dio is an OCaml 5 trading engine built for high-frequency trading. Strategies are authored and extended through a scripting configuration. The engine loads a strategy, validates it, and runs it over a library of registered actions. The bundled strategy is Jacobs Ladder, a grid that buys dips and sells into reversals. It runs against Kraken, Hyperliquid, Lighter, Interactive Brokers, and Alpaca.
Each traded asset runs in its own OCaml domain. Order intents go through a lock-free executor, and market data arrives on lock-free ring buffers. A capital-survival oracle sizes each position from the asset's all-time drawdown history. A terminal dashboard connects to the running engine over a Unix domain socket.
dio trades real money and ships without warranty. Test on a testnet before risking capital.
The deployment guide has the
full walkthrough. Create a working directory, pull the image, extract the
starter files, then fill in config.json and .env.
The image contains the example config, env template, and compose file:
IMAGE=ghcr.io/malciller/dio-domains:latest
docker pull $IMAGE
docker run --rm -v "$PWD:/out" --entrypoint cp $IMAGE /usr/share/doc/dio/config.example.json /out/config.json
docker run --rm -v "$PWD:/out" --entrypoint cp $IMAGE /usr/share/doc/dio/.env.example /out/.env
docker run --rm -v "$PWD:/out" --entrypoint cp $IMAGE /usr/share/doc/dio/compose.yaml /out/compose.yaml
docker compose up -d # start the engine
docker compose run --rm dashboard # attach the dashboard (Ctrl-p Ctrl-q to detach)The image is published to GitHub Container Registry
(ghcr.io/malciller/dio-domains) and Docker Hub (malciller/dio-domains). The
commands use GHCR. Substitute the Docker Hub name to use that registry.
The published image targets linux/amd64 and runs under emulation on Apple
Silicon. The engine reads config.json and .env from the working directory and
writes state to the dio-data volume.
Full reference: configuration and deployment.
| Executable | Function |
|---|---|
dio |
The engine. Trades the instruments in config.json. |
dio-dashboard |
Terminal UI. Connects to a running engine over the Unix domain socket. |
dio-oracle |
Runs the sizing pipeline offline and prints the decision surface. |
dio targets high-frequency trading. Trading domains pin to performance cores,
background work stays on efficiency cores, and real-time SCHED_FIFO priority is
available through DIO_TRADING_RT_PRIO on a dedicated host. The strategy
interpreter resolves action handlers at load, prewarms guards, and hoists
per-cycle closures out of the steady state. The OxCaml flambda2 build uses
local_ allocation mode, and CI fails on any compiler alert.
Latency and allocation are observable through every phase of the cycle:
- A histogram profiler reports p50, p90, p95, p99, and p999, with a nanosecond tier for sub-microsecond samples.
- A busy-spin canary detects process-wide stop-the-world pauses and scheduler descheduling that per-domain GC counters miss.
- Per-cycle GC counters attribute a spike to collector activity or to CPU work.
Each entry in config.json names a strategy, for example
"strategy": "jacobs_ladder". The engine loads strategies/<name>.strategy,
validates it, and compiles it to its internal step/action representation at
startup. The script is a small dotted language. The compiled .json form is
also accepted.
when book.updates:
skip.nan.price:
if $platform.price.nan:
stop
dio strategy validate <file> checks a strategy statically.
dio strategy compile <file.strategy> emits its JSON form. The
strategy script guide documents
the language and the action vocabulary.
All documentation lives at https://diophantsolutions.com/dio/: deployment, configuration, strategy scripts, capital oracle, risk and terms, and the specification.
Bug reports and feature requests go through GitHub issues: https://github.com/malciller/dio-domains/issues. Do not paste API keys or other secrets into an issue.
MIT. See LICENSE.