OpenAI-compatible API

Environment variables

Most OpenAI SDKs and tools pick these up automatically:

export OPENAI_BASE_URL="https://apixly.ai/v1"
export OPENAI_API_KEY="YOUR_API_KEY"

Chat completions

curl https://apixly.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

{
  "choices": [
    {
      "message": { "role": "assistant", "content": "Hello! How can I help?" },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 9,
    "completion_tokens": 12,
    "total_tokens": 21
  }
}

Add "stream": true to receive server-sent events, exactly like the official API.

Official SDK

Only the base_url changes — everything else is standard OpenAI SDK usage:

from openai import OpenAI

client = OpenAI(
    base_url="https://apixly.ai/v1",
    api_key="YOUR_API_KEY",
)
resp = client.chat.completions.create(
    model="gpt-5.5",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

Image generation

POST to the Images API; the result comes back as base64 image data under data[].b64_json.

curl https://apixly.ai/v1/images/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2",
    "prompt": "A cat astronaut floating in space, digital art",
    "size": "1024x1024"
  }'

{
  "created": 1782500000,
  "data": [{ "b64_json": "<BASE64_IMAGE_DATA>" }]
}

Image editing

Edit an existing image: POST multipart/form-data with the source image and a text prompt.

curl https://apixly.ai/v1/images/edits \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -F model="gpt-image-2" \
  -F image="@photo.png" \
  -F prompt="Add a wizard hat and a soft magical glow" \
  -F size="1024x1024"

{
  "created": 1782500000,
  "data": [{ "b64_json": "<BASE64_IMAGE_DATA>" }]
}

Responses API

POST /v1/responses: model, input and optional instructions; stream=true uses SSE. Use a model/group with Responses support. Chat Completions support alone does not guarantee Responses support. GET /v1/responses is a WebSocket upgrade route, not a stored-response listing.

POST /v1/responses/compact and POST /v1/responses/input_tokens are upstream-dependent Responses extensions. The gateway forwards only safe /v1/responses/*subpath segments; path registration does not guarantee the upstream implements that operation. Use final response usage for billing.

curl https://apixly.ai/v1/responses \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"YOUR_RESPONSES_MODEL","instructions":"Answer briefly.","input":"Hello","stream":true}'
# Read SSE through a terminal response event, including usage.

Model-specific request contracts

Discover available IDs with GET /v1/models using your key. The public price catalog is not a guarantee of current routing. Existing valid requests need no changes; optional fields are model-specific.

POST /v1/chat/completions: model + messages; stream=true uses SSE. Tools, JSON schema, image/PDF inputs and reasoning support depend on the selected model and route; text availability alone does not enable them.

gpt-6.1-sol: Responses reasoning.effort (Chat: reasoning_effort) accepts low/medium/high/xhigh/max/ultra. none/minimal and thinking.type=disabled are rejected before upstream execution. temperature/top_p are removed for this model; do not use them as sampling controls. Other GPT models have their own supported effort set.