sstudio 0.0.2

SmartStudio Post-Training SDK — Python client and CLI for the SmartStudio platform.

sstudio

sstudio is the Python SDK and command-line client for the SmartStudio post-training platform. It wraps the platform's HTTP APIs behind a small, stable Python surface and an sstudio CLI — handling authentication, request normalization, and unified job tracking.

Install

pip install sstudio

Authenticate

sstudio login --api-key sk-xxxx
sstudio whoami

Or from Python:

from sstudio import SmartStudioClient

with SmartStudioClient(api_key="sk-xxxx") as client:
    print(client.me())

Configure

Project settings live in sstudio.yaml (base model, platform endpoint, storage paths, …). Values are resolved with the precedence command-line flag > SSTUDIO_* env var > sstudio.yaml > built-in default.

sstudio config validate
sstudio config show

Scope

The SDK is organized around the post-training workflow:

Module	Purpose
data	Document parsing and training/eval dataset generation
train	SFT-LoRA / SFT-Full / SFT-Plus / DPO / GKD fine-tuning
eval	Benchmark scoring and LLM/VLM-as-judge evaluation
serve	Model deployment
job	Unified async job status, logs and results

Requirements

Python 3.8+

---

© SmartStudio. All rights reserved.