Pioneer CLI

Command-line interface for the Pioneer AI training platform.

██████╗ ██╗ ██████╗ ███╗   ██╗███████╗███████╗██████╗ 
██╔══██╗██║██╔═══██╗████╗  ██║██╔════╝██╔════╝██╔══██╗
██████╔╝██║██║   ██║██╔██╗ ██║█████╗  █████╗  ██████╔╝
██╔═══╝ ██║██║   ██║██║╚██╗██║██╔══╝  ██╔══╝  ██╔══██╗
██║     ██║╚██████╔╝██║ ╚████║███████╗███████╗██║  ██║
╚═╝     ╚═╝ ╚═════╝ ╚═╝  ╚═══╝╚══════╝╚══════╝╚═╝  ╚═╝

Installation

npm

npm install -g @fastino-ai/pioneer-cli

Manual Install (requires Bun)

# Install Bun if needed
curl -fsSL https://bun.sh/install | bash

# Clone and install
git clone https://github.com/fastino-ai/pioneer-cli.git
cd pioneer-cli/
bun install

# Run directly
bun run src/index.tsx --help

Quick Install (coming soon)

curl -fsSL https://pioneer.ai/install.sh | sh

Usage

# Show help
pioneer --help

# Authentication
pioneer auth login       # Enter API key interactively
pioneer auth logout      # Clear stored API key
pioneer auth status      # Check if logged in

# Agent
pioneer agent                                      # Interactive chat (default)
pioneer agent --mode research                       # Research mode (Pro subscription required)

# Training Jobs
pioneer job list
pioneer job get <id>
pioneer job logs <id>
pioneer job delete <id>
# To create a training job, use the agent — it will help you pick a base model
# and datasets conversationally:
pioneer agent

# Models
pioneer model base-models                             # List available base models (tabular)
pioneer model endpoints list                          # List model catalog entries
pioneer model endpoints create                        # Create a model entry (interactive by default)
pioneer model endpoints create --model "qwen/Qwen3-8B-Instruct" --repo https://github.com/fastino-ai/Pioneer
pioneer model endpoints get <model-id>                # Endpoint details (incl. attached dataset count)
pioneer model endpoints update <model-id> --description "Updated model metadata"
pioneer model endpoints delete <model-id>
pioneer model endpoints quality-metrics <model-id>
pioneer model endpoints deploy <model-id> [--job <training-job-id>] [--reason "..." ] [--all]
                                                       # If --job is omitted, pick from a list of deployable jobs interactively.
                                                       # The list is filtered to jobs whose base model matches the endpoint; pass --all to bypass that filter.
pioneer model endpoints rollback <model-id> <deployment-id>

pioneer model artifacts list                          # List trained and deployed artifacts
pioneer model artifacts trained                       # List trained artifacts
pioneer model artifacts deployed                      # List deployed artifacts
pioneer model artifacts download <job-id>             # Get a signed download URL for the artifact
pioneer model artifacts upload <job-id> --to hf --repo username/model  # Push trained artifact to HF

# Datasets
pioneer dataset list                                  # List remote + local datasets (tabular)
pioneer dataset get <name[:version]>                  # Dataset details (key/value layout)

# Most read commands also accept --json for the raw JSON payload.

# Aliases
pioneer model_endpoints list
pioneer model_artifacts list

Agent command

agent starts in interactive mode by default (standard workflow). For the only explicit alternate mode:

• --mode research: Pro mode with deeper research response style.
• agent resume: list recent conversations and resume a selected session.
• agent sessions: explicit alias for listing sessions and resuming one.

# Interactive (default)
pioneer agent

# Standard mode is the default:
pioneer agent

# Research mode (Pro subscription required)
pioneer agent --mode research

# Open a session list and pick one to continue
pioneer agent resume

# Equivalent: explicit sessions command
pioneer agent sessions

# Resume a specific conversation id
pioneer agent resume 4f2a...

When you start any of these commands, the CLI opens a conversational prompt and keeps accepting follow-up messages in the same session.

Interactive standard mode uses the same WebSocket agent as the Pioneer app’s sidebar: Felix tools plus local CLI tools (bash, read_file, write_file, edit_file, list_directory, search_files, and browse_url — HTTP GET for public pages with HTML stripped; no JavaScript). Full Browser Use (driving your open Pioneer UI with click/type/navigate against the live DOM) remains web-only; the CLI surfaces a clear hint if those tools appear.

End-to-End Example

This example shows the complete workflow from running jobs to using agent support.

Step 1: Start a planning session

Use agent for quick iteration while preparing jobs and reviewing results.

Step 2: Open the interactive helper

# Open a short interactive helper session
pioneer agent

Step 3: Create a Training Job

Use pioneer agent to create a training job. The agent will guide you through picking a base model, selecting datasets, and configuring hyperparameters without needing to remember flags or dataset IDs:

pioneer agent

Step 4: Monitor Training Progress

# List all training jobs
pioneer job list

# Check job status
pioneer job get <job-id>

# View training logs
pioneer job logs <job-id>

Step 5: Use Your Trained Model

Once training completes:

# List trained models
pioneer model artifacts trained

# Download the trained model
pioneer model artifacts download <job-id>

Configuration

The CLI stores configuration in ~/.pioneer/config.json.

Configuration File

The CLI stores your API key in ~/.pioneer/config.json after running pioneer auth login.

{
  "apiKey": "your-api-key",
  "hfToken": "your-hf-token"
}

Authentication

Pioneer API

To use the CLI, you'll need a Pioneer API key. Get one at https://app.pioneer.ai/api-keys.

By default, the CLI points at production (https://api.pioneer.ai). For dev/staging and local debugging, set PIONEER_API_URL explicitly (for example: https://api-dev.pioneer.ai or http://localhost:5001).

# Login with your API key
pioneer auth login

# Check authentication status
pioneer auth status

# Logout (clears saved key)
pioneer auth logout

Hugging Face (Optional)

For uploading model artifacts to Hugging Face, you'll need an HF token. Get one at https://huggingface.co/settings/tokens .

# Set your Hugging Face token
pioneer auth hf

# Check token status
pioneer auth hf status

# Clear token
pioneer auth hf logout

Note: Hugging Face-backed dataset import/export is currently behind the hidden dataset workflow, so the main recommended HF flow today is model artifact upload via model artifacts upload --to hf.

Development & Testing

Building & Running

cd pioneer-cli
bun install
bun run dev          # Hot reload
bun run typecheck    # Type checking

Release

Before publishing, bump package.json to a version that is not already on npm:

npm version patch --no-git-tag-version

Then verify the npm package tarball locally:

bun run release:check-version
bun run release:pack

release:pack writes the verified package to release-artifacts/*.tgz and checks that the tarball contains only:

README.md
dist/index.js
dist/yoga.wasm
package.json

To publish, run the Publish Pioneer CLI npm Package GitHub Actions workflow from main. Leave dry_run=true to build and upload the tarball as a GitHub Actions artifact without publishing. Set dry_run=false to publish the verified tarball to npm and attach the same .tgz to a vX.Y.Z GitHub Release.

Tech Stack

• Runtime: Bun
• UI: Ink (React for CLI)
• Language: TypeScript

License

MIT