Skip to content

ChatCompletion API

The ChatCompletion API provides access to AI language models supported by Unique AI for generating text completions. (The request does not modify the frontend and is not related to any chats in the unique platform)

Overview

Generate AI responses using various language models with OpenAI-compatible message format.

Methods

unique_sdk.ChatCompletion.create - Generate AI completions

Send messages to an AI model and receive completions.

Parameters:

  • user_id (required)
  • company_id (required)
  • model (required) - Model identifier (e.g., "AZURE_GPT_4o_2024_1120")
  • messages (required) - Array of message objects in OpenAI format
  • options (optional) - Model parameters like temperature

Available Models:

Common models include:

  • AZURE_GPT_4o_2024_1120 - GPT-4 Turbo
  • AZURE_GPT_4_32K_0613 - GPT-4 with 32K context
  • AZURE_o3_2025_0416 - Latest o3 model

Tip

Use LLM Models API to get the current list of available models.

Example - Basic Completion:

completion = unique_sdk.ChatCompletion.create(
    user_id=user_id,
    company_id=company_id,
    model="AZURE_GPT_4o_2024_1120",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Explain quantum computing in simple terms."}
    ]
)

print(completion.choices[0].message.content)

Example - With Options:

completion = unique_sdk.ChatCompletion.create(
    user_id=user_id,
    company_id=company_id,
    model="AZURE_GPT_4o_2024_1120",
    messages=[
        {"role": "system", "content": "You are a creative storyteller."},
        {"role": "user", "content": "Write a short story about a robot."}
    ],
    options={
        "temperature": 0.8,      # Higher for more creativity
        "max_tokens": 500,
        "top_p": 0.9
    }
)

print(completion.choices[0].message.content)

Example - Multi-turn Conversation:

messages = [
    {"role": "system", "content": "You are a helpful coding assistant."},
    {"role": "user", "content": "How do I reverse a string in Python?"},
    {"role": "assistant", "content": "You can use slicing: `reversed_str = my_str[::-1]`"},
    {"role": "user", "content": "Can you show me another way?"}
]

completion = unique_sdk.ChatCompletion.create(
    user_id=user_id,
    company_id=company_id,
    model="AZURE_GPT_4o_2024_1120",
    messages=messages
)

print(completion.choices[0].message.content)

Message Format

Messages follow the OpenAI Chat API format:

{
    "role": "system" | "user" | "assistant",
    "content": "message text"
}

Roles:

  • system - Instructions for the AI's behavior
  • user - User messages
  • assistant - AI responses (for conversation history)

Use Cases

Structured Output Generation 
completion = unique_sdk.ChatCompletion.create(
    user_id=user_id,
    company_id=company_id,
    model="AZURE_GPT_4o_2024_1120",
    messages=[
        {
            "role": "system",
            "content": "Extract information as JSON: {\"name\": \"\", \"email\": \"\", \"phone\": \"\"}"
        },
        {
            "role": "user",
            "content": "John Doe, john@example.com, 555-0123"
        }
    ],
    options={"temperature": 0}
)

import json
extracted_data = json.loads(completion.choices[0].message.content)

Response Format

{
    "choices": [
        {
            "message": {
                "role": "assistant",
                "content": "Response text here..."
            },
            "finish_reason": "stop",
            "index": 0
        }
    ],
    "usage": {
        "prompt_tokens": 50,
        "completion_tokens": 100,
        "total_tokens": 150
    }
}

Best Practices

Control Temperature by Use Case 
# Factual/analytical tasks - low temperature
factual = unique_sdk.ChatCompletion.create(
    model="AZURE_GPT_4o_2024_1120",
    messages=[...],
    options={"temperature": 0.2}
)

# Creative tasks - high temperature
creative = unique_sdk.ChatCompletion.create(
    model="AZURE_GPT_4o_2024_1120",
    messages=[...],
    options={"temperature": 0.8}
)