signalpilot 0.6.1

SignalPilot CLI - Your Trusted CoPilot for Data Analysis

SignalPilot Installer CLI

This installer CLI is a bootstrap installer that sets up the SignalPilot-AI Jupyter extension in one command.

The CLI is NOT the product. It's a convenience installer. The SignalPilot Jupyter extension (agentic harness) is the actual product.

What You're Installing

SignalPilot is a Jupyter-native AI agentic harness that investigates data by connecting to your organizational context:

Four core capabilities:

• 🔌 Multi-Source Context — Auto-connects to db warehouse, dbt lineage, query history, Slack threads, Jira tickets, and past investigations via MCP
• 🔄 Long-Running Agent Loop — Plans, executes, iterates until task complete with analyst-in-the-loop approval (not single-shot completions)
• 🧠 Multi-Session Memory — Remembers past hypotheses, validated assumptions, known data quirks across investigations
• 📚 Skills & Rules — Custom analysis patterns (skills) + team coding standards (rules) + business logic

Security: Zero data retention • Read-only access • Local-first execution • SOC 2 in progress

Quick Install

Prerequisites: macOS, Linux, or Windows (WSL) • Internet connection

Don't have uv? Install it first (takes 10 seconds):

curl -LsSf https://astral.sh/uv/install.sh | sh

Install SignalPilot:

uvx signalpilot@latest

What happens:

• Creates ~/SignalPilotHome workspace with starter notebooks
• Installs isolated Python 3.12 + Jupyter Lab + SignalPilot extension
• Installs data packages (pandas, numpy, matplotlib, seaborn, plotly)
• Optimizes Jupyter cache for fast startup
• Launches Jupyter Lab at http://localhost:8888

Time: ~2 minutes

Why uv?

• 10-100x faster than pip/conda for package installation
• SignalPilot runs on it — native integration with kernel
• Modern Python package management with better dependency resolution

Launch Jupyter Lab Anytime

Once installed, start Jupyter Lab with:

uvx signalpilot@latest lab

What this does:

• Opens Jupyter Lab in your current directory
• Uses home environment from ~/SignalPilotHome/.venv
• SignalPilot extension pre-loaded
• Opens browser at http://localhost:8888

⚠️ Smart Detection: If a local .venv with jupyter is detected in your current directory, you'll see a red warning. Use --project flag to use it instead.

Keeping SignalPilot Updated

SignalPilot automatically checks for updates when you launch Jupyter Lab. When an update is available, you'll see a notification:

For minor updates:

╭─────────────── 📦 SignalPilot Update ───────────────╮
│ Update Available: 0.11.8 (installed: 0.11.7)        │
│ Run 'sp upgrade' to update                          │
╰──────────────────────────────────────────────────────╯

For major updates:

╭─────────────── 📦 SignalPilot Update ───────────────╮
│ Important Update: 0.12.0 (installed: 0.11.7)        │
│ This is a MAJOR update                              │
╰──────────────────────────────────────────────────────╯
Upgrade now? [y/n] (n):

Manual Upgrade

Upgrade both the CLI and library anytime:

uvx signalpilot@latest upgrade

Upgrade your project's local environment:

cd /path/to/project
uvx signalpilot@latest upgrade --project

Note: Update checks happen in the background and never slow down Jupyter startup. You can disable them in ~/SignalPilotHome/.signalpilot/config.toml if desired.

📖 Full upgrade guide: docs/UPGRADE-USER-GUIDE.md

What Gets Installed

Python Packages:

• signalpilot-ai — AI agent integration (the actual product)
• jupyterlab — Modern Jupyter interface
• pandas, numpy — Data manipulation
• matplotlib, seaborn, plotly — Visualization
• python-dotenv, tomli — Configuration utilities

Directory Structure:

~/SignalPilotHome/
├── user-skills/       # Custom analysis patterns
├── user-rules/        # Team coding standards
├── team-workspace/    # Shared notebooks (git-tracked)
├── demo-project/      # Example notebooks
├── pyproject.toml     # Python project config
├── start-here.ipynb   # Quick start guide
└── .venv/             # Python environment

Working in Different Modes

SignalPilot offers three ways to launch Jupyter Lab:

Default Mode (Current Folder + Home Environment)

cd ~/projects/my-analysis
uvx signalpilot@latest lab

What this does:

• Opens Jupyter Lab in your current directory
• Uses home environment from ~/SignalPilotHome/.venv
• Perfect for quick exploration without setting up new environment

⚠️ Warning: If you have a local .venv with jupyter, you'll see a red warning prompting you to use --project flag.

Project Mode (Current Folder + Local Environment)

cd ~/projects/custom-analytics
uvx signalpilot@latest lab --project

What this does:

• Opens Jupyter Lab in your current directory
• Uses local .venv in that directory (fails if missing)
• Great for project-specific work with custom dependencies

Requirements:

• A .venv must exist in current directory
• Must have jupyterlab and signalpilot-ai installed

Create project environment:

mkdir ~/projects/custom-analytics && cd ~/projects/custom-analytics
uv venv --seed --python 3.12
source .venv/bin/activate
uv pip install jupyterlab signalpilot-ai pandas numpy matplotlib plotly
uvx signalpilot@latest lab --project

Home Mode (SignalPilotHome Workspace + Home Environment)

uvx signalpilot@latest lab --home
# Or use the shortcut:
uvx signalpilot@latest home

What this does:

• Opens Jupyter Lab in ~/SignalPilotHome directory
• Uses home environment from ~/SignalPilotHome/.venv
• Default workspace with all your skills, rules, and team notebooks

Pass Jupyter Lab Arguments

You can pass any Jupyter Lab flags after the command:

# Custom port
uvx signalpilot@latest lab --port=8889

# Disable browser auto-open
uvx signalpilot@latest lab --no-browser

# Combine with mode flags
uvx signalpilot@latest lab --project --port=8889
uvx signalpilot@latest home --no-browser

# Bind to all interfaces (remote access)
uvx signalpilot@latest lab --ip=0.0.0.0 --port=9999

All standard jupyter lab arguments work.

Alternative Installation Methods

Option 1: Run with uvx (Recommended)

uvx signalpilot@latest

No permanent installation needed. Perfect for most users. Always gets the latest version.

Option 2: Install with uv tool

uv tool install signalpilot
sp init

Installs sp command globally. Use sp lab, sp home to launch later.

Note: Global installations don't auto-update. Reinstall periodically:

uv tool install --force signalpilot

Option 3: Install with pip

pip install signalpilot
sp init

Works but slower than uv (10-100x). May have dependency conflicts.

Requirements

• Python 3.10 or higher
• uv package manager (recommended)

Links

• Homepage
• Full Documentation
• Installation Guide
• 5-Minute Quickstart
• GitHub

License

MIT License - See LICENSE file for details