Create a chat completion
Send a JSON POST request to https://api.arnict.com/v1/chat/completions with your API key. Include a public model ID and a messages array.
Node.js
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.arnict.com/v1",
apiKey: process.env.ARNICT_API_KEY,
});
const completion = await client.chat.completions.create({
model: "zai/glm-5.3-flash-uncensored",
messages: [{ role: "user", content: "Say hello in one sentence." }],
max_tokens: 256,
});
console.log(completion.choices[0].message.content);Compatibility scope
Arnict supports the chat-completions API. Other APIs or features offered by your SDK are not automatically available. Model-specific capabilities are listed in the catalog.
Request parameters
| Parameter | Type | How to use it |
|---|---|---|
| model | string, required | An available public model ID. |
| messages | array, required | Conversation messages with a role and content. Include previous turns when your application needs conversation context. |
| max_tokens | positive integer | An output limit within the selected model’s published maximum. |
| max_completion_tokens | positive integer | An alternative output-limit field. Use one output-limit field per request. |
| stream | boolean | Use true for streamed events. The default is false. |
| stream_options | object | For streaming, use {"include_usage": true} to request token usage. |
| temperature | number | A sampling setting, where supported by the selected model. |
| top_p | number | A sampling setting, where supported by the selected model. |
| stop | string or array | Stop sequences, where supported by the selected model. |
Output limits vary by model. Leave room for the input and output within the model’s context window. Start with a modest output limit and increase it when your application needs longer responses.
Check optional capabilities
Tool calls, structured output, reasoning controls and image inputs depend on the model. Use the model page and validate your request with that model before relying on a feature.
Read the result
| Field | Meaning |
|---|---|
| id | Identifier for the completion. |
| model | The public model ID used for the request. |
| choices | The generated choices. For a standard request, read choices[0].message.content. |
| choices[].finish_reason | Why generation ended. For example, stop indicates completion and length indicates a limit was reached. |
| usage | Token counts, when provided. See Pricing for how token categories affect cost. |
Read X-Request-ID from the HTTP response headers when diagnosing a request. Keep the ID and status code in your application’s operational logs; avoid logging credentials or sensitive content.
- Stream the responseReceive incremental content instead of waiting for a complete JSON response.
- Understand token pricingRead input, cached input and output usage.