KodeKloud Notes

Learn how to send multiple prompts in one go using OpenAI’s Python client. This guide covers setup, code structure, and best practices for efficient batch processing.

Prerequisites
Install & Import
Define Your Prompts
Helper Function for a Single Prompt
Batch Processing Loop
Inspecting Results
Run the Script
Reference Links

Prerequisites

Python 3.7+
An OpenAI API key
openai Python package

Note

Store your API key as an environment variable for security:

export OPENAI_API_KEY="sk-your-api-key-here"

Alternatively, pass it directly in code (not recommended for production).

Install & Import

Install the OpenAI Python client:

pip install openai

Then import and initialize the client:

from openai import OpenAI

client = OpenAI(api_key="sk-your-api-key-here")

Define Your Prompts

Build a list of user prompts to process in batch:

prompts = [
    "Tell me a story about a warrior princess",
    "Generate a list of 5 business ideas",
    "Explain the theory of relativity in simpler terms",
    "Write a poem about Michael Jordan"
]

Feel free to extend this list to dozens or hundreds of items.

Helper Function for a Single Prompt

Encapsulate the API call in a reusable function:

def process_prompt(prompt: str) -> str:
    response = client.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "user", "content": prompt}],
        max_tokens=250,
        temperature=0.8
    )
    return response.choices[0].message.content

Model Parameters

Parameter	Description	Example
model	The chat model to use	`gpt-4`
messages	Conversation history for the model	`[...]`
max_tokens	Maximum number of tokens in the reply	`250`
temperature	Controls randomness (0.0–1.0)	`0.8`

Batch Processing Loop

Iterate through all prompts and collect responses:

results = []

for prompt in prompts:
    reply = process_prompt(prompt)
    results.append(reply)

Warning

Batch requests can incur higher costs and rate limits. Monitor your usage in the OpenAI Dashboard.

Inspecting Results

Print each prompt alongside its generated response:

for idx, (prompt, reply) in enumerate(zip(prompts, results), start=1):
    print(f"Prompt {idx}: {prompt}")
    print(f"Response {idx}:\n{reply}\n")

Sample output:

Prompt 1: Tell me a story about a warrior princess
Response 1:
Once upon a time in the verdant kingdom of Eldoria, a fierce warrior princess...

Prompt 2: Generate a list of 5 business ideas
Response 2:
1. Eco-friendly packaging startup
2. Virtual event planning service
...

Prompt 3: Explain the theory of relativity in simpler terms
Response 3:
The theory of relativity, proposed by Albert Einstein, shows how time and space are linked...

Prompt 4: Write a poem about Michael Jordan
Response 4:
In courts of hardwood, he stood so tall...

Run the Script

Save the code to batch_processing.py and execute:

python batch_processing.py

Extend or customize the process_prompt function to add streaming output, error handling, or alternative model parameters as needed.

Batch Processing

Table of Contents