mingfenghong/paperseek

★ 13PythonAudience · researcherComplexity · 3/5LicenseSetup · moderate

Mindmap

mindmap
  root((paperseek))
    What it does
      Accepts plain-language questions
      Ranks papers by relevance
      Follows citation links
      Exports results as CSV
    Data Sources
      OpenAlex
      Crossref
      Web of Science
    Tech Stack
      Python
      SQLite
      Docker
    Audience
      Researchers
      Students
      Literature reviewers

mindmap root((paperseek)) What it does Accepts plain-language questions Ranks papers by relevance Follows citation links Exports results as CSV Data Sources OpenAlex Crossref Web of Science Tech Stack Python SQLite Docker Audience Researchers Students Literature reviewers

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Type a plain-language research question and get an AI-ranked list of relevant academic papers from OpenAlex or Web of Science.

USE CASE 2

Build a citation graph to surface related research that keyword searches would miss by following citation links for top results.

USE CASE 3

Export a filtered paper list as CSV to include in a literature review or reference manager.

USE CASE 4

Connect a local Ollama model or any major AI provider to power the relevance ranking without sending data to third-party services.

Tech stack

PythonSQLiteDockerVercelOpenAI APIAnthropic APIOllama

Getting it running

Difficulty · moderate Time to first run · 30min

Requires an API key for at least one supported AI provider (OpenAI, Anthropic, Google Gemini, etc.) or a running Ollama instance.

Use freely for any purpose including commercial, as long as you keep the copyright and license notice.

In plain English

PaperSeek is a research literature discovery tool that accepts a question in plain language, then uses an AI model to search academic databases, rank the results by relevance, and export a reviewable list of papers. The README is primarily in Chinese with an English version available. The project is in alpha and is licensed under Apache 2.0. The core workflow starts with you typing a research question in Chinese or English. The tool generates appropriate search queries for the selected data source, checks how many results come back, and automatically broadens or narrows the query across up to five iterations until it reaches the target number of results you specify. Each candidate paper is then scored for relevance by the AI model with a short explanation. Optionally, the tool can follow citation links for the top-matching papers, pulling in both the papers they cite and the papers that cite them, which helps surface related work that a keyword search might miss. Data sources supported include OpenAlex, which is open and free to use, Crossref for DOI and publisher metadata, and the Web of Science Starter API for users with institutional access. Results include title, authors, journal, year, DOI, abstract, citation count, and links. The tool works through a local web interface at port 8765 or a command-line interface. The web interface has four sections: a search workspace where you configure and monitor a run in real time, a results table where you filter and select papers to export as CSV, a citation graph showing directional links between papers, and a history view of past searches stored locally in SQLite. You can connect almost any AI model. Supported providers include OpenAI, Anthropic, Google Gemini, DeepSeek, and many Chinese services including Alibaba DashScope, Kimi Moonshot, Tencent Hunyuan, Baidu Qianfan, and others. Local models through Ollama also work. Deployment options include running it directly with Python, using Docker for a complete setup, or deploying to Vercel for a lightweight hosted version.

Copy-paste prompts

Prompt 1

Using PaperSeek with OpenAlex, find the 20 most relevant papers on transformer architectures for time-series forecasting, rank them by relevance, and export as CSV.

Prompt 2

Set up PaperSeek with Docker, configure it with the Anthropic API key, and run a search for 'large language model alignment' with citation graph enabled for the top 5 results.

Prompt 3

I want to do a literature review on retrieval-augmented generation using PaperSeek. Show me how to set the target result count, enable citation following, and filter results by year before exporting.

Prompt 4

Configure PaperSeek to use a local Ollama model instead of a cloud API, then search for papers on 'diffusion models for image synthesis' and explain how the query broadening across 5 iterations works.

Open on GitHub → Explain another repo

← mingfenghong on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.