Getting Started — Fovux in 5 Minutes

This quickstart walks through the end-to-end local workflow on a YOLO dataset.

1. Install From Source

git clone https://github.com/oaslananka/fovux-kit
cd fovux-kit/fovux-mcp
uv sync --frozen --extra dev

2. Register Fovux In Your MCP Client

{
    "mcpServers": {
        "fovux": {
            "command": "fovux-mcp",
            "env": {
                "FOVUX_HOME": "~/.fovux"
            }
        }
    }
}

3. Inspect The Dataset

Prompt your MCP client with:

Inspect my dataset at ~/data/mini_yolo and summarize the class balance.

Fovux uses dataset_inspect to return class counts, orphan warnings, and sample images.

4. Start Training

Train yolov8n.pt on this dataset for 20 epochs and tag it baseline.

Fovux uses train_start. The call returns immediately while training continues in the background.

5. Watch Progress

Check the current status of my latest run.

Fovux uses train_status. In Fovux Studio, the live dashboard can subscribe to /runs/{run_id}/metrics.

6. Evaluate And Diagnose

Evaluate the best checkpoint and explain the main failure modes.

Use eval_run, eval_per_class, and eval_error_analysis to understand both headline metrics and the likely causes behind misses.

7. Export And Benchmark

Export the best checkpoint to ONNX and benchmark it on CPU.

This combines export_onnx and benchmark_latency, producing a deployable artifact plus p50/p95/p99 latency numbers.

8. Open Fovux Studio

Build the extension from source:

cd ../fovux-studio
pnpm install --frozen-lockfile
pnpm build

Then open the dashboard, dataset inspector, or export wizard inside VS Code.