Chat and Text Completions

Chat Completions is the recommended text generation API. It supports multi-turn messages, streaming, tool calls, multimodal input and structured output. The legacy Text Completions API remains available for compatibility.

Endpoints

Method	Path	Description
`POST`	`/v1/chat/completions`	OpenAI-compatible chat completions
`POST`	`/v1/completions`	Legacy text completions using `prompt`

Request Fields

Field	Type	Required	Description
`model`	string	Yes	Model ID
`messages`	array	Required for Chat	Conversation messages. FIM models may use `prefix` / `suffix` instead
`prompt`	string or array	Required for Completions	Legacy text completion input
`stream`	boolean	No	Return an SSE stream
`stream_options`	object	No	Streaming options
`max_tokens`	integer	No	Maximum output tokens
`max_completion_tokens`	integer	No	Newer maximum output tokens field
`temperature`, `top_p`, `top_k`	number	No	Sampling controls
`stop`	string or array	No	Stop sequence
`tools`, `tool_choice`	array/object	No	Tool calling configuration
`parallel_tool_calls`	boolean	No	Allow parallel tool calls
`response_format`	object	No	Structured output such as JSON object or JSON schema
`reasoning_effort`	string	No	Reasoning effort for reasoning models
`modalities`, `audio`	object	No	Multimodal or audio model parameters
`metadata`	object	No	Pass-through metadata
`store`, `service_tier`, `safety_identifier`	any	No	OpenAI extension fields; channel settings may filter them

curl

curl https://api.tensoraxis.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $TENSORAXIS_API_KEY" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {"role": "system", "content": "You are a concise assistant."},
      {"role": "user", "content": "Describe TENSORAXIS in one sentence"}
    ],
    "temperature": 0.7
  }'

Python

from openai import OpenAI

client = OpenAI(
    api_key="your-tensoraxis-api-key",
    base_url="https://api.tensoraxis.com/v1",
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

TypeScript

import OpenAI from 'openai'

const client = new OpenAI({
  apiKey: process.env.TENSORAXIS_API_KEY,
  baseURL: 'https://api.tensoraxis.com/v1',
})

const response = await client.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: 'Hello!' }],
})
console.log(response.choices[0].message.content)

Streaming

{
  "model": "gpt-4o",
  "messages": [{"role": "user", "content": "Write a three-line poem"}],
  "stream": true
}

With stream enabled, the response is SSE. Clients should read data: events until the stream finishes.

Multimodal Input

messages[].content can be a string or an array of content parts. The backend recognizes text, image_url, input_audio and related content types; actual support depends on the selected model and upstream channel.