NVTrader documentation

Everything you need to operate, modify, and interpret NVTrader — the multi-agent autonomous trading platform built on NVIDIA DGX Spark. Twenty-five specialized agents on the A2A event bus today (26 active by default + 4 example agents available via the Agent Builder).

Getting started

Three short articles get you from a fresh DGX Spark to a trading paper portfolio in about 20 minutes.

Platform

Per-page reference for everything in the NVTrader UI. Each page covers what it shows, how to drive it, every setting, and how to read the outputs.

Engines

Reference for the six core engines that make the platform autonomous. Each engine doc covers what it is, how to call it, how to tune it, and how to interpret its outputs — with the exact code paths to look at if you need to modify it.

Reference

Specifications, vocabularies, and the technical brief you'd share with a CEO or partner.

Documentation conventions

• live — feature is shipped, running today on the demo host. Cited code paths exist.
• beta — feature works but is rough or partially scoped; expect breakage if you push it.
• no badge — stable surface, documented for completeness.
• Code paths cite from the repo root, e.g. src/traderspace/bus/bus.py.
• REST examples assume the demo host. Replace http://127.0.0.1:8015 with your deployment URL.
• Each page is structured: Overview → How it works → How to use it → How to interpret it → How to modify it → API surface → Troubleshooting.