Arxie v0.1.0 — Pre-release

First pre-release of Arxie — an AI research assistant that reads real papers, not its training data.

Features:

• Live retrieval from Semantic Scholar + arXiv
• Full-text PDF analysis (methods, results, tables — not just abstracts)
• Multi-hop deep search across papers (ra query --deep)
• Literature review generation with thematic grouping (ra lit-review)
• Citation graph exploration (ra trace)
• Confidence scoring per claim (supporting vs contradicting evidence)
• Interactive conversational research sessions (ra chat)
• Verified inline citations with DOI/arXiv links
• FastAPI REST API + Docker support
• 130 unit tests, 100-question eval suite

Eval baseline (GPT-4o-mini):

• Citation precision: 86% (target ≥85%)
• Claim support ratio: 100% (target ≥80%)
• Tool success rate: 99.8% (target ≥90%)

Install:

git clone https://github.com/mmTheBest/arxie.git
cd arxie && python -m venv .venv && source .venv/bin/activate && pip install -e .