Abstracts Explorer Documentation
Welcome to the documentation for Abstracts Explorer! This package provides tools for downloading, storing, and analyzing conference paper abstracts using LLM-based semantic search.
Web Interface
Abstracts Explorer includes a browser-based UI for searching, chatting, rating, and visualizing conference abstracts.
The interface provides four tabs — Search, AI Chat, Interesting Papers, and Clusters — described in detail in the Web Interface guide. A live demo is available at abstracts.hzdr.de.
Features
Download conference abstracts from multiple sources (NeurIPS, ICLR, ICML, and more via plugins)
Plugin system for downloading from workshops and other conferences
Store abstracts in a SQL database (SQLite or PostgreSQL) with full metadata
Create vector embeddings for semantic search
Cluster and visualize paper embeddings with multiple algorithms
MCP server for LLM-based cluster analysis and topic exploration
RAG (Retrieval-Augmented Generation) chat interface for querying papers
Web interface for browsing and searching papers
Registry support for sharing paper databases and embeddings between instances via OCI registries (e.g. ghcr.io)
Command-line interface for easy interaction
Configuration system with .env file support
Quick Start
Install the package:
# Install uv if you haven't already
curl -LsSf https://astral.sh/uv/install.sh | sh
# Install the package with all dependencies
uv sync --all-extras
Download conference abstracts:
# Download abstracts for a specific conference and year
uv run abstracts-explorer download --year 2025
# Or download from a workshop using plugins
uv run abstracts-explorer download --plugin ml4ps --year 2025
Alternatively, download pre-built data from an OCI registry (no LLM backend required):
uv run abstracts-explorer registry download \
-r ghcr.io/thawn/abstracts-data \
--conference neurips --year 2025
Create embeddings for semantic search:
uv run abstracts-explorer create-embeddings
Search papers:
uv run abstracts-explorer search "machine learning"
Chat with papers using RAG:
uv run abstracts-explorer chat
Documentation Contents
User Guide