gpt_academic

Purpose

GPT Academic (gpt_academic) is a practical interactive interface project for large language models (LLMs), especially optimized for academic scenarios, and can be used for tasks such as paper reading, polishing, writing, code interpretation, and translation.

Main Capabilities

• Provides a unified interactive interface for LLMs such as GPT and GLM;
• Optimizes the paper reading, polishing, and writing experience;
• Modular design, supports custom shortcut buttons and function plugins;
• Supports project analysis and self-translation/self-explanation functions for Python, C++, and other projects;
• Supports PDF / LaTeX paper translation and summarization, and additionally adds the Doc2x PDF paper translation function;
• Supports parallel querying of multiple LLM models;
• Supports local models such as chatglm3;
• Integrates Tongyi Qianwen, deepseekcoder, iFlytek Spark, ERNIE Bot, llama2, rwkv, claude2, moss, etc.;
• Supports coexistence of multiple api-keys, and multiple keys can be filled in the configuration file;
• Supports temporary replacement of API_KEY (after entering a temporary key in the input area and pressing Enter, it takes effect);
• Supports natural language invocation of all function plugins (Void Terminal);
• Supports conversation saving and loading of historical archives;
• Mermaid-based flowcharts and mind maps;
• Provides functions such as custom fonts, interface themes, and custom menus;
• Compatible with and encourages trying domestic Chinese large language foundation models.

Usage

Provides multiple installation and running methods:

1. Direct run (Windows, Linux, or MacOS):

git clone --depth=1 https://github.com/binary-husky/gpt_academic.git
cd gpt_academic

In config.py , configure variables such as API_KEY. The program will first check whether config_private.py exists and use it to override config.py for configurations with the same name; configuration through environment variables is supported, with read priority: environment variables > config_private.py > config.py.

Install dependencies (recommended Python version 3.9 ~ 3.11):

python -m pip install -r requirements.txt

Or use Anaconda:

conda create -n gptac_venv python=3.11
conda activate gptac_venv
python -m pip install -r requirements.txt

Or use uv:

uv venv --python=3.11
source ./.venv/bin/activate
uv pip install --verbose -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/

Run:

python main.py

1. Use Docker: Modify docker-compose.yml and run after retaining the required scheme:

docker-compose up

Schemes include: full capabilities (large image including cuda and latex), online models only (ChatGPT + GLM4 + ERNIE Bot + spark, etc., recommended for most people), and ChatGPT + GLM3 + MOSS + LLAMA2 + Tongyi Qianwen (requires familiarity with the Nvidia Docker runtime), etc.

1. Windows one-click run script: Windows users who are completely unfamiliar with the Python environment can download the one-click run script published in Releases to install the version without local models.

2. Other deployment methods: supports Sealos one-click deployment, WSL2, and secondary URL deployment (such as http://localhost/subpath ).

Applicable Scenarios

• Translation, summarization, polishing, and writing assistance for academic papers;
• Interpretation and self-translation/self-explanation of project source code;
• Scenarios requiring simultaneous comparison or invocation of multiple LLM models;
• Users who want to use local models or domestic Chinese large models;
• Developers who need custom shortcut buttons and plugin extensions.

Limitations and Notes

• When installing dependencies, please choose requirements.txt the version specified in;
• Official Gradio has many compatibility issues, so be sure to install Gradio using requirements.txt;
• Some browser translation plugins can interfere with the frontend operation;
• You can change the theme by modifying the config.py THEME option in;
• The project is divided into the master branch (main branch, stable version) and the frontier branch (development branch, beta version);
• The new GUI frontend is in testing;
• Specific fees, prices, platforms, and other information are not yet available.