Reference

OpenAI Chat Completions compatible API

An OpenAI Chat Completions compatible API accepts the same JSON schema as OpenAI's /v1/chat/completions endpoint.

That means you can keep your existing SDKs and just switch the base URL and API key.

Streaming, structured output, function calling, and multimodal inputs follow the same overall patterns as OpenAI's Chat Completions API.

Definition of OpenAI Chat Completions compatible API

A compatible API mirrors OpenAI's chat completions request and response format so existing clients can talk to it without rewrites.

Why OpenAI Chat Completions compatible API matters

Drop-in compatibility reduces migration time.
Reuse existing SDKs, tools, and client-side helpers.
Keep the same message format across providers.
Extend your integration with streaming, vision, structured output, and retry logic without retooling.

How it works

Point your client at https://api.abliteration.ai/v1.
Send a Bearer token using your abliteration.ai API key.
Set model to an abliteration.ai model id such as abliterated-model.
Enable streaming with stream: true if desired.
Use content arrays for vision-capable models when sending images.
Request structured JSON with response_format: { type: "json_object" }.
Provide functions and function_call to enable function calling.

Example request

curl https://api.abliteration.ai/v1/chat/completions \
  -H "Authorization: Bearer $ABLIT_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "abliterated-model",
    "messages": [{"role":"user","content":"Summarize this in one sentence."}],
    "stream": true
  }'

FAQ

Frequently Asked Questions

Do I need to change my message schema?

No. Use the same role/content message format as OpenAI Chat Completions.

Does compatibility include streaming and vision?

Yes. Streaming works with stream: true. Vision works when you use a vision-capable model and send images via the content array format.

Does it support function calling and structured output?

Yes. Send functions plus function_call for tool invocation and response_format: { type: "json_object" } for structured JSON.