Questions to answer before you commit.

Bybit spot. The data fetcher, execution layer, and built-in strategies are all built around the Bybit API. dgbit is intentionally exchange-specific so the abstractions can be precise rather than lowest-common-denominator.

Both. The Python package is importable for backtests and strategy authoring. The framework also ships a FastAPI server (dgbit-api), an NNG service bus across data/backtest/strategy workers, and a Vue 3 dashboard. You can run as much or as little of that stack as you need.

Four built-in strategies: Wavelet Reversal (Daubechies wavelet decomposition for mean reversion), MA Crossover (trend following), RSI (momentum), and Bollinger Bands (volatility breakout). Custom strategies subclass BaseStrategy and register with strategy_registry.

Install with pip install dgbit, fetch klines with BybitDataFetcher.get_kline_data, configure with BacktestConfig (initial capital, transaction fee), and call Backtester(config).run(data). The result object exposes total_return, win_rate, and max_drawdown.

Yes. Set BYBIT_API_KEY and BYBIT_API_SECRET in .env. Use BYBIT_TESTNET=true to stage against Bybit testnet before flipping to production. The execution endpoint /api/execution/orders places orders; /api/execution/positions returns open positions.

docker-compose up -d after copying dgbit-api/.env.example to .env. All services come up together. Logs stream from the api container with docker-compose logs -f api.

Because dgbit binds to a single exchange, the backtester's fee model, kline boundaries, and order assumptions map one-to-one to live Bybit behavior. There is no multi-exchange abstraction hiding a mismatch between simulated and real fills.

MIT. The Python package on PyPI and the Docker image at cryptuon/dgbit are both free to use, modify, and self-host.

It can be operated programmatically. Every action in the Vue 3 dashboard is also a REST call plus a WebSocket event, so a scheduler, a CI job, or an autonomous agent can POST a backtest to /api/backtests, poll /api/jobs/{uuid}, request a signal from /api/strategies/{name}/signal, place an order via /api/execution/orders, and subscribe to job.*, trade.*, and signal.generated events to close the loop. dgbit does not ship an agent — it ships the operable substrate one runs on.

A one-off script usually re-implements its own data fetch, its own backtest assumptions, and its own order handling — and the backtest quietly disagrees with live. dgbit gives you a shared strategy interface for backtest and live, a real backtester with fee and drawdown accounting, position tracking with risk controls, and a REST + WebSocket surface for operation. You keep writing Python; you stop rebuilding the plumbing.

Commercial platforms bundle many exchanges, hosted infrastructure, and a subscription. dgbit is the opposite trade: MIT-licensed, self-hosted, and Bybit-only. You own the code and the keys, run it on a single small VPS, and get precision on one venue instead of breadth across many. If you need multi-exchange coverage or a managed SLA, a commercial platform is the honest fit; if you want a transparent, self-hosted framework, dgbit is.

No. dgbit is infrastructure, not a strategy or a signal service. It makes no promise, implied or otherwise, about profit. Backtest metrics describe historical behaviour on the data you supply; past performance does not guarantee future results, and any live trading is at your own risk.

The cheapest path is pip install dgbit (or docker-compose up) on one small VPS — no cluster or managed broker needed. Running live with real capital adds requirements: exchange reconnect/backoff, order and position reconciliation on restart, risk limits enforced before orders leave the process, secrets kept out of the image, and monitoring that alerts on disconnects and rejected orders. Start on BYBIT_TESTNET=true. The repo ROADMAP.md details this path.