CameoDB v0.3.10 stable
Why Quickstart Build Query Operate Downloads Blog
01 / 01
Home / Engineering notes / Goran Cvijanovic
Author

Goran Cvijanovic

Goran Cvijanovic writes the CameoDB engineering notes: release write-ups, architecture and benchmarks from the engine.

  • Release notes 8 write-ups, from v0.2.2 to today.
  • Architecture The storage, the search and the mesh, from the inside.
  • Put to work Agents, datasets, benchmarks and CameoDB.ai.
LinkedIn →
Release v0.3.10
15 November 20265 min read

CameoDB v0.3.10: Compressed Documents, CJK Search and Space Back After Deletes

CameoDB v0.3.10 stores documents compressed, searches Chinese, Japanese and Korean by word, gives deleted space back on its own and reports min and max over every match.

ReleaseStorageSearch
CameoDB v0.3.9: OAuth for MCP, Ghost Fields and Delete by QueryCameoDB v0.3.9 signs MCP clients in with OAuth 2.1, mints scoped API keys over HTTP, searches one text in several analyzers with ghost fields and deletes by query.5 Nov4 min read CameoDB v0.3.8: Zero-Downtime Certificates, Signed Releases and CIDR SearchCameoDB v0.3.8 rotates certificates on SIGHUP with no restart, ships cosign-signed artifacts with SPDX and CycloneDX SBOMs, and adds CIDR queries on IP fields.25 Oct3 min read
Release notes

Every release,
written up

8 releases since v0.2.2, each with what changed, why it matters and how to upgrade.

Release notes and version history →
v0.3.10Compressed Documents, CJK Search and Space Back After Deletes15 Nov v0.3.9OAuth for MCP, Ghost Fields and Delete by Query5 Nov v0.3.8Zero-Downtime Certificates, Signed Releases and CIDR Search25 Oct v0.3.7Two-Phase Paging, Millisecond Recovery and a Windows Service19 Oct v0.3.5Commits on a Clock and 370,000 Docs a Second28 Sep v0.3.1Secure by Default, With Paging and Exact Sorting16 Aug v0.2.3Battle-Tested Stability for Agentic AI Workflows24 Jul v0.2.2A Native MCP Server, Sorting and Compressed Sources5 Mar
Architecture

How the engine
is built

One write path through Redb and Tantivy, ACID without consensus on the data path, and a mesh that needs no leader.

Storage, search and mesh →
ACID Guarantees in a Leaderless World: CameoDB's Transaction ModelHow CameoDB keeps every write ACID across a leaderless mesh, with no consensus on the data path.26 Feb8 min read Redb + Tantivy: The Atomic Dual-Engine at CameoDB's CoreACID key-value storage and full-text search behind one write path, with Redb as the source of truth.12 Mar5 min read Search Everything: Full-Text Indexing with Tantivy Under the HoodInverted indexes, BM25 scoring, phrase queries and field boosting: how Tantivy powers search in CameoDB.20 Mar6 min read Leaderless Mesh: How CameoDB Scales Without ConsensusZero master nodes, no Raft, no Paxos. Consistent hashing places the data and Kademlia finds the peers.28 Mar7 min read
Put to work

From a dataset
to an agent’s answer

Load a million rows in one command, hand them to an agent over MCP, measure what the engine does with them, or have the whole layer installed for you.

MCP guides and posts →
Native MCP Server: Give AI Agents Direct Database AccessHow CameoDB's built-in MCP server lets Claude, Cursor and Windsurf query your data directly.10 Apr5 min read From Zero to a Million Jokes: Loading Hugging Face Datasets into CameoDBOne CLI command, one million Reddit jokes from Hugging Face, fully indexed and searchable.15 Apr4 min read Fast by Design: Benchmarking CameoDBWhere CameoDB's speed comes from, and the numbers cameodb-bench measures on a single node.3 Apr6 min read Introducing CameoDB.ai: Your Intelligence Layer, InstalledMeet CameoDB.ai: we install your intelligence layer, so your AI answers from your own data.21 Oct6 min read
© 2026 CameoDB. Licensed under Apache-2.0, MIT and FSL-1.1-Apache-2.0 Built with Rust 🦀