simple-openai-requests 1.1.0

Unify common OpenAI API requests use cases into a simple interface

Simple OpenAI Request

This script provides a simple interface for making OpenAI API requests with various use cases:

• Synchronous requests: parallel requests with retry when rate limit hit
• Batch requests: create, check and merge result of batch API requests
• Caching: exact match caching to avoid redundant API calls

Main Function

make_openai_requests(conversations, model, use_batch=False, use_cache=True, ...)

Key Parameters:

• conversations: List of conversations/messages (supports multiple formats)
• model: OpenAI model to use (e.g., "gpt-3.5-turbo")
• use_batch: Set to True for batch API, False for synchronous API
• use_cache: Enable/disable caching
• api_key: OpenAI API key (if not provided, it will be read from the environment variable OPENAI_API_KEY)

Additional Options:

• generation_args: Additional arguments for the API call (e.g., max_tokens, temperature)
• cache_file: Path to the cache file (default: environment variable SIMPLE_OPENAI_REQUESTS_CACHE_FILE or default as ~/.gpt_cache.db)
• batch_dir: Directory for batch processing files (default: environment variable SIMPLE_OPENAI_REQUESTS_BATCH_DIR or default as ~/.gpt_batch_requests)
• full_response: Return full API response or just the message content
• user_confirm: If True, prompts for user confirmation before making API requests
• base_url: use when calling vLLM APIs.

and other parameters in function make_openai_requests()'s documentation.

Return Format:

The function returns a list of dictionaries, where each dictionary contains:

• index: The index of the conversation
• conversation: The original conversation
• response: The API response (full response object if full_response=True, otherwise just the message content)
• is_cached_response: Boolean indicating if the response was from cache
• error: Any error message (None if no error occurred)

Installation

Option 1: Install using pip

You can install the Simple OpenAI Request package using pip:

pip install simple-openai-requests

Option 2: Install from source

To install the package from source, follow these steps:

1. Clone the repository:

   git clone https://github.com/lehoanganh298/simple_openai_requests.git
   

2. Navigate to the project directory:

   cd simple_openai_requests
   

3. Install the package:

   pip install .
   

Usage Examples:

1. Simple string prompts

from simple_openai_requests import make_openai_requests

conversations = [
    "What is the capital of France?",
    "How does photosynthesis work?"
]

results = make_openai_requests(
    conversations=conversations,
    model="gpt-3.5-turbo",
    use_batch=False,
    use_cache=True
)

for result in results:
    print(f"Question: {result['conversation'][0]['content']}")
    print(f"Answer: {result['response']}\n")

Output

INFO - Cache disabled
Making 2 synchronous API requests using model 'gpt-3.5-turbo'? ([y]/n): 
API Requests: 100%|████████████████████████████████| 2/2 [00:04<00:00,  2.09s/request]
INFO - Completed 2 synchronous API requests
INFO - Successful requests: 2/2
INFO - Failed requests: 0/2
[
  {
    "index": 0,
    "conversation": [
      {
        "role": "user",
        "content": "What is the capital of France?"
      }
    ],
    "response": "The capital of France is Paris.",
    "error": null,
    "is_cached_response": false
  },
  ...
]

2. Conversation format

conversations = [
    [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "What is the best way to learn programming?"}
    ],
    [
        {"role": "system", "content": "You are a knowledgeable historian."},
        {"role": "user", "content": "Explain the significance of the Industrial Revolution."}
    ]
]
results = make_openai_requests(
    conversations=conversations,
    model="gpt-4",
    use_batch=True,
    use_cache=False,
    generation_args={"max_tokens": 150}
)
for result in results:
    print(f"Question: {result['conversation'][-1]['content']}")
    print(f"Answer: {result['response']}\n")

Output

Making 2 batch API requests using model 'gpt-4o-mini' with batch run name 'ac3275cd-8b2e-4d8d-97a0-092916b6b9ce'? ([y]/n): 
INFO - Batch request file created: /Users/user/.gpt_batch_requests/ac3275cd-8b2e-4d8d-97a0-092916b6b9ce_batch_request.jsonl
INFO - Batch submitted with file ID: file-wEdX4pNawx7XmUocTVcwzAF9
INFO - Batch processing started. Batch ID: batch_TgV5j1LPf3nYWwKgBCfF4qZ7

3. Indexed conversation format

conversations = [
    {
        "index": 0,
        "conversation": [
            {"role": "system", "content": "You are a math tutor."},
            {"role": "user", "content": "Explain the Pythagorean theorem."}
        ]
    },
    {
        "index": 1,
        "conversation": [
            {"role": "system", "content": "You are a creative writing assistant."},
            {"role": "user", "content": "Give me a writing prompt for a short story."}
        ]
    }
]
results = make_openai_requests(
    conversations=conversations,
    model="gpt-3.5-turbo",
    use_batch=False,
    use_cache=True,
    max_workers=2
)
for result in results:
    print(f"Index: {result['index']}")
    print(f"Question: {result['conversation'][-1]['content']}")
    print(f"Answer: {result['response']}\n")