v0.1.0 — ConfLens

First public release of ConfLens — a NiceGUI web app that browses conference papers, classifies them against a theme with an LLM, discovers topics, and synthesises per-topic findings.

Highlights

• Sources: ACL Anthology (e.g. acl-2026) and IJCAI accepted-paper pages; pluggable adapter interface.
• Theme classification: batched, structured-output relevance scoring with confidence + rationale.
• Topic modelling: LLM taxonomy + assignment (optional BERTopic backend).
• Per-topic synthesis: description + 5–10 common findings across each topic's papers.
• LLM providers: Anthropic (default), OpenAI, and LiteLLM (custom endpoint).
• Exports: PPTX slide deck, JSON, CSV; per-topic paper lists with PDF links.
• Caching: scrapes and LLM results cached on disk; --clear-cache command.
• Runs anywhere: Python 3.13 + uv, or Docker / docker-compose.

Install

uv sync            # + --extra all for OpenAI/LiteLLM
cp .env.example .env
uv run conference-analyzer

Licensed under MPL-2.0.