zerogpu-sdk (Python SDK)

Official Python client for the ZeroGPU API: responses, chat completions, moderations, embeddings, and audio (transcriptions and speech), authenticated with your API key (x-api-key header). Published from ZeroGPU API SDKs (pypi/ in that repository).

Install

pip install zerogpu-sdk

Quick start — Responses

input may be a string or a list of message dicts / InputMessage, depending on the model.

import os
from zerogpu import ZerogpuApi

client = ZerogpuApi(
    api_key=os.environ["ZEROGPU_API_KEY"],
)

response = client.responses.create_response(
    model="zlm-v1-followup-questions-edge",
    input="In one short sentence, what is a habit tracker?",
    text={"format": {"type": "text"}},
)
print(response)

Optional metadata (Responses)

client.responses.create_response(
    model="gliner-multi-pii-v1",
    input="Your text…",
    metadata={"mask": "label", "usecase": "redact"},
)

Chat completions

from zerogpu import ChatMessage

client.chat.create_chat_completion(
    model="gliner-multi-pii-v1",
    messages=[ChatMessage(role="user", content="Your text…")],
    metadata={"mask": "label", "usecase": "redact"},
)

Moderations

Classify text against OpenAI's 13 safety categories. model defaults to zlm-v1-moderation-edge.

moderation = client.moderations.create_moderation(input="I want to hurt them.")
print(moderation.results[0].flagged, moderation.results[0].categories)

Embeddings

384-dimensional vectors from all-minilm-l6-v2 or bge-small-en-v1.5. Pass a list to embed a batch.

embeddings = client.embeddings.create_embedding(
    model="all-minilm-l6-v2",
    input=["first document", "second document"],
)
print(embeddings.data[0].embedding)

Audio

Transcribe with whisper-tiny. json (default) and verbose_json return a TranscriptionResponse; text, srt, and vtt return a string.

with open("speech.mp3", "rb") as f:
    transcript = client.audio.create_transcription(file=f)
print(transcript.text)

Generate speech with chatterbox-nano; the result is the audio bytes. Pass voice_sample to clone a voice from a reference clip.

audio = client.audio.create_speech(input="Hey, how are you today?", response_format="mp3")
with open("speech.mp3", "wb") as f:
    f.write(audio)

Every client has an async twin: AsyncZerogpuApi exposes the same methods with await.

API reference

• ZeroGPU docs
• Responses
• Chat completions
• Moderations
• Embeddings
• Transcriptions
• Speech

Repository

Source and release workflow: github.com/zerogpu/SDK (Python package lives under pypi/).