HTTP clients
The text contract uses POST https://apiwild.com/v1/chat/completions, an API WILD bearer key and an Idempotency-Key. Use an exact model ID from the selected catalogue, subject to route activation and the key’s chat capability. Reuse the same request identifier and identical JSON body if the result is uncertain.
OpenAI-compatible clients
Use the OpenAI SDK Chat Completions client with base URL https://apiwild.com/v1 and an API WILD key kept on your server. Supply an Idempotency-Key, use an exact activated model ID and disable automatic SDK retries. Text replies, function-call data and buffered SSE output use the documented subset. Responses, Assistants and Anthropic message adapters are not supported.
import os
import uuid
from openai import OpenAI
client = OpenAI(
api_key=os.environ["APIWILD_API_KEY"],
base_url="https://apiwild.com/v1",
max_retries=0,
)
request_id = str(uuid.uuid4()) # Save this ID with the request body before sending.
response = client.chat.completions.create(
model="MiniMax-M2.7-highspeed",
messages=[{"role": "user", "content": "Reply with the single word OK."}],
max_tokens=128,
stream=False,
extra_headers={"Idempotency-Key": request_id},
)
print(response.choices[0].message.content)Read availability and usage
Public model metadata is a catalogue reference. Authenticated model and usage endpoints describe your available routes and account records. These reads do not activate inference or add credit. Follow the API reference before connecting a client.
Business systems
Manual web search, Wikipedia article summaries and arithmetic are available through the authenticated research utility endpoint. Browser dictation and read aloud are available where supported. Hosted audio APIs are planned for a later sprint. Messenger, CRM writes and autonomous repository execution require separate integrations.