SPACE BUNNY API
Move from a working prompt to a product call.
Use the same Space Bunny workflow in your own application. Send long context, images or compatible video URLs, choose reasoning effort, and keep the API key behind your server.
from openai import OpenAI
client = OpenAI(
base_url="https://spacebunny.app/api/v1",
api_key=SPACE_BUNNY_API_KEY,
)
response = client.chat.completions.create(
model="stealth/space-bunny-alpha",
messages=[{"role": "user", "content": prompt}],
extra_body={"reasoning": {"effort": "low"}},
)ONE ENDPOINT, SEVERAL WORKFLOWS
Keep the request familiar. Add depth only when the task needs it.
Space Bunny follows a chat-completions shape so your application can start small and add context, media, structured output or tools as the workflow grows.
Large working context
Bring documents, repository files, research notes and conversation history into one request when the task depends on the wider picture.
Text and visual input
Ask about source code, screenshots, diagrams or compatible video URLs alongside a written instruction.
Five reasoning levels
Start with low for routine work and move toward medium, high, xhigh or max when deeper analysis earns its cost.
Text, JSON and tools
Request machine-readable JSON or approved tool calls, then validate the response before your application acts.
FROM PLAYGROUND TO PRODUCTION
A short path from an experiment to an integration.
You do not need to redesign the workflow when it moves from the browser to your server.
Try a real task
Use the Playground to test a prompt you can verify, then note the context, output shape and reasoning effort that worked.
Create a server-side key
Store the key in an environment variable or secret manager. Never place it in browser code, public prompts or source control.
Send the same request
Use the OpenAI client pattern, set the Space Bunny model ID, and add multimodal content, JSON or tools only when needed.
THE REQUEST SHAPE
Start with the smallest useful call.
The model and messages are the core fields. Reasoning, response format, tools and output limits are deliberate controls, not decoration.
{
"model": "stealth/space-bunny-alpha",
"messages": [
{
"role": "system",
"content": "Be careful and concise."
},
{
"role": "user",
"content": "Review this migration plan."
}
],
"reasoning": { "effort": "low" },
"response_format": { "type": "json_object" },
"max_completion_tokens": 2048
}APPLICATION-OWNED CONTROL
The model can propose. Your application decides.
Treat text, JSON and tool calls as model output. Validate schemas, permissions, resource scope, rate limits, idempotency and approval boundaries before any external side effect.
READY TO SEND A SIGNAL
Test the request before you wire the rest of the system.
Run one verifiable task in the Playground, then use the docs to move the same shape behind your server.