Skip to content

Starting Pioneer

This is the verified Windows development workflow for the current repository build. It uses the browser/Vite surface; the Tauri shell is optional.

Before you start

You need:

  • Python 3.11 or later;
  • Node.js 20 or later with npm; and
  • Ollama if you want to use a local model provider.

The examples assume the repository is at C:\Users\Owner\repos\pioneer.etl. If your checkout is elsewhere, replace that path. The default durable Pioneer home is %USERPROFILE%\PioneerAI.

1. Check Ollama

In PowerShell:

Invoke-RestMethod -Uri "http://127.0.0.1:11434/api/version"

Expected: an object with a version value. If that succeeds, Ollama is already serving. Do not start a second ollama serve process.

If you use an OpenAI-compatible or other cloud provider, you can continue without Ollama and configure that provider during setup or in Settings.

2. Start the backend

Open a second PowerShell terminal:

Set-Location "C:\Users\Owner\repos\pioneer.etl"
$env:PIONEER_HOME = Join-Path $env:USERPROFILE "PioneerAI"
& ".\.venv\Scripts\python.exe" -m pioneer.cli serve --host 127.0.0.1 --port 8420

Expected: Starting Pioneer.AI on http://127.0.0.1:8420, followed by the server startup messages.

PIONEER_HOME selects the durable environment for configuration, sessions, Memory, saved Flows, run history, logs, attachments, and tool state. The backend reads it when the process starts; changing it in PowerShell does not retarget an already-running backend.

3. Start the frontend

Open a third PowerShell terminal:

Set-Location "C:\Users\Owner\repos\pioneer.etl\frontend"
npm.cmd run dev -- --host 127.0.0.1 --port 1420 --strictPort

Expected: Vite reports a local URL on port 1420. --strictPort prevents an unnoticed fallback to another port.

4. Open and verify Pioneer

Open http://127.0.0.1:1420/#/chat.

The top bar should show Backend Online. LLM Online separately means the configured provider answered its availability check.

Open Machine and confirm:

  • the active profile ID and Pioneer home are the ones you intended;
  • the profile source is PIONEER_HOME environment for the command above;
  • backend evidence says the API is ready; and
  • provider reachability and local model inventory are plausible.

Configured is not observed

A saved provider endpoint or model name is configuration intent. It does not prove that the provider is reachable or the model is installed. Use Models and Machine for current observations.

First-run setup

A fresh ordinary profile opens the setup wizard after the frontend can read the backend health and stable profile identity. The wizard:

  1. introduces Chat, Canvas, Machine, and the command palette;
  2. saves one provider, endpoint, optional credential, and default model;
  3. performs a non-generative provider check and optionally sends one real First Chat;
  4. inspects the durable Memory inventory without changing it; and
  5. records success, failure, and skipped checks separately.

For Ollama, the setup model selector includes only currently installed models whose provider evidence says they support completion. Setup never installs a model automatically.

Stop cleanly

Press Ctrl+C once in the frontend terminal and once in the backend terminal. Normally leave the independently running Windows Ollama application running.

Durable configuration and databases remain. Live streams, pending approval waiters, runtime-only mounts, and in-memory agents do not survive the stop.

Next steps