allenai

asta-paper-finder: Find Papers from Natural-Language Queries

asta-paper-finder is a single-query paper-finding agent that uses LLM-based analysis and relevance judgments to locate and rank research papers. This frozen evaluation snapshot is for reproducing results or running the core search workflow locally.

MaintainedPythonApache-2.0AI AgentsAPIPaper Search
View on GitHub Compare
Stars
251
Forks
32
Stars, last 30 days
+1
Last commit
12 days ago
Contributors
3

Our take

Maintaineda few commits in the last 90 days

Health score

31/100

How it is scored
Activity
29
Community
31
Issues
21
Pull requests
43

+1 stars in the last 30 days

  • Fewer contributors than 82% of the projects we track

This is a recently active but deliberately frozen evaluation snapshot, useful for reproducing or inspecting PaperFinder's single-query search, with adoption risk from its small, concentrated contributor base and lack of reported tests or CI.

Good fit if

  • You need the released PaperFinder workflow for evaluation or local API use, rather than the full online product.
  • You can provide the required API credentials and run Python 3.12.8 or later.
  • You are comfortable working with an Apache-2.0-licensed project whose recent activity is limited to a few commits and one merged pull request, with most contributions coming from one person.

Look elsewhere if

  • You need multi-turn interaction, progress updates, long-lived conversations, or a graphical results interface.
  • You need a actively maintained product integration or broad community ownership. The repository has no release in the past year and only a few contributors.
  • Your workflow requires documented automated tests or CI, neither of which is reported for this repository.
All health signals
Last commit2026-10-29 (12 days ago)
Commits, last 90 days2
Releases, last 12 months0
Contributors3 (top contributor: 75% of commits)
Issues closed, last 90 days0
Pull requests merged, last 90 days1
Project age15 months

Checked on 2026-11-11 with the GitHub API.

Project history

Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.

Selected snapshot

Stars

252

Forks

32

Change since Nov 17, 2026: 0 stars·0 forks

Overview

asta-paper-finder is AllenAI's standalone PaperFinder agent, an LLM-based tool for finding sets of research papers from a natural-language query. It aims to help researchers and evaluators locate papers by content and metadata, then rank results using relevance and criteria such as publication timing or influence.

This repository is a frozen, locally runnable snapshot of the single-turn search workflow, intended in part for reproducing evaluation results. It is not the full online PaperFinder product: it omits features such as multi-turn conversations, progress updates, and a graphical results widget.

Key Features

  • Transforms a natural-language paper request into a structured query.
  • Routes the query to a workflow suited to the paper-seeking intent.
  • Retrieves papers and uses LLM-based judgments to assess abstracts and snippets for relevance.
  • Ranks results using content relevance alongside requested criteria such as early or influential work.
  • Offers infer, fast, and diligent operation modes. The README describes infer and fast as equivalent, with diligent doing more exhaustive fetching.
  • Supports an upper-bound date for limiting results and an option to read results from a disk cache.
  • Provides a FastAPI server with a POST endpoint for search requests.

Use Cases

  • Researchers who want to find papers matching a detailed topic description without composing multiple database searches.
  • Evaluation teams reproducing the reported PaperFinder results with the released snapshot.
  • Developers who want to run and inspect the single-query paper-search API locally.
  • Users seeking papers under extra constraints, such as an upper publication-date bound or an emphasis on influential work.

What you need

Detected in the repository

  • Python >=3.12.8 (from pyproject.toml)

License in plain words

Apache-2.0permissive

  • Commercial use: yes
  • Modify and redistribute: yes
  • You must keep: the license, the NOTICE file and a note of your changes
  • Share your changes: no
  • Includes an explicit patent grant from the contributors:

A summary, not legal advice: the LICENSE file is what applies.

Getting Started

Create agents/mabool/api/conf/.env.secret with the required API keys, then run:

make sync-dev
cd agents/mabool/api && make start-dev

See the README for the required keys and API request details.

Alternatives

  • local-deep-research: It researches across web, academic, and private sources to produce cited reports, while asta-paper-finder focuses on finding and ranking papers for one query.
ProjectLanguageLicenseStarsStatus
asta-paper-finderPythonApache-2.0251Maintained
local-deep-researchPythonMIT9.2kActive

Considerations

  • The repository describes itself as a snapshot that is not actively maintained, even though recent commits indicate some maintenance activity. Treat it as a stable evaluation release, not as a promise of ongoing product development.
  • Running the agent requires keys for OpenAI, Semantic Scholar, Cohere, and Google, so operation depends on access to those services and their APIs.
  • The online agent has additional interaction features and access to datasets and indices not included in this public snapshot.
  • The project requires Python 3.12.8 or later. No Docker setup, automated tests, or CI are reported in the supplied project facts.
  • The Apache-2.0 license permits broad use, subject to its license terms.

Found this useful?

Share it with someone who would like asta-paper-finder.

Comparisons