User guide¶
Watch the Desktop walkthrough for one preparation, order, fill and verified Replay.
Install and build using the project README. You can explore the demo without market recordings or a model API key.
Testing a beta? Start with the private beta guide.
Cantonese illustrated walkthrough: screenshots, page purposes, manual orders, saved results, Agent settings and verified Replay.
- First experiment: demo, workspace, data, saved Agents and Replay.
- Execution modes: choose Controlled or Live and understand the controls.
- Trade with your own model: a Python file, such as a neural network, that returns a target position or orders.
- Trade several symbols: one experiment and one account across several symbols.
- Read a saved result: every figure on the result page, including Sharpe ratio, win rate, turnover and holding time.
- Evaluate an Agent against baselines: disguise the market, repeat runs, and test the difference with confidence intervals.
- Local models, budgets and lineage: keep data on premises, cap model spending and prove a run is unchanged.
- Explain, sweep and review a study: decision timelines, PnL attribution, sweeps, run diffs and model risk reports.
- Command line: run a rule or model evaluation or compare independent runs.
- Troubleshooting: messages and symptoms, and what to do.
- Support and limitations: available workflows, OS evidence and simulation boundaries.
To write your own Agent, continue with Agent development. For Python application integration, use the Platform API.
Complete Desktop workflow and native menus
Complete Desktop workflow (Traditional Chinese / Cantonese)
Maintained source: docs/user-guide/README.md