paper summarizer

Paper Summarizer

A Slack Bot for summarizing arXiv papers, powered by OpenAI LLMs.

Features

This repository offers 2 main features:

• Summarize an arXiv paper which you mention in a Slack channel.
• Once a day, post summaries of papers which AK mentions.

Currently, the bot supports only papers posted on arXiv.

How to use

0. Get API keys for OpenAI and Slack

• OpenAI: https://platform.openai.com/account/api-keys
• Slack: https://api.slack.com/apps

1. Add the bot to your Slack workspace and make it be able to react with user's mention

• This article may be helpful: Enabling interactions with bots

2. Clone this repository

git clone https://github.com/discus0434/paper-summarizer.git

3. Build a Docker image

If you do not use CUDA (i.e., you use only CPU), you can modify Dockerfile as follows:

- FROM paddlepaddle/paddle:2.4.1-gpu-cuda11.2-cudnn8.2-trt8.0
+ FROM paddlepaddle/paddle:2.4.1

Then, build a Docker image:

cd paper-summarizer/docker && make build

4. Set environment variables

OPENAI_ORGANIZATION might be optional.

echo "OPENAI_ORGANIZATION=org-XXX" >> .env
echo "OPENAI_API_KEY=sk-XXX" >> .env
echo "SLACK_INCOMING_WEBHOOK_URL=https://hooks.slack.com/services/XXX/XXX/XXX" >> .env
echo "SLACK_SIGNING_SECRET=XXX" >> .env
echo "SLACK_BOT_TOKEN=xoxb-XXX" >> .env

5. Run the API and forward the port

You can run the API with the following command:

make run

If your PC is Windows, run below in git bash:

make run-win
docker exec -it <container id> bash
python3 app.py

Then, forward the port by using ngrok or something like that.

Requirements

• Computer with x86-64 architecture
• Docker

License

This repository is licensed under AGPLv3. See LICENSE for more information.

References

• For PDF to text conversion, using PaddlePaddle model.
• The docker image is based on paddlepaddle/paddle.