Skip to content

Latest commit

 

History

History
86 lines (58 loc) · 3.42 KB

README.md

File metadata and controls

86 lines (58 loc) · 3.42 KB
Raw

Processing Prompts by Batch (probat)

This project provides a Python script to interact with multiple Large Language Models (LLMs) from various APIs like Google Gemini, Deepseek, OpenAI (Harvard), Anthropic, GPT-4Free, and Qwen. The script allows batch processing of prompts, splitting lengthy inputs into manageable chunks, and retrieving responses from the specified LLM.

NOTICE

We are currently using the Deep Seek V2 API as the default.

Branch

Jiajun Zou's G4F implement

https://github.com/jzou19957/Unlimited-Excel-Processing-through-GPT-3.5-API

Features

  • Multi-API Support: Easily switch between different language models including Google Gemini, Deepseek, OpenAI (Harvard), Anthropic, GPT-4Free, and Qwen by setting api_choice.
  • Chunk Splitting: Automatically splits long prompts into smaller chunks based on specified separators to ensure they fit within token limits.
  • Text Cleaning: Converts newline characters to <br /> for formatted output.
  • Configurable Parameters: Adjust batch size, timeout, and token length thresholds for efficient processing.
  • Batch Processing: Processes multiple prompts at a time, writing outputs to an external file.

Requirements

  • Python 3.9 or newer
  • Dependencies for specific APIs:
    • google.generativeai for Google Gemini
    • openai for Deepseek and Qwen
    • anthropic for Anthropic API
    • requests for OpenAI (Harvard)
    • g4f for GPT-4Free

Installation

  1. Before running the script, ensure you have Python installed on your system, and then install the required SDK using pip:

For Deep Seek and Qwen users:

pip install -U openai

For Gemini users:

pip install -q -U google-generativeai

For anthropic users:

pip install anthropic

For GPT-4Free users:

pip install g4f
  1. You can obtain a API key from the LLM API website(For exampel, deep seek users can visit https://platform.deepseek.com). After acquiring your API key, save the key to api_key.txt in the root directory of the current repository.

image

Usage

  1. Place your text prompts in a file named prompts.txt, with one prompt per line.
  2. (Optional) Add a prefix prompt to the prompt_prefix.txt file to include a prefix for each prompt if needed.
  3. Change api_choice = "deepseek" in probat.py to the LLM you want to use and save the file. The available options are: gemini, deepseek, openai_harvard, anthropic, call_g4f, and qwen.
  4. Run probat.py.
  5. The script will process all prompts and save the outputs in output.txt. If this file already exists, it will be overwritten.

Configuration

You can adjust the following configurations at the beginning of the script:

  • TEMP_BATCH_SIZE: Number of prompts processed in each batch (default is 10).
  • TIMEOUT: The base delay time between API calls (default is 0.5 seconds).
  • TIMEOUT_OFFSET: The offset for randomizing delay time (default is 0.5).
  • LEN_THRESHOLD: Character length threshold for splitting prompts (default is 2000).
  • SEPARATOR_LIST: List of separators to use when splitting long prompts.

Disclaimer

Always adhere to the API provider's usage policies and guidelines.

License

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International license