AI Agent Engineer

AI Agent Engineer

Qwen API Calls

3 min read

Qwen API Calls

Official source

Based on Alibaba Cloud Model Studio documentation for the first Qwen API call, text generation, streaming, and Function Calling. Start with the OpenAI-compatible interface; learn the DashScope SDK as native capabilities require it.

Common entry points and regional boundary

FormUse
client.chat.completions.create()Text, history, images, JSON, and tool calls.
stream=TrueStream output.
client.embeddings.create()OpenAI-compatible text vectors.
dashscope.Generation.call()Native DashScope text generation.

The key and regional endpoint must match:

Regionbase_url
Mainland Chinahttps://dashscope.aliyuncs.com/compatible-mode/v1
Singapore (international)https://dashscope-intl.aliyuncs.com/compatible-mode/v1

A dedicated Workspace can use an endpoint containing {WorkspaceId}; copy it from the current console or official documentation.

python -m pip install --upgrade openai
$env:DASHSCOPE_API_KEY = Read-Host 'DASHSCOPE_API_KEY' -MaskInput
import os
from openai import OpenAI
 
client = OpenAI(
    api_key=os.environ["DASHSCOPE_API_KEY"],
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
)
messages = [
    {"role": "system", "content": "You are a concise technical teacher."},
    {"role": "user", "content": "Explain an API call in three points."},
]
response = client.chat.completions.create(model="qwen3.7-plus", messages=messages)
print(response.choices[0].message.content, response.usage)

On 2026-07-22, the official text-model page listed qwen3.7-plus as a balanced Agent/general-task recommendation with 1M context, thinking, Function Calling, built-in tools, and structured-output support. The matrix is not proof of entitlement for every region, account, endpoint, or plan. Re-select low-latency, code, vision, or long-context models in the model list and target-region console.

History, streaming, vision, and JSON

The compatible interface is stateless; append assistant text and the next user turn to the same ordered messages list before sending it again.

stream = client.chat.completions.create(
    model="qwen3.7-plus",
    messages=[{"role": "user", "content": "Write a Python API study plan."}],
    stream=True,
    stream_options={"include_usage": True},
)
for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].delta.content or "", end="", flush=True)
    elif chunk.usage:
        print("\nTokens:", chunk.usage.total_tokens)

The final usage-only chunk can have no choices; guard it before indexing.

import json
 
vision = client.chat.completions.create(
    model="qwen3-vl-plus",
    messages=[{"role": "user", "content": [
        {"type": "image_url", "image_url": {"url": "https://help-static-aliyun-doc.aliyuncs.com/file-manage-files/zh-CN/20241022/emyrja/dog_and_girl.jpeg"}},
        {"type": "text", "text": "Explain the chart trend and name its axes."},
    ]}],
)
 
structured = client.chat.completions.create(
    model="qwen3.7-plus",
    messages=[{"role": "user", "content": "Return JSON with topic, days, and tasks."}],
    response_format={"type": "json_object"},
)
data = json.loads(structured.choices[0].message.content)

Do not mix text-model, Qwen-VL, and Qwen-Omni input shapes. JSON mode needs an explicit JSON instruction plus local parsing and field/range validation.

Tools, embeddings, and native SDK

Use the compatible tool loop exactly as a normal Chat Completions contract: validate model JSON locally; execute only allowlisted, authorized tools; append one role: "tool" result with the matching tool_call_id; then send the full history for a final answer. Some thinking models require retaining reasoning_content; check current Function Calling notes for the selected model.

embedding_response = client.embeddings.create(
    model="text-embedding-v4",
    input=["Python is suitable for rapid development.", "Rust emphasizes performance and memory safety."],
)
vectors = [item.embedding for item in embedding_response.data]
print(len(vectors), len(vectors[0]))

Indexing and query need the same embedding model and dimension; available region, dimensions, and input type vary by model.

python -m pip install --upgrade dashscope
import os
import dashscope
from dashscope import Generation
 
dashscope.api_key = os.environ["DASHSCOPE_API_KEY"]
response = Generation.call(
    model="qwen3.7-plus",
    messages=[{"role": "user", "content": "Explain an API call in three points."}],
    result_format="message",
)
print(response.output.choices[0].message.content)

The compatible interface eases migration; DashScope SDK exposes Alibaba-native features. Select one primary call style in a project and do not read both response objects as though their shapes were identical.

Common mistakes and sources

  • Mismatched key and regional endpoint.

  • Directly indexing choices in the final usage-only stream chunk.

  • Mixing text, vision, and Omni input shapes.

  • Returning no tool_call_id-matched tool result.

  • Relying on aliases while ignoring upgrade and retirement notices.

  • Mixing OpenAI SDK and DashScope SDK response conventions.

  • First Qwen API call

  • Text generation

  • Streaming

  • Function Calling

  • Vision

  • Embeddings

Return to Vendor AI API Reference Index; general HTTP reliability remains in API Learning Path.

Backlinks