OpenAI API Python Sample Code and Example Output (GPT-4o Compatible)

In this article, we introduce sample code for using the OpenAI API.

The OpenAI API is not limited to OpenAI's own service; other inference engines such as vLLM also expose OpenAI-compatible API servers, and it is becoming the de facto standard for LLM serving APIs. So it is worth getting familiar with the coding conventions.

A simple code sample for accessing the API of OpenAI's GPT series looks like the following. Let's receive the generated output incrementally via streaming.

Sample Code: Quick Start

import asyncio
import os
import traceback

from openai import AsyncOpenAI


async def main() -> None:
    try:
        # Specify the model name
        # model="gpt-4-turbo" # $10.00/MTok for input ,$30.00/MTok for output
        # model="gpt-4o" # $5.00/MTok for input ,$15.00/MTok for output
        model = "gpt-3.5-turbo-0125"  #

        # Get the API key from an environment variable
        api_key = "your api key"
        
        client = AsyncOpenAI(
            api_key=api_key
        )

        stream = await client.chat.completions.create(
            model=model,
            stream=True,
            messages=[
                {"role": "system", "content": "You are a sincere and helpful assistant."},
                {"role": "user", "content": "Hello"}
            ],
            stream_options={"include_usage": True},  # Output usage (number of input/output tokens)
        )

        async for chunk in stream:
            print(f"chunk__{chunk}")
    except Exception as e:
        print(f"An unexpected error occurred: {e}\n{traceback.format_exc()}")
    finally:
        pass


asyncio.run(main())

Example Output

ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content='', function_call=None, role='assistant', tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content='Hello', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' there', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content='!', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' How', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' can', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' I', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' be', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' of', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' help', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' to', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' you', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=' today', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content='?', function_call=None, role=None, tool_calls=None), finish_reason=None, index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[Choice(delta=ChoiceDelta(content=None, function_call=None, role=None, tool_calls=None), finish_reason='stop', index=0, logprobs=None)], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=None)
ChatCompletionChunk(id='chatcmpl-9iGtdyZ43HFebZV22QOaZPIzgeStA', choices=[], created=1234567890, model='gpt-3.5-turbo-0125', object='chat.completion.chunk', system_fingerprint=None, usage=CompletionUsage(completion_tokens=14, prompt_tokens=31, total_tokens=45))

Sample Code: Parsing the Streamed Chunk Contents

Let's parse each chunk and extract the various pieces of data it contains.

import asyncio
import traceback

from openai import AsyncOpenAI


async def main() -> None:
    try:
        model = "gpt-3.5-turbo-0125"
        api_key = "your api key"
        client = AsyncOpenAI(api_key=api_key)

        stream = await client.chat.completions.create(
            model=model,
            stream=True,
            messages=[
                {"role": "system", "content": "You are a sincere and helpful assistant."},
                {"role": "user", "content": "Hello"}
            ],
            stream_options={"include_usage": True},  # Output usage
        )

        first_chunk = None
        last_chunk = None
        finish_reason = None
        full_content = ""
        role = None
        created = None
        model = None
        completion_id = None

        async for chunk in stream:

            object_type = chunk.object

            if object_type == "chat.completion.chunk":

                if first_chunk is None:
                    # On the first chunk
                    first_chunk = chunk

                    model = first_chunk.model
                    created = first_chunk.created
                    completion_id = first_chunk.id

                    if chunk.choices:
                        first_choice = chunk.choices[0]
                        role = first_choice.delta.role

                    # Information available from the first chunk
                    print(f"completion_id: {completion_id}")
                    print(f"created: {created}")

                    # Information available only from the first chunk
                    print(f"model: {model}")
                    print(f"role: {role}")

                    print("streaming text: ", end="", flush=True)

                last_chunk = chunk

                if chunk.choices:
                    first_choice = chunk.choices[0]

                    if first_choice.delta.content:
                        # Text generated in this iteration
                        delta_str = first_choice.delta.content

                        print(delta_str, end="", flush=True)  # Print the generated text incrementally

                        full_content += delta_str  # Append to the full text

                    if finish_reason is None:
                        finish_reason = first_choice.finish_reason

        print()

        if last_chunk:
            # Process the data from the last chunk

            usage = last_chunk.usage

            print(f"Full Content: {full_content}")
            print(f"Finish Reason: {finish_reason}")

            if usage:
                print(f"ttl tokens: {usage.total_tokens}")
                print(f"num input tokens:: {usage.prompt_tokens}")
                print(f"num output tokens: {usage.completion_tokens}")
            else:
                print("Usage information not available")

    except Exception as e:
        print(f"An unexpected error occurred: {e}\n{traceback.format_exc()}")
    finally:
        pass


asyncio.run(main())

Execution Result

completion_id: chatcmpl-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
created: 123456789
model: gpt-3.5-turbo-0125
role: assistant
streaming text: Hello there! How can I be of help to you today?
Full Content: Hello there! How can I be of help to you today?
Finish Reason: stop
ttl tokens: 51
num input tokens:: 31
num output tokens: 20