PaperTrail / README.md
AdarshRL's picture
Publish PaperTrail from pinned public source
f7f8a1e verified
|
Raw History Blame Contribute Delete
1.89 kB
---
title: PaperTrail
emoji: 📄
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.30.0
python_version: 3.12
app_file: app.py
license: apache-2.0
datasets:
- allenai/qasper
models:
- google/embeddinggemma-2
- BAAI/bge-reranker-v2-m3
- Qwen/Qwen3-1.7B
---
# PaperTrail
PaperTrail helps you inspect research papers, find passages, and review answers alongside their sources. The application code is available in the [PaperTrail GitHub repository](https://github.com/Adarsh096/PaperTrail). `SOURCE.json` records the exact source revision used by this Space.
Choose a saved example to browse the sample corpus without making model requests. Live answers require both screening and answer providers to be available. When either is not configured, saved examples and browsing remain available, and live answers are disabled. For a deployment that enables live answers, configure `GROQ_API_KEY` and `OPENROUTER_API_KEY`. Answers can be incomplete or wrong, so check the cited passages before relying on them.
You can also upload a PDF of up to 5 MB and 20 pages. Uploads are processed for the current session and do not enter the shared paper corpus. Do not upload personal, confidential, or unpublished material.
Live questions and relevant paper excerpts may be sent to external AI providers for screening, reranking, or answer generation. Provider limits or outages can prevent a live answer. Usage limits apply over a rolling 24-hour window, with up to 4 questions and 2 PDF-processing attempts per visitor. People sharing a network connection may share these limits. Saved examples and browsing do not use these allowances.
Sample-paper and dataset acknowledgements and license details are in [DATA_LICENSES.md](DATA_LICENSES.md). Model files are downloaded from their upstream repositories when the matching feature is enabled; they are not bundled with this project.