Pixell Agent Kit

A lightweight developer kit for packaging AI agents into portable, standardized APKG files.

Installation

Using pipx (Recommended)

pipx install pixell-kit

Using Homebrew

brew install pixell-kit

Using pip

pip install pixell-kit

Quick Start

# Create a new agent project
pixell init my_agent

# Run locally for development
cd my_agent
pixell run-dev

# Build into APKG package
pixell build

# Inspect the package
pixell inspect my_agent-0.1.0.apkg

Configuration

Pixell Kit supports flexible configuration management to avoid entering credentials repeatedly. You can configure API keys and app IDs at multiple levels with the following precedence order:

1. Environment Variables (Highest Priority)

export PIXELL_API_KEY=your-api-key
export PIXELL_APP_ID=your-app-id
export PIXELL_ENVIRONMENT=prod

2. Project-Level Configuration

Create .pixell/config.json in your project directory:

{
  "api_key": "your-api-key",
  "app_id": "your-default-app-id",
  "default_environment": "prod",
  "environments": {
    "prod": {"app_id": "your-production-app-id"},
    "staging": {"app_id": "your-staging-app-id"},
    "local": {"app_id": "your-local-app-id"}
  }
}

3. Global Configuration

Create ~/.pixell/config.json for user-wide settings:

{
  "api_key": "your-api-key",
  "app_id": "your-default-app-id"
}

Configuration Commands

# Interactive setup (recommended for first-time users)
pixell config init

# Set individual values
pixell config set --api-key your-api-key
pixell config set --app-id your-app-id
pixell config set --env-app-id prod:your-prod-app-id
pixell config set --env-app-id staging:your-staging-app-id

# Set global configuration (affects all projects)
pixell config set --global --api-key your-api-key

# View current configuration
pixell config show
pixell config show --global

Simplified Deployment

Once configured, you can deploy without specifying credentials every time:

# Deploy to production (uses stored credentials)
pixell deploy --apkg-file my_agent-0.1.0.apkg

# Deploy to staging (uses environment-specific app ID)
pixell deploy --apkg-file my_agent-0.1.0.apkg --env staging

# Deploy to local development
pixell deploy --apkg-file my_agent-0.1.0.apkg --env local

Environment and Secrets

Phase 1: Required .env in APKG

• Every agent package must include a .env at the project root.
• Builds fail if .env is missing.
• The builder always includes .env in the APKG.
• The validator warns on potential secrets and non-portable absolute paths.

Scaffold:

• pixell init generates a .env.example. Copy to .env and fill values.

Notes:

• Treat .env as sensitive; it is packaged. Use placeholders for shared artifacts.

Phase 2: Runtime Environment Injection (Dev parity)

• The dev server automatically loads .env and applies variables to the process environment.
• Precedence (dev): .env > base environment.
• Logs show variable keys only, never values.

Phase 3: Service-Bound Secrets (Dev parity)

• Optional secrets providers can inject runtime secrets without baking them into .env.
• Provider selection is controlled by environment variables:
  • PIXELL_SECRETS_PROVIDER=static with PIXELL_SECRETS_JSON (JSON object)
  • PIXELL_SECRETS_PROVIDER=env to pass-through current process env
  • PIXELL_SECRETS_PROVIDER=aws to use AWS Secrets Manager with:
    • PIXELL_AWS_SECRETS (comma-separated secret names/ARNs)
    • optional PIXELL_AWS_REGION
• Precedence (dev): provider > .env > base env.

Example (static):

export PIXELL_SECRETS_PROVIDER=static
export PIXELL_SECRETS_JSON='{"OPENAI_API_KEY":"runtime","DB_HOST":"database"}'

Example (AWS):

export PIXELL_SECRETS_PROVIDER=aws
export PIXELL_AWS_SECRETS=my/app/secrets,another/secret
export PIXELL_AWS_REGION=us-east-1

Best Practices

• Use 0.0.0.0 for bind addresses inside containers (not localhost).
• Avoid absolute, machine-specific paths in .env.
• Never log secret values; only keys. The kit adheres to this.

PAR Guidance (separate runtime)

• Apply precedence in the agent subprocess:
  1. Runtime deployment env (highest)
  2. .env from APKG
  3. Base runtime environment (lowest)
• Optionally add service-bound providers per deployment context.

Features

• 📦 Package any AI agent into portable APKG files
• 🚀 Local development server with hot-reload
• ✅ Manifest validation and package integrity
• 🔐 Optional package signing with GPG
• 🐍 Python 3.11+ support (TypeScript coming soon)

SDK Runtime

The Pixell SDK provides runtime infrastructure for agent execution, including task queue processing, user context management, and progress reporting.

Installation

pip install pixell-sdk

Quick Start

import asyncio
from pixell.sdk import UserContext, TaskConsumer

async def handle_task(ctx: UserContext, payload: dict) -> dict:
    # Report progress
    await ctx.report_progress("starting", percent=0)

    # Access user data
    profile = await ctx.get_user_profile()

    # Call OAuth APIs on behalf of the user
    events = await ctx.call_oauth_api(
        provider="google",
        method="GET",
        path="/calendar/v3/calendars/primary/events"
    )

    await ctx.report_progress("completed", percent=100)
    return {"status": "success", "events": len(events.get("items", []))}

async def main():
    consumer = TaskConsumer(
        agent_id="my-agent",
        redis_url="redis://localhost:6379",
        pxui_base_url="https://api.pixell.global",
        handler=handle_task,
    )

    async with consumer:
        await consumer.start()

if __name__ == "__main__":
    asyncio.run(main())

Core Components

UserContext Methods

# OAuth API calls (Google, GitHub, Slack, TikTok, etc.)
result = await ctx.call_oauth_api(provider, method, path, body?, headers?)

# User data access
profile = await ctx.get_user_profile()
files = await ctx.get_files(filter?, limit?)
content = await ctx.get_file_content(file_id)
conversations = await ctx.get_conversations(limit?, since?)
history = await ctx.get_task_history(agent_id?, limit?)

# Progress reporting
await ctx.report_progress(status, percent?, message?)
await ctx.report_error(error_type, message, recoverable?)

Error Handling

from pixell.sdk import (
    AuthenticationError,  # Invalid/expired token
    RateLimitError,       # Rate limit exceeded (check retry_after)
    APIError,             # API error response
    ConnectionError,      # Network failure
    TaskTimeoutError,     # Task exceeded timeout
)

try:
    result = await ctx.call_oauth_api(...)
except RateLimitError as e:
    retry_after = e.details.get("retry_after", 60)
    await asyncio.sleep(retry_after)
except AuthenticationError:
    # Token invalid - cannot retry
    raise

Configuration Options

consumer = TaskConsumer(
    agent_id="my-agent",
    redis_url="redis://localhost:6379",
    pxui_base_url="https://api.pixell.global",
    handler=handle_task,
    concurrency=10,        # Max concurrent tasks (default: 10)
    poll_interval=1.0,     # Queue poll interval in seconds
    task_timeout=300.0,    # Task timeout in seconds (default: 5 min)
)

Redis Queue Keys

• pixell:agents:{agent_id}:tasks - Main task queue
• pixell:agents:{agent_id}:processing - Tasks being processed
• pixell:agents:{agent_id}:dead_letter - Failed tasks
• pixell:tasks:{task_id}:progress - Progress pub/sub channel

Documentation

See the full documentation for detailed usage.

For SDK tutorials and advanced patterns, see SDK_TUTORIAL.md.

Release & Deployment

The pixell-sdk package is automatically published to PyPI via GitHub Actions (not manual PyPI API uploads).

Automatic Publishing

Every push to main triggers the publish workflow:

1. Tests run - All tests must pass
2. Version check - Compares version in pyproject.toml against PyPI
3. Publish - If version is new, automatically publishes to PyPI

Publishing a New Version

# 1. Update version in pyproject.toml
# 2. Commit and push to main
git add pyproject.toml
git commit -m "chore: bump version to X.Y.Z"
git push origin main

# 3. GitHub Actions handles the rest automatically

Manual Trigger

You can also trigger publishing manually via GitHub Actions:

1. Go to Actions → Publish to PyPI on Main Push
2. Click Run workflow
3. Optionally check Force publish to republish an existing version

Workflow Files

• .github/workflows/publish-main.yml - Main publish workflow (push to main)
• .github/workflows/publish.yml - Release-based publish (on GitHub releases)
• .github/workflows/test.yml - CI tests

Required Secrets

The following secret must be configured in GitHub repository settings:

• PYPI_API_TOKEN - PyPI API token for publishing

License

This project is licensed under the GNU Affero General Public License v3.0.

For organizations that do not wish to comply with AGPL-3.0 requirements, commercial licensing options are available. Contact us at engineering@pixell.global.