams that standardize on permissive licenses cannot treat it as an Apache-2.0 substitute, and anyone providing the database as part of a network service should have counsel review section 13. Running the unmodified program for an internal application is a different decision from offering database functionality to third parties. The license deserves review early, before migration code and operational tooling make a change expensive.

FalkorDB also carries the duties of a stateful Redis-compatible service. Persistent data belongs under /var/lib/falkordb/data in the quick-start example. A real deployment needs controlled network exposure, authentication, backups, restore drills, upgrade tests, and resource limits. The browser interface on port 3000 is useful during a trial, but it should not be treated as evidence that those operational decisions have been made.

Version 4.20.4 shipped while 642 issues and PRs remained open

GitHub recorded 5,700 stars and a last push on August 30, 2026. Release v4.20.4 arrived on August 20 with a replica crash fix, a persisted-graph loading change, and serializer APIs. GitHub's combined open count was 642; a separate search split that into 489 issues and 153 pull requests. The dates and active pull-request queue show current maintenance, while the volume also signals a large migration and compatibility workload.

The Rust repository has several distinct test routes: graph unit tests, end-to-end and function tests, concurrency and MVCC suites, flow tests, and a Technology Compatibility Kit selection. That breadth is reassuring once the native prerequisites are in place. Our 7-second test failure means we did not independently confirm any of them at commit dd99ae9. Trial the Docker image, then reproduce the source build in the project's container before committing staff time to a fork.

Alternatives

ProjectWhat it isPick it when
Neo4jA widely used property-graph database with Cypher and a large surrounding product ecosystem.pick this instead when broad tooling, training material, and established Cypher operations matter more than Redis protocol compatibility.
MemgraphAn in-memory Cypher-compatible graph database written in C++.pick this instead when streaming graph analysis and Memgraph's operational model fit your workload better.
JanusGraphA distributed graph database designed to sit over external storage and indexing systems.pick this instead when horizontal scale over a chosen storage backend matters more than a Redis module and OpenCypher.

What people are saying

  1. [github-trending] FalkorDB/FalkorDB

Sources

  1. FalkorDB repository
  2. FalkorDB README at commit dd99ae9
  3. FalkorDB v4.20.4 release
  4. FalkorDB graph build script at commit dd99ae9
  5. FalkorDB RediSearch build script at commit dd99ae9
  6. FalkorDB HNSW ingestion issue 1710
  7. FalkorDB SSPL-1.0 license

More data reviews

turso · TrackersListCollection · dash · getcontact-cli · awesome-zhuiju-free · iggy · the whole board →