Targon SDK

Python SDK for building and deploying serverless applications on the Targon platform. It includes the command-line client, decorator-friendly runtime APIs, and publishing utilities.

---

Requirements

• Python 3.9+
• Targon account with API access
• Optional: Docker (for advanced image customization)

---

Installation

From PyPI

pip install targon-sdk

From source

git clone https://github.com/manifold-inc/targon-sdk.git
cd targon-sdk
pip install -e .

---

Quick Start

1. Configure credentials

   targon setup
   # Follow the prompts to store your API key securely

2. Define an app

   # my_app.py
   import targon
   import subprocess
   
   app = targon.App("my-first-app")
   
   @app.function()
   @targon.web_server(port=8000)
   def serve():
       subprocess.Popen("python -m http.server 8000", shell=True)
   
   @app.local_entrypoint()
   def main():
       print("Hello from Targon!")

3. Deploy

   targon deploy my_app.py

4. Iterate locally

   targon run my_app.py --message "hello"

---

CLI Overview

Command	Description
targon setup	Store or update API credentials.
targon deploy <file.py>	Build and deploy an app module.
targon run <file.py>	Execute a local entrypoint in an ephemeral session.
targon app list	List deployed/running apps.
targon app functions <app_id>	Inspect functions for a given app.
targon app delete <app_id>	Delete an app and its deployments.

---

Examples

Visit examples/ for ready-to-run templates:

• gettin_started/getting_started.py – minimal hello world workflow.
• gettin_started/web_endpoint_simple.py – FastAPI endpoint via @targon.fastapi_endpoint.
• web/web_endpoint_asgi.py – full ASGI application deployment.
• llm/vllm_example.py – vLLM inference service running behind @targon.web_server.
• gen-ai/ – generated media workloads.

Deploy an example directly:

targon deploy examples/gettin_started/getting_started.py

---

Development Notes

• The SDK mirrors common serverless-style patterns: decorator-based registration, cloudpickle serialization, and runtime-provisioned images.
• Core packages used at runtime:
  • cloudpickle for function transport
  • grpcio / grpcio-tools for RPC communication
  • aiohttp for async HTTP interactions
• Tests live under tests/. Run them with pytest (optional dependency).
• When contributing protocol changes, regenerate stubs with the matching grpcio-tools version to avoid runtime mismatches.

---

Support & Feedback

Open an issue or reach out to dev@manifold.inc for questions, feature requests, or bug reports. Contributions are welcome!