Jev Trading

A structured decision for one stock: the final action among Buy, Sell, and Hold, the model's probabilities for the three actions, and any data-quality or policy restrictions, plus a record of the evidence. You can review the history, input snapshots, and raw model responses later.

No. The demo uses synthetic data and a fixed response, needs no API key or AIStock, and never automatically switches to paid analysis. Live analysis with cloud inference adds AIStock and a Jev API key; the local-model route needs model weights and a compatible service.

Bun and uv, both verified with a version command after install. The launcher runs uv sync --frozen --no-dev automatically to prepare the Python 3.12 or newer environment, and the first launch needs internet access to download dependencies.

No. The README is explicit: signals only, with no brokerage account integration, automatic stop-loss, target position sizing, or return backtesting. Action probabilities are the model's distribution over the three actions, not the probability of making a profit.

Run the local service and use its HTTP API: the README's starting-points table lists integrating a script or application when a running local service is available, alongside the web workbench and the demo. bun run serve starts the same service without opening a browser.

No. The demo uses synthetic data and a fixed response, calls neither cloud nor local models, and never auto-switches to paid analysis. A real analysis with cloud inference requires AIStock plus a Jev API key, and it records the input snapshot and raw model response for review.

It is the ethanalgox community project. Startup has been validated on macOS only and native Windows startup is unverified, and the current app uses Simplified Chinese labels, with the README walking through the controls in English. No license or benchmark results appear in the metadata.