arx-agent is an AI research assistant that helps developers and researchers quickly understand and work with arXiv papers. It ingests PDFs, splits them into pages or logical chunks, creates embeddings, and serves retrieval-augmented results to downstream components.
PDFs are sourced from arXiv (https://arxiv.org) by default — the project expects arXiv research paper PDFs but can be adapted to other sources if needed.
- PDF ingestion and page-level splitting so each page or chunk can be embedded independently.
- Storage of embeddings and metadata in a vector store (collection-per-project/document pattern supported).
- Retrieval logic that returns relevant chunks so downstream components can attach those to prompts.
- Clone the repository
git clone git@github.com:bijaydas/arx-agent.git
cd arx-agent- Install project dependencies
Primary command (common uv install command):
uv sync- Add OpenAI API key
- Put your OpenAI key where your terminal can see it.
export OPENAI_API_KEY="sk-..."- Or put the same line into a local
.envfile and make sure you never commit that file.
Execute a search against a topic
uv run src/arx_agent/cli.py "Novel Systems Thinking approach"CLI screenshot:
