Autonomous Local Research Agent

An autonomous, fully local research agent that:

• Continuously fetches relevant arXiv papers
• Builds and incrementally updates a vector database
• Uses local LLMs (Ollama) for retrieval-augmented reasoning
• Generates research summaries, insights, and code examples
• Runs fully offline after ingestion

Features

• Incremental PDF embedding with fault tolerance
• Local RAG using Chroma + SentenceTransformers
• Ollama-based LLM inference (no cloud dependency)
• Autonomous research cycles
• arXiv ingestion with deduplication
• Resilient to bad PDFs and LLM outages

Requirements

• Python 3.10+
• Ollama
• NVIDIA GPU recommended (works on CPU)

Setup

pip install -r requirements.txt
ollama pull llama3:8b-instruct
python run_agent.py

Architecture

Example Research Output