Chat Completions

Create a chat completion using the OpenAI-compatible schema. The same endpoint can target OpenAI, Claude, Gemini, Qwen, DeepSeek and other backends — BUZZ resolves the upstream from the model field and converts protocols when needed.

Authentication

BUZZ accepts the standard OpenAI bearer-token form:

The Anthropic x-api-key header is honored on /v1/messages and /v1/models; for /v1/chat/completions use Authorization: Bearer.

Authorization: Bearer <YOUR_BUZZ_KEY>
Content-Type: application/json

Example request

curl -X POST https://buzzai.cc/v1/chat/completions \
  -H "Authorization: Bearer $BUZZ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-haiku-4-5-20251001",
    "messages": [
      {"role": "user", "content": "reply with exactly: hello world"}
    ],
    "max_tokens": 32
  }'

from openai import OpenAI

client = OpenAI(
    base_url="https://buzzai.cc/v1",
    api_key="",
)

resp = client.chat.completions.create(
    model="claude-haiku-4-5-20251001",
    messages=[
        {"role": "user", "content": "reply with exactly: hello world"}
    ],
    max_tokens=32,
)

print(resp.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://buzzai.cc/v1",
  apiKey: process.env.BUZZ_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "claude-haiku-4-5-20251001",
  messages: [
    { role: "user", content: "reply with exactly: hello world" },
  ],
  max_tokens: 32,
});

console.log(resp.choices[0].message.content);

Response

{
  "id": "chatcmpl-AbCdEf...",
  "object": "chat.completion",
  "created": 1748246400,
  "model": "claude-haiku-4-5-20251001",
  "choices": [
    {
      "index": 0,
      "message": {"role": "assistant", "content": "hello world"},
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {"prompt_tokens": 14, "completion_tokens": 2, "total_tokens": 16}
}

Body parameters

BUZZ accepts every field defined in the OpenAI CreateChatCompletionRequest schema. The table below lists all 35 official fields plus top_k, which is widely used by non-OpenAI upstreams.

Core fields

Sampling

Tools and structured output

Reasoning and modalities

Caching, accounting and identity

When the model is a Claude model

The same /v1/chat/completions endpoint can target Claude models (e.g. claude-haiku-4-5-20251001, claude-sonnet-4-6, claude-opus-4-7). BUZZ rewrites the request into the Anthropic Messages API and converts the response back to OpenAI shape — but several OpenAI-only fields have no Anthropic equivalent and are silently dropped.

Fields silently dropped on Claude paths

The Claude request builder does not include any of these fields, so values you pass are ignored. Plan accordingly:

Sampling overrides on Claude

When a Claude model runs in thinking mode or is opus-4-7, sampling parameters are overridden by BUZZ before the request leaves the gateway:

• For opus-4-7 with thinking suffix: temperature, top_p, top_k are all cleared, thinking.type=adaptive, output_config.effort=high.
• For other models with -thinking suffix: thinking.type=enabled, budget_tokens = 80% of max_tokens, top_p cleared, temperature forced to 1.0.
• Without thinking, temperature / top_p / top_k pass through normally.

Field translations

Response

Non-streaming responses follow the standard OpenAI chat.completion shape.

usage object

OpenAI fields are returned verbatim. BUZZ adds extension fields for cache and reasoning visibility.

Streaming

Set "stream": true to receive a Server-Sent Events stream. Each event is one data: {...} line, terminated by data: [DONE].

Stream chunk shape

{
  "id": "chatcmpl-...",
  "object": "chat.completion.chunk",
  "created": 1748246400,
  "model": "claude-haiku-4-5-20251001",
  "choices": [
    {
      "index": 0,
      "delta": {"role": "assistant", "content": "Hello"},
      "logprobs": null,
      "finish_reason": null
    }
  ]
}

Set stream_options.include_usage = true to receive a final chunk where choices is empty and usage is populated.

Claude-stream translations

When the model is a Claude model, BUZZ converts Anthropic event: stream into OpenAI chunks:

Identifying which upstream answered

Because both OpenAI and Claude are reachable through this endpoint, you can fingerprint which family handled a given response without parsing the model name:

Errors

OpenAI-format error envelope:

{
  "error": {
    "code": "",
    "message": "Invalid token (request id: 202605260708...)",
    "type": "buzz_error"
  }
}

See also

• POST /v1/messages — native Anthropic format
• GET /v1/models — discover available models
• Tool Use concept
• Prompt Caching concept
• Streaming guide