Desktop walkthrough¶
The current Desktop keeps data management, trading controls and Agent editing inside the main window. The video predates this navigation change and shows some of them as separate windows; use the illustrated guide for current screen layouts and return actions.
The short tutorial shows the actual Market Lab Desktop with synthetic market data and temporary accounts. Its captions are visible in the video; a separate English caption track is also available. No model/provider is called and no API key is used.
Watch or download the tutorial
1. Prepare an experiment¶
Start the demo after installing the source edition:
python -m sim_platform desktop --demo
Open New experiment. Choose the market recording, simulated account, decision maker and start point. Preview market data is optional. Start experiment prepares the data automatically and opens the experiment paused. Press Play or Step to advance it; Controlled mode lets you advance one market event at a time.
2. Submit and inspect a simulated order¶
With a Human experiment open, use Controls & records to enter a limit order. The tutorial submits a buy order against the synthetic AAPL book. Stepping forward shows the accepted order fill through the real execution engine. Inspect the order, position and cash together, then stop the experiment before starting another one.
The synthetic dataset is deliberately small. This walkthrough explains behavior, not strategy quality or investment performance.
3. Find the Agent editor¶
Open Agents to create a built-in template or import a Python strategy/carrier. The tutorial opens the editor to show this entry point; it does not run a model-backed Agent. Saved changes create an immutable revision for future experiments.
For a real Agent experiment, choose its model and granted tools, set the matching credential reference, and prepare an imported project's dependencies. Follow the Agent project guide and SDK guide.
4. Run a Live baseline¶
The tutorial next creates a Live experiment using the built-in Buy → Hold → Sell rule. Market time advances at 1× while decisions run. This rule is an understandable baseline, not an AI model. Agent experiments also use background Live execution.
Compare Controlled and Live modes before selecting your own run.
5. Verify and review a recorded run¶
After the Live rule run ends, open its operation log in Replay and select Verify & open. Verification checks the original financial events using a separate account. The recording checks that the replay matched, events were verified and model calls stayed at zero. Play, Next action and seek then change the viewing position without executing the decision maker again.
See first use for data, account and credential setup, and support and limitations for the V0.1 scope.
User guide · All documentation