文件列表:

archive/
assets/
current/
py/
src/
tags/
templates/
test/
translated-papers/
LICENSE
configs.yaml

[](https://github.com/deep-diver/hf-daily-paper-newsletter/actions/workflows/newsletter.yml) # Daily Newsletter for Daily Papers [](https://groups.google.com/g/hf-daily-paper-newsletter) This repository sends out a newsletter at `0 20 * * *` collected from [ Daily Papers](https://huggingface.co/papers) to the subscribers. If you want to subscribe, please join the [newsletter group](https://groups.google.com/u/2/g/hf-daily-paper-newsletter ). Everything is automated via GitHub Action, so no paid maintanance at all. # How this works? This whole project works in 2 steps broadly in the perspective of GitHub Action workflow. 1. [collect papers](https://github.com/deep-diver/hf-daily-paper-newsletter/blob/main/.github/workflows/collect_papers.yml) 1. Download the metadat of daily papers from [ Daily Papers](https://huggingface.co/papers) API Endpoint. The target date is the yesterday since it is unclear when the full list of daily paper is fully updated. The target date is dynamically calculated via `date` command. 2. Run `go run main.go parse` command to **(1)** parse the downloaded the metadata file (JSON), **(2)** generate categories/tags of each paper via [Gemini API](https://ai.google.dev/), and **(3)** generate reformatted `yaml` file for each paper which is suitable for newsletter program. Each `yaml` file will be stored inside the `current` directory. 2. [send newsletter](https://github.com/deep-diver/hf-daily-paper-newsletter/blob/main/.github/workflows/newsletter.yml) 1. Grasp all the `yaml` files from the `current` directory. 2. Fill in the email HTML [template](https://github.com/deep-diver/hf-daily-paper-newsletter/tree/main/templates) with the contents from the `yaml` files. 3. Archiving each papers' `yaml` file under the `archive` directory 4. Generate markdown files based on the existing categories/tags. Add link of the corresponding archived paper `yaml` file to each markdown file. 5. Send out email to the receivers which are specified in the [config.yaml](https://github.com/deep-diver/hf-daily-paper-newsletter/blob/main/configs.yaml) file. 6. Clean up the `current` directory. Experimentally, summary of each paper is generated by [Upstage](https://www.upstage.ai/) 's [SOLAR-10.7B-Instruct](https://huggingface.co/upstage/SOLAR-10.7B-Instruct-v1.0) model. Since abstract is too long to be contained in an email, abstracts needed to be shrinken. # Appearance  # Todos - [X] Auto Translation to other languages (first target -> Korean)

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：