v0.43.0 · a recommendation engine — “also bought” in real time · full-text search off the heap · non-blocking index builds · source-available, free for production use
OxiDB
A fast, multi-model database.
One database process, several shapes of data: JSON documents with MongoDB-style queries, aggregation pipelines, geospatial and graph stages · a full SQL engine with joins, window functions, stored procedures and an EF Core provider — reachable over the PostgreSQL wire protocol, so psql, JDBC, Npgsql and DBeaver connect unmodified · Gorilla-compressed time series with continuous rollups · a Redis-compatible key-value store with MULTI/EXEC and Lua · S3-compatible object storage that aws-cli and boto3 speak to unmodified · MQTT/AMQP messaging, bridged across protocols · with full-text & vector search over all of it.
ACID transactions, snapshot-consistent reads, Raft replication, point-in-time recovery, SCRAM/TLS/RBAC and AES-GCM encryption at rest are built in — one ~20 MB zero-dependency binary that starts with no configuration, runs embedded in your process, and compiles to WASM for the browser.
one 20 MB binary · zero dependencies · embed or run as a server
One 20 MB binary. Zero dependencies. Everything below: OxiDB — a database engine in Rust: ▪ Document DB (Mongo-style queries & aggregation) ▪ Full SQL engine (joins, DDL, stored procedures) ▪ ACID transactions — OCC + SELECT FOR UPDATE + group commit ▪ Raft clustering & replication ▪ Redis-compatible in-memory store (RESP) with MULTI/EXEC/WATCH + EVAL ▪ MQTT + AMQP (RabbitMQ) brokers, bridged, cross-protocol pub/sub ▪ Full-text search: HTML, PDF, DOCX, XLSX, OCR ▪ S3-compatible blob storage (SigV4, multipart, lifecycle) ▪ Time-series: OHLCV candles, window functions, gap filling ▪ TTL indexes, point-in-time recovery ▪ AES-GCM encryption at rest ▪ REST + WebSocket + JWT auth + security rules ▪ Prometheus metrics, explain & slow-query profiler ▪ Multi-database, RBAC, SCRAM auth, TLS, audit log ▪ Runs embedded too — Python, Go, .NET (EF Core), JS, Julia, Swift ▪ Compiles to WASM Postgres + Mongo + Redis + Elastic + S3. 20 MB.
Architecture
One binary, one server process. Every client speaks its own native wire protocol; the router authenticates and dispatches each request to the right engine; all engines share one durable storage foundation.
Cross-protocol pub/sub: a message published over RESP (OxiMem) reaches MQTT subscribers, and vice-versa — one bus, two protocols. The three query engines are per-database; OxiMem's keyspace and S3 buckets are global. Everything above runs embedded in-process too, with no server or protocol at all.
Showcase — ColdChain
A cold-chain compliance system running on every engine at once, live. Sensors in trucks report temperature over MQTT; two years later an auditor asks you to prove a shipment never left its contracted range.
The point is not that OxiDB can do these things — the list above already says that. It is that ColdChain is an ordinary .NET application. It uses EF Core, MQTTnet, StackExchange.Redis and the AWS SDK: four libraries that have never heard of OxiDB and are not configured to know. They are pointed at localhost ports and behave exactly as they would against the systems they were written for.
So the interesting number is not a benchmark. It is the count of things you did not deploy:
Six systems to install, secure, monitor, back up, upgrade on six schedules and keep consistent with each other — against five ports on one binary, where the backup is one archive taken at one instant with one restore. The dashboard shows what the two processes actually cost while it does it.
Open the live demo runs continuously · six probes · two of them broken on purpose
Install OxiDB in 60 seconds
From zero to a running server. Single static binary, no dependencies, no installer.
1. Download the server
# Linux (x86_64)
curl -LO https://oxidb.baltavista.com/releases/v0.43.0/oxidb-server-v0.43.0-linux-amd64.tar.gz
tar xzf oxidb-server-v0.43.0-linux-amd64.tar.gz
# macOS (Apple Silicon)
curl -LO https://oxidb.baltavista.com/releases/v0.43.0/oxidb-server-v0.43.0-darwin-arm64.tar.gz
tar xzf oxidb-server-v0.43.0-darwin-arm64.tar.gz
# Windows: download oxidb-server-v0.43.0-windows-amd64.zip from /downloads/2. Start the server
# Listens on 127.0.0.1:4444 by default; data goes to ./oxidb_data
./oxidb-server
# Or override with env vars:
OXIDB_ADDR=0.0.0.0:4444 OXIDB_DATA=/var/lib/oxidb ./oxidb-server3. Connect and run your first query
# Download the CLI
curl -LO https://oxidb.baltavista.com/releases/v0.43.0/oxidb-v0.43.0-linux-amd64.tar.gz
tar xzf oxidb-v0.43.0-linux-amd64.tar.gz
# Open the REPL against the running server
./oxidb --host 127.0.0.1 --port 4444
oxidb> insert users {"name": "Alice", "age": 30}
oxidb> find users {"age": {"$gte": 18}}
oxidb> update users {"name": "Alice"} {"$inc": {"age": 1}}Prefer to embed instead of running a server?
# Rust
cargo add oxidb
# Python (TCP client / embedded)
pip install oxidb
pip install oxidb-embedded
# Go
go get oxidb
# .NET
dotnet add package OxiDb.Client.TcpNeed a different platform, source build, or the WebAssembly bundle? See Downloads · Full Quick Start · All clients.
Why OxiDB
Three engines, one binary
A MongoDB-style document store, a full relational SQL engine, and an InfluxDB-style time-series engine — in one process, one file, no glue. Use one or all three; each owns separate storage.
Fast
2-3x faster than MongoDB on indexed queries at 1M docs, 2189x on indexed counts — and beats PostgreSQL across the EF Core benchmark. A bytes-first wire path with a direct JSONB→OxiWire converter.
MongoDB-style documents
JSON queries with $eq, $gt, $in, $regex, $or, $elemMatch, $expr, plus a 16-stage aggregation pipeline (JOINs via $lookup, GROUP BY via $group, window & time-series stages).
Full SQL engine
INNER/LEFT/RIGHT/FULL joins, GROUP BY, window functions, CTEs (incl. WITH RECURSIVE) and LATERAL. An EF Core provider passing all 3832 official spec tests, with ADO.NET, Dapper, and migrations. Instant, online ALTER TABLE.
Time-series engine
InfluxDB-style, with Gorilla-compressed columnar streams (~0.3 bytes/point), line-protocol ingest, continuous-aggregate rollups, and rate() / percentile aggregations.
ACID + crash-safe backups
Multi-collection OCC transactions, 3-phase commit with WAL, CRC32 crash recovery. Crash-atomic MANIFEST checkpoints, low-lock online backup/restore, and point-in-time recovery.
Full-Text & Vector Search
TF-IDF ranked full-text search across 8 document formats (PDF, DOCX, XLSX, HTML, XML, JSON, images/OCR). HNSW vector index with cosine, euclidean, and dot product.
Redis-compatible + MQTT
OxiMem — a Redis-compatible in-memory store (RESP, MULTI/EXEC/WATCH, EVAL/Lua, persistence) — plus a full MQTT 3.1.1 broker (wildcards, retained, QoS, LWT) and an AMQP 0-9-1 (RabbitMQ-protocol) work-queue broker, bridged through amq.topic.
S3 Blob Storage
A full S3-compatible HTTP API (SigV4, multipart, Range, lifecycle) — aws-cli, boto3, and the MinIO SDKs work unmodified. Store files alongside documents.
Sharding + Raft Replication
Horizontal sharding via oxipool (CRC32 routing, scatter-gather) on per-shard Raft groups. Persistent log survives restarts. Verified at 1M records under mid-stream failover.
Flexible Deployment
Embed as a Rust library with zero network overhead, or run as a single static TCP server with SCRAM auth, TLS, and RBAC. One binary, no dependencies, no installer.
Multi-Language
Official clients for Rust, Python, Go, .NET (ADO.NET + EF Core + LINQ + Embedded), Java, Julia, Swift, and JS/TS — plus a WebAssembly build.