Anthropic-Compatible API
ZBStream also exposes an Anthropic-compatible Messages API (/v1/messages). Point the official Anthropic SDK at ZBStream by changing the base URL — cURL, Python, Node.js, and Go are all supported.
Endpoint & Authentication
The Anthropic-compatible base URL has no /v1 suffix — the official SDK appends /v1/messages automatically. Authenticate with the x-api-key header (Authorization: Bearer is also accepted) and include anthropic-version.
x-api-key: $ZBSTREAM_API_KEY
anthropic-version: 2023-06-01
Content-Type: application/json| Parameter | Value | Description |
|---|---|---|
| Base URL | https://api.zbstream.com | The SDK appends /v1/messages automatically |
| Auth | x-api-key: <key> | Same ZBStream API key as the OpenAI interface |
| Version | anthropic-version: 2023-06-01 | Recommended for forward compatibility |
Send a Messages Request
The Messages API requires max_tokens, and the system prompt is a top-level field rather than part of the messages array. The answer is returned in the content array; usage is reported in usage.input_tokens / usage.output_tokens.
curl https://api.zbstream.com/v1/messages \
-H "x-api-key: $ZBSTREAM_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4-pro",
"max_tokens": 1024,
"messages": [
{ "role": "user", "content": "你好，请介绍一下 ZBStream API。" }
]
}'Streaming Responses
Add "stream": true to receive an SSE stream. Event types include message_start, content_block_delta, and message_stop.
curl https://api.zbstream.com/v1/messages \
-H "x-api-key: $ZBSTREAM_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4-pro",
"max_tokens": 1024,
"stream": true,
"messages": [{ "role": "user", "content": "请逐步回答" }]
}'Common Request Fields
| Field | Required | Description |
|---|---|---|
| model | Yes | Model id, matching /v1/models |
| messages | Yes | Message array with user / assistant roles |
| max_tokens | Yes | Maximum output tokens |
| system | No | Top-level system prompt |
| temperature | No | Sampling temperature, model-dependent range |
| stream | No | Enable SSE streaming |
| tools | No | Tool definitions following the Anthropic schema |
Differences from the OpenAI API
- max_tokens is required by the Messages API but optional in the OpenAI API.
- The system prompt is a top-level field, not a role=system message.
- The response body is an array of content blocks (text, tool use, …) rather than a single message.content string.
- Usage is reported as usage.input_tokens / usage.output_tokens, corresponding to prompt_tokens / completion_tokens in the OpenAI API.
- Both interfaces share the same API key and credit account, billed at each model's rates.