Run your first search locally
Install PaperSeek on your own machine, connect a model, and turn a question into a reading list.
Prefer to skip installation? Run your first search online instead.
1. Install
Use Python 3.12 or newer and an isolated environment.
python3 -m venv .venv
source .venv/bin/activate
python -m pip install "paperseek==1.0.0b1"py -3.12 -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install "paperseek==1.0.0b1"For a 1.0.0-beta1 source checkout, use python -m pip install ./backend instead. After changing frontend source, follow the contributor build instructions.
2. Start
paperseek serveOpen http://127.0.0.1:8765. Keep the terminal running while you use the local app.
The package includes the web interface; Node.js is not required to run it.
3. Connect a model
Open Settings → Model APIs at the bottom of the sidebar and choose Add connection. Select LLM, a provider and a model, then enter your key. For a custom service, check the base URL and protocol as well. You can use Test connection to check the model draft, then click Save to keep it. Under Defaults, select the saved connection for Search and scoring. Return through Search in the sidebar.
OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages are distinct protocols. A model name alone does not identify the right one.
Model draft testing does not save the connection. Data-source testing has different save behavior, described below. To supply deployment defaults instead, see configuration.
4. Ask a question
How does open-source software obtain funding for long-term maintenance?
If you have an OpenAlex key, open Settings → Data sources → OpenAlex → Configure, enter it, and click Save. Test connection in this data-source dialog also saves the current key and options before testing; a failed test or subsequently cancelling the dialog does not undo the save.
Return to Search in the sidebar and choose OpenAlex. Only restrict the research field under More options when needed. Leave the budget on Auto for a first run, or choose a budget to match the breadth of your question.
Select Start search, or press Ctrl+Enter / ⌘+Enter. Follow the stages while PaperSeek searches and ranks.
5. Read the results
Open Results, inspect abstracts and relevance explanations, and export CSV or JSON. Community saves history locally.
Reopen a run from Search history in the sidebar, or click All beside that heading for the full list. Use Citation evidence to explore available citation data, or select papers for an experimental review. Open Bibliometrics from the sidebar to analyze the full retrieved candidate corpus.
If something fails
Start with troubleshooting. For a local setup check:
paperseek doctor --jsonA valid configuration does not guarantee available credit or uninterrupted upstream service throughout a long search.
What next?
- Self-hosting: Docker deployment, data directories, network access, and backups.
- Search the literature: budgets, field filters, stopping and reconnecting.