EventFlux

Stream processing engine built in Rust. Write SQL, process events, skip the infrastructure headache.

Why EventFlux

You need to detect patterns in event streams, aggregate metrics, or react to conditions in real-time.

Your options today:

• Flink — needs Kubernetes, 4GB+ JVM heap, dedicated ops
• Kafka Streams — needs a Kafka cluster plus Java expertise
• Build it yourself — months of work

For 100k events/sec, that's overkill.

EventFlux runs as a single binary. No cluster. No JVM. No YAML manifests. Just SQL.

Quick Start

# Docker
docker run -v ./app.sql:/app.sql ghcr.io/eventflux-io/eventflux /app.sql

# Or build from source (--recursive pulls in the vendored SQL parser submodule)
git clone --recursive https://github.com/eventflux-io/eventflux.git
cd eventflux
cargo build --release --features connectors-all
./target/release/run_eventflux app.sql

Already cloned without --recursive? Fetch the submodule before building:

git submodule update --init --recursive

Connector feature flags

The default build is fully minimal — core engine plus the built-in timer source and log sink, no external connectors. Each connector is a cargo feature named after its SQL extension name:

Feature	Connector	In default build?
kafka	Kafka source + sink (needs cmake to build)	No
rabbitmq	RabbitMQ source + sink	No
websocket	WebSocket source + sink	No
connectors-all	All of the above	No

Docker images are built with connectors-all, so they include everything. Building from source, pick what you need:

cargo build --release --features rabbitmq          # just RabbitMQ
cargo build --release --features connectors-all    # everything

Prerequisites (for building)

• Rust 1.85+
• Protocol Buffer compiler (for gRPC features)
• Git submodules initialized (see above) — the SQL parser is a vendored submodule

Example

CREATE
STREAM Trades (symbol STRING, price DOUBLE, volume INT);

SELECT symbol, AVG(price), SUM(volume)
FROM Trades WINDOW TUMBLING(1 min)
GROUP BY symbol
INSERT
INTO Summaries;

That's it. No boilerplate. No config files. Just SQL.

How It Compares

	EventFlux	Flink	Kafka Streams
Deployment	Single binary	Kubernetes cluster	Kafka cluster
Memory	50-100MB	4GB+ JVM	1GB+ JVM
Language	SQL	Java/SQL	Java
Setup time	Minutes	Hours/days	Hours
Scale ceiling	~500k eps	Millions+	Millions+

Choose EventFlux when you want simple deployment and SQL-first development.

Choose Flink when you need massive scale or batch+stream processing.

Documentation

Full docs at eventflux.io:

• Getting Started — install and run your first query
• SQL Reference — windows, joins, patterns, aggregations
• Connectors — Kafka, RabbitMQ, WebSocket, HTTP
• Examples — real-world use cases
• Architecture — how it works under the hood

IDE Support

EventFlux Studio — VS Code extension with:

• Syntax highlighting for .eventflux files
• Schema-aware autocomplete
• Real-time error diagnostics
• Query visualization

Install from VS Code marketplace or search "EventFlux Studio" in extensions.

Performance

• 1M+ events/sec on a single node
• Sub-millisecond latency
• Zero GC pauses
• Starts in milliseconds

When to Use

Good fit: IoT backends, e-commerce tracking, analytics pipelines, SaaS telemetry, fraud detection.

Not a fit: You need 100+ connectors, or you're already running Flink at scale.

Status

Active development. Core CEP works. 1,400+ tests passing. See ROADMAP.md for details.

Contributing

cargo nextest run   # run tests
just lint           # clippy
cargo fmt           # format
just check-all      # all quality checks

See DEV_GUIDE.md for setup.

License

Apache-2.0