studiolm 0.1.0

Python SDK for the StudioLM API – chat completions and image generation

studiolm

Official Python SDK for the StudioLM API — chat completions and image generation.

---

Installation

pip install studiolm

For local development (from this repo):

pip install -e .

---

Quick Start

import studiolm

client = studiolm.Client(api_key="sk-...")

# Generate an image and save it
image = client.generate(
    "A serene mountain lake at dawn, cinematic lighting, 8K",
    model="imagen-v3",
    size="1024x1024",
    style="vivid",
)
image.save("masterpiece.png")

# Chat completions
response = client.chat.completions.create(
    model="gemma-3-12b-it-qat",
    messages=[{"role": "user", "content": "What is the capital of France?"}],
)
print(response["choices"][0]["message"]["content"])

You can also set the API key via environment variable:

export STUDIOLM_API_KEY="sk-..."

import studiolm

client = studiolm.Client()  # reads STUDIOLM_API_KEY automatically

---

Image Generation

Parameters

Parameter	Type	Default	Description
prompt	str	required	Text description of the image
model	str	auto	Model display name (e.g. "imagen-v3")
size	str	"1024x1024"	Output resolution — see size table below
style	str	"vivid"	preview, natural, vivid, upscaled
aspect_ratio	str	derived from size	square, portrait, landscape
negative_prompt	str	—	Elements to avoid
seed	int	random	Reproducibility seed
response_format	str|list	"url"	url, b64_json, hxd, or a list

Supported sizes

Size	Aspect ratio
512x512	square
768x768	square
1024x1024	square
832x1216	portrait
512x768	portrait
1216x832	landscape
768x512	landscape

Add custom sizes at runtime:

import studiolm
studiolm.SIZE_PRESETS["640x640"] = "square"

Examples

# Portrait, natural style
image = client.generate(
    "A knight standing in a misty forest",
    model="imagen-v3",
    size="832x1216",
    style="natural",
)
image.save("knight.png")

# Landscape with negative prompt and seed
image = client.generate(
    "Cyberpunk city at night",
    size="1216x832",
    style="vivid",
    negative_prompt="blurry, low quality",
    seed=42,
)
image.save("city.png")

# Image-to-image from URL
image = client.generate(
    "Transform into Studio Ghibli style",
    reference_image_url="https://example.com/photo.jpg",
    denoising_strength=0.65,
)
image.save("ghibli.png")

# Image-to-image from local file
import base64
with open("my_photo.png", "rb") as f:
    b64 = base64.b64encode(f.read()).decode()

image = client.generate(
    "Make it look like a watercolor painting",
    reference_image=f"data:image/png;base64,{b64}",
    denoising_strength=0.5,
)
image.save("watercolor.png")

# Get URL + base64 in one request
image = client.generate(
    "A galaxy nebula",
    response_format=["url", "b64_json"],
)
print(image.url)
image.save("nebula.png")  # uses b64_json for saving

List image models

models = client.images.available_models()
for m in models:
    print(m["display_name"], "-", m["description"])

---

Chat Completions

# Basic
response = client.chat.completions.create(
    model="gemma-3-12b-it-qat",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain quantum entanglement briefly."},
    ],
    temperature=0.7,
    max_tokens=500,
)
print(response["choices"][0]["message"]["content"])

# Streaming
for chunk in client.chat.completions.create(
    model="gemma-3-12b-it-qat",
    messages=[{"role": "user", "content": "Write a haiku about coding"}],
    stream=True,
):
    delta = chunk["choices"][0].get("delta", {})
    if "content" in delta:
        print(delta["content"], end="", flush=True)
print()

# Web search (smart mode)
response = client.chat.completions.create(
    model="gemma-3-27b-it-qat",
    messages=[{"role": "user", "content": "What happened in AI news today?"}],
    web_search="auto",
)

# JSON mode
response = client.chat.completions.create(
    model="gemma-3-12b-it-qat",
    messages=[{"role": "user", "content": "Return a JSON list of 3 fruits."}],
    response_format="json",
)

---

Models

# List available chat/text models
models = client.models.list()
for m in models:
    print(m["id"])

# List image generation models
image_models = client.images.available_models()

---

Context manager

with studiolm.Client(api_key="sk-...") as client:
    image = client.generate("A sunset over the ocean")
    image.save("sunset.png")

---

Custom base URL (self-hosted)

client = studiolm.Client(
    api_key="sk-...",
    base_url="http://localhost:8000",
)

Or via environment variable:

export STUDIOLM_BASE_URL="http://localhost:8000"

---

License

MIT