targon-sdk 0.1.0

Targon sdk to interacte with the targon rentals and serverless

Targon SDK

Modal-style Python SDK for deploying serverless functions to Targon.

Installation

cd targon-sdk
pip install -e .

Quick Start

1. Setup credentials:

targon setup
# Enter your API key when prompted

2. Create 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

Examples

Simple Web Server

import targon
import subprocess

app = targon.App("web-server")

@app.function()
@targon.web_server(port=8000)
def serve():
    subprocess.Popen("python -m http.server 8000", shell=True)

vLLM Inference Server

See examples/vllm_qwen.py for a complete example of deploying vLLM with Qwen models.

targon deploy examples/vllm_qwen.py

Architecture

• targon-sdk: Python client library (this directory)
• heim: Image builder + container registry
• tha: Serverless backend (Knative)

API Reference

targon.App(name)

Create a new application.

@app.function(image=None, gpu=None, cpu=None, memory=None, timeout=300)

Decorator to register a function with the app.

@targon.web_server(port, startup_timeout=5.0)

Decorator for functions that start a web server process.

@targon.fastapi_endpoint(method="GET")

Decorator for FastAPI-style HTTP endpoints.

CLI Commands

• targon deploy <file.py> - Deploy an app
• targon list - List deployments
• targon logs <uid> - View function logs
• targon setup - Configure API credentials

Development

The SDK follows Modal's design patterns:

• Decorator-based function registration
• Cloudpickle serialization for function transport
• Image building with Dockerfile generation
• Container registry integration
• Knative deployment backend