speed-the-arxiv 0.1.0

A web app for searching and managing arXiv papers with SciRate integration

speed-the-arxiv

You know the drill. Wake up, coffee, arXiv. Same sections, same keywords, same authors — every single day. Speed-the-arxiv exists so you can do that in seconds instead of minutes, and maybe enjoy the coffee a bit more.

Define your searches in .yaml configs (or create them right from the UI), hit a button, and get a clean page of results with collapsible abstracts, LaTeX rendering, SciRate scores, and one-click BibTeX — no copy-pasting DOIs into twelve different tabs.

Try it with limited functionality: https://mekise.pythonanywhere.com/

Setup

Clone the repository:

git clone https://github.com/mekise/speed-the-arxiv.git
cd speed-the-arxiv

Install dependencies using setup.py:

pip install .

Or install them manually:

pip install flask requests feedparser pyyaml habanero waitress beautifulsoup4 aiohttp

Usage

1. Edit search/config.yaml to set your sections, keywords, and authors. You can also create, edit, duplicate, and delete configs directly from the UI.
2. Run the app:

   python speedthearxiv.py
   

3. A browser window opens at http://localhost:8080/. Select a search from the landing page — cache age is displayed next to each search so you know how fresh the results are.
4. Results are loaded via AJAX with an inline spinner on the clicked button. Use the highlight buttons, sort controls, text filter, and pagination to navigate results.

Search configuration

Each .yaml file in search/ supports the following parameters:

Parameter	Description
max_results	Maximum number of results to fetch
past_days	Number of days to look back from today
literal	Literal (exact phrase) keyword search
run_scirate	Fetch SciRate scores for each paper
arxiv_sortby	arXiv sort field (submittedDate, relevance, lastUpdatedDate)
arxiv_sortorder	arXiv sort order (descending, ascending)
sortby	Local sort keys (e.g. ['date', 'scirate'])
sortorder_rev	true for descending, false for ascending
and_or_*	Logical connectors between keys (+OR+, +AND+, +ANDNOT+)
keys.sections	arXiv category list (e.g. [quant-ph, cs.LG])
keys.keyauthors	Author list
keys.keywords	Keyword list

See search/config.yaml for a complete example.

Features

• Configurable searches — sections, keywords, authors, logical operators, date ranges, all in tidy .yaml files you can create, edit, duplicate, and delete from the browser.
• Dark & light theme — because not everyone wants to stare into the void (or into the sun).
• Sort, filter, paginate — sort by date, SciRate score, or title. Filter results in real time. 25 papers per page so your browser doesn't melt.
• SciRate scores — fetched asynchronously with concurrency limits and timeouts, so you get scores without hammering the internet.
• BibTeX on demand — auto-generated for every paper, copy to clipboard with one click. Need a specific DOI? The Crossref lookup on the landing page has you covered.
• Keyword & author highlighting — toggle-able, so the important bits jump out at you.
• MathJax — LaTeX in titles and abstracts rendered properly, as nature intended.
• Caching — results are cached locally with a visible age indicator so you know exactly how stale your data is.
• Robust error handling — timeouts, API failures, and unreachable endpoints all get friendly messages instead of blank screens.

Why no ML?

A recommendation engine narrows your reading to papers that look like what you already read. That's great for efficiency, terrible for serendipity. Speed-the-arxiv casts a wider net on purpose — you set the parameters, the search distills the field, and every now and then something unexpected catches your eye. That's the good stuff.

Screenshot