Hundreds of abstracts. Days to conference. Which talks matter to you?
Abstracts Explorer is an agentic AI tool that helps researchers navigate conference literature — download abstracts, search semantically, chat with an AI that cites its sources, and visualise the topic landscape.
Open source · Apache-2.0 · runs locally or on your institute's server.
| Layer | What it does |
|---|---|
| Data | Downloads & stores abstracts from 8 conferences (NeurIPS · ICLR · ICML · CHI · HAICON · …) |
| Embeddings | Converts abstracts to vectors via an LLM embedding model (Blablador / LM Studio) |
| Retrieval | Semantic (cosine similarity) search |
| Reasoning | RAG chat: the LLM reads relevant abstracts, then answers with citations |
| Agency | 6 MCP tools the LLM calls autonomously to analyse topics, trends, and clusters |
| Interface | Flask web UI — four tabs, no coding required |
| Tab | What you can do |
|---|---|
| 🔍 Search | Keyword and semantic search with filters (year, track, decision type) |
| 💬 Chat | Ask anything — query rewriting, abstract retrieval, cited answers, MCP tools |
| ⭐ Interesting Papers | Rate 1–5 stars, export as ZIP / JSON / Markdown |
| 📊 Statistics | Topic trend chart and interactive cluster landscape |
The LLM picks the right tool automatically from your plain-English question.
| Tool | Triggered when you ask… |
|---|---|
get_conference_topics |
"What are the main themes at NeurIPS 2025?" |
analyze_topic_relevance |
"How prominent is uncertainty quantification?" |
get_topic_evolution |
"How has diffusion model research trended since 2020?" |
search_papers |
"Find papers about graph neural networks" |
get_cluster_visualization |
"Show me the topic landscape" |
get_paper_details |
"Who wrote that paper? Where is the poster?" |
curl -fsSL https://raw.githubusercontent.com/thawn/abstracts-explorer/main/scripts/install-quadlets.sh | bashuv sync --all-extras
uv run abstracts-explorer download --conference haicon --year 2026
uv run abstracts-explorer create-embeddings
uv run abstracts-explorer web-ui # → http://localhost:5000echo "LLM_BACKEND_AUTH_TOKEN=<your_blablador_token>" > .env
curl -o docker-compose.yml https://raw.githubusercontent.com/thawn/abstracts-explorer/refs/heads/main/docker-compose.yml
docker compose up -d # → http://localhost:5000- 🔍 Using the Web UI
- 🐳 Docker / Podman Guide
- 📖 Installation Guide
- ⚙️ Configuration
- 🔌 Plugins Guide
- 🤖 MCP Server
- 📚 Full Docs
Contributions welcome — see Contributing Guide.
Apache License 2.0 — see LICENSE.