Arcade DB

07/08/2026

ArcadeDB 26.8.1 is now available.

This release closed 381 issues across 669 commits, and it is the largest single step we have taken on concurrency and storage efficiency.

What changed, measured:

• Concurrent writes to unrelated records on the same page no longer conflict. In our insert benchmark, conflicts dropped from roughly 1750 out of 2000 to zero.
• Deleting a super-node vertex with 100k edges went from 2374ms to 446ms.
• Geospatial indexes now store one entry per point instead of eleven: 5.3x cheaper to index, 11x smaller on disk.
• Time-series TAG columns are dictionary-encoded, cutting page write amplification from 23x to 0.7x.
• Bloom filters on compacted LSM indexes make absent-key lookups 1.9x faster.

We also closed 13 security advisories, including authentication and per-database authorization on the MongoDB, Redis, and Gremlin wire protocols.

The upgrade requires no schema migration, no database rewrite, and no export/import. Take a backup, swap the jars, restart.

Full release notes: https://arcadedb.com/blog/arcadedb-26-8-1/?utm_source=fb&utm_campaign=26-8-1

14/07/2026

ArcadeDB has spoken Bolt for a while, and the honest version of our claim was narrow: the official Neo4j drivers connect, and simple queries work. That is not the same thing as compatibility. The gap is where users get hurt: the driver connects on Monday, and on Thursday someone reads a datetime back out of a query and gets a string.

ArcadeDB 26.7.2 closes that gap with real certification:

🔹 39 conformance scenarios across 9 areas (connection, auth, transactions, bookmarks, multi-database, result handling, type round-trip, errors, protocol negotiation)
🔹 All 5 official Neo4j drivers under test: Java, JavaScript, Python, .NET, Go. No mocks.
🔹 14 pinned driver versions, so a driver-side release that breaks compatibility trips our nightly run within a day
🔹 A live compatibility matrix that regenerates itself every night and is published where anyone can read it, including the two cells that are not green

The testing exposed real protocol bugs, all now fixed: native temporal, Duration and Point structures over Bolt; Neo.TransientError codes so driver retry logic actually fires; HA-aware neo4j:// routing; and proper Bolt 5.x negotiation instead of a silent downgrade.

A compatibility claim you cannot fail is not a compatibility claim. So we built one that can, runs nightly, and publishes the result either way.

If you evaluate a graph database by pointing an existing driver at it and running existing Cypher, that deserves a real answer instead of a compatibility adjective.

Read the full engineering write-up: https://arcadedb.com/blog/bolt-driver-compatibility-certification/

07/07/2026

How do you debug a distributed database running across a Kubernetes cluster? You make it observable from day one.

Our latest engineering post covers ArcadeDB's new cloud observability stack, built on four independently deployable pillars:

📊 Metrics depth: RED timers (Rate, Errors, Duration) for HTTP requests, queries, and transactions, plus percentile histograms and SLO buckets.

🔭 Distributed tracing: OpenTelemetry with W3C traceparent support. Context propagates through Raft replication, so you can follow a single request from the HTTP layer through query ex*****on and transaction commit, across nodes.

🧾 Structured logging: JSON logs with trace, span, and request IDs for painless correlation, built with zero new dependencies.

❤️ Kubernetes health probes: a new process-level liveness endpoint and an HA-aware readiness endpoint.

The design principle we're most proud of: everything is opt-in and defaults to off. A single instrumentation point emits both metrics and spans, so with no tracer registered you pay zero tracing overhead. Upgrades stay byte-for-byte compatible with existing deployments, which means teams can adopt observability with no downtime.

It integrates cleanly with the stacks you already run: Grafana, Prometheus, and Tempo.

https://arcadedb.com/blog/arcadedb-cloud-observability-opentelemetry-kubernetes/

ArcadeDB adds vendor-neutral OpenTelemetry tracing, RED latency metrics with OTLP export, structured JSON logging with correlation IDs, and Kubernetes-standard health probes, all opt-in and fully backward compatible.

03/07/2026

ArcadeDB 26.7.1 is here 🚀

This release represents 420+ commits and 238 resolved issues, and thanks to 100% database format compatibility, upgrading requires no migration.

What's new:

🛡️ High Availability hardening. Failed database operations now trigger per-database quarantine instead of halting entire nodes. Diverged followers can self-recover to leader state, snapshot integrity is strengthened with fsync and manifest verification, and TRUNCATE is now HA-safe.

🔍 Search & Query. Native BM25 full-text scoring with field boosts, alignment with ISO GQL standards, and EXPLAIN/PROFILE coverage for search operations.

📊 Observability. OpenTelemetry distributed tracing with optional OTLP export, structured JSON logging with correlation IDs, and Kubernetes health probes.

🔒 Security. Polyglot scripting restricted to admin privileges, centralized gRPC authorization, and enhanced path-traversal protection.

Also under the hood: dependency upgrades to Netty 4.2.15, Lucene 10.5.0, and JVector 4.0.0-rc.8, plus MongoDB protocol enhancements.

Available now on Docker Hub, Maven Central, and GitHub Releases.

Read the full release notes: https://arcadedb.com/blog/arcadedb-26-7-1/

ArcadeDB 26.7.1 is a large stability, resilience and security release with 420+ commits and 238 resolved issues: Raft/HA hardening with per-database quarantine and divergence self-recovery, OpenTelemetry distributed tracing, native BM25 full-text scoring, ISO GQL alignment, and a broad security swee...

16/06/2026

Deploy a highly available ArcadeDB cluster on Kubernetes with a single command.

Our new guide shows how the official Helm chart handles the hard parts for you: a 3-node cluster with Raft consensus, automatic failover, persistent storage, rolling upgrades, and secure production defaults. Go from a single pod in 30 seconds to a full production deployment, with day-two operations covered too.

Read it here: https://arcadedb.com/blog/deploy-arcadedb-cluster-kubernetes-helm/

04/06/2026

ArcadeDB 26.6.1 is out!

🔒 End-to-end TLS/SSL for HA clusters
💾 Durability + crash-recovery hardening (WAL fsync on commit, gap-safe recovery)
🛡️ Security hardening + full CodeQL cleanup 280+ commits, 66 issues fixed.

https://arcadedb.com/blog/arcadedb-26-6-1/

12/05/2026

🚀 Announcing ArcadeDB 26.5.1

We've just shipped a major release: 270+ commits and 128 issues resolved. The headline is hybrid retrieval, but there's a lot more inside.

What's new:

• Sparse Vector Index (LSM_SPARSE_VECTOR) for BM25 / SPLADE-style retrieval, with WAND and BlockMax-WAND pruning that scales to 100M+ documents
• Server-side Hybrid Retrieval via vector.fuse(...) (RRF, DBSF, LINEAR strategies) — combine dense, sparse, and lexical signals without client round-trips
• INT8 quantization end-to-end across ingest, storage, and query (dramatically lower disk and RSS)
• EXTERNAL property storage: a paired-bucket layout that isolates heavy values (vectors, large JSON, blobs) so scans stay hot
• Query partitioning with a partition-aware planner that prunes unnecessary partitions
• Offline HA cluster bootstrap via snapshot-and-restore (no full re-replication)
• Production-ready Helm chart, aligned with the Raft-based HA subsystem
• Extensive OpenCypher correctness improvements
• FIND REFERENCES is back, easing OrientDB migrations

100% backward compatible. No export/import needed when upgrading.

Read the full release notes 👉 https://arcadedb.com/blog/arcadedb-26-5-1/

Huge thanks to the community for reporting issues, sending PRs, and shaping this release. 🙌

ArcadeDB 26.5.1 ships a brand-new sparse vector index with server-side hybrid retrieval, INT8 quantization end-to-end, EXTERNAL property storage, query partitioning, offline HA cluster bootstrap, and an extensive batch of OpenCypher correctness fixes.

24/04/2026

ArcadeDB 26.4.2 is here.

This release packs 340+ commits and 100+ resolved issues into one of our biggest updates yet.

What's new:

A brand-new Raft-based High Availability stack built on Apache Ratis. Leader election and log replication are now handled by a battle-tested consensus implementation, giving you stronger guarantees for replicated writes and cleaner operations on Kubernetes and bare metal.

The Query Profiler is now available across every entrypoint (HTTP, Studio, embedded, drivers) and shows the full time breakdown, including record deserialization and result serialization.

A massive OpenCypher correctness round with dozens of fixes: filter pushdown for list predicates, dynamic index usage for MATCH WHERE ID(n), CREATE INDEX and CREATE CONSTRAINT IS TYPED support, WALK / TRAIL / ACYCLIC path modes, and many subquery fixes.

Plus: Bolt wire protocol with TLS, batch SQL UPDATE / DELETE / TRUNCATE, case-insensitive COLLATE indexes, production-mode auto-fsync with a startup security checklist, Studio polish (horizontal scroll, progress bars, clickable datasets), and major performance gains including an 80%+ RAM drop on the Neo4j importer.

Download and full release notes: https://github.com/ArcadeData/arcadedb/releases/tag/26.4.2

Thanks to everyone who contributed issues, pull requests, and feedback.

ArcadeDB 26.4.2 Release Notes Overview ArcadeDB 26.4.2 is a major release with over 340 commits and 100+ resolved issues. The headline is a brand-new Raft-based High Availability stack built on Apa...

01/04/2026

BREAKING NEWS: Anthropic Acquires ArcadeDB to Power "Bigfoot" - On Path to Super Intelligence https://arcadedb.com/blog/anthropic-acquires-arcadedb-bigfoot-superintelligence/

Send a message to learn more

25/03/2026

ArcadeDB v26.3.2 introduces Graph Analytical Views — a CSR-based OLAP engine that coexists with our transactional engine. We run the official LDBC benchmark against 8 GraphDBs and OLAP DBMS (DuckDB)
https://arcadedb.com/blog/graph-olap-engine-the-fastest-graph-analytics-with-zero-compromises/