StackFlow AI Platform
Model Library
Speech Recognition & Audio
Development Examples
OpenAI-Compatible API
Chat Completions API
The chat completions endpoint accepts a list of messages and uses a language model running on the device to generate a response. The endpoint path is /v1/chat/completions.
Before making a request, install llm-openai-api, the LLM service, and a matching model package on the device. Replace the device IP and model ID in the examples with your actual configuration. The examples below use an AX650 model for AI Pyramid; for other devices, select a platform-compatible model from the model library.
Curl Call
curl http://192.168.20.186:8000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-xxxxxxxx" \
-d '{
"model": "qwen2.5-1.5B-Int4-ax650",
"messages": [
{"role": "developer", "content": "You are a helpful home assistant."},
{"role": "user", "content": "Write a one-sentence bedtime story about a unicorn."}
]
}'
Python Call
from openai import OpenAI
client = OpenAI(
api_key="sk-",
base_url="http://192.168.20.186:8000/v1"
)
completion = client.chat.completions.create(
model="qwen2.5-0.5B-p256-ax630c",
messages=[
{"role": "developer", "content": "You are a helpful assistant."},
{"role": "user", "content": "Hello!"}
]
)
print(completion.choices[0].message)Request Parameters
| Parameter Name | Type | Required | Example Value | Description |
|---|---|---|---|---|
| messages | array | Yes | [{"role": "user", "content": "Hello"}] | Conversation history composed of multiple messages. Text, image, audio, and other modalities are supported (depending on the model). |
| model | string | Yes | qwen2.5-0.5B-p256-ax630c | The model ID used to generate the reply. Multiple models are supported. Refer to the Model Introduction for selection. |
| audio | - | No | - | Audio output is not currently supported. |
| function_call | - | No | - | Function calling is not currently supported. |
| max_tokens | integer | No | 1024 | The maximum number of tokens the model is allowed to generate. Output will be truncated if exceeded. |
| response_format | object | No | "json_object" | Specifies the output format of the model. Currently, only "json_object" is supported. |
Response Example
ChatCompletionMessage(content='Hello! How can I assist you today?', refusal=None, role='assistant', annotations=None, audio=None, function_call=None, tool_calls=None)Page Tools