xiry.

AI · GitHub repo

Local Deep Researcher

Fully local web research and report writing assistant using Ollama or LMStudio

From the Xiry library · Updated 11 October 2026

Visit github.com ↗

Local Deep Researcher is an open-source research assistant built with LangGraph that operates entirely on local hardware. Given a research topic, it generates search queries, browses the web, summarizes findings, self-critiques to find knowledge gaps, and loops repeatedly until it produces a comprehensive cited markdown report. It integrates with local LLM runtimes like Ollama and LMStudio alongside free search engines like DuckDuckGo to run with zero API costs.

What it's for

  • Conduct deep research on complex technical or general topics without sending data to cloud LLMs
  • Generate cited markdown research briefs and overviews automatically
  • Run recursive web searches using local reasoning models like DeepSeek R1

Features

  • Iterative research loop that reflects on summaries and identifies knowledge gaps
  • Works with local inference engines including Ollama and LMStudio
  • Default support for key-free web search via DuckDuckGo
  • Optional integration with external search APIs like SearXNG, Tavily, and Perplexity
  • Interactive workflow visualization and parameter tweaking via LangGraph Studio
  • Configurable maximum research loops and optional full-page content scraping
  • Outputs structured markdown reports with source citations

Pricing

Open source. Free and open source (MIT license)

Details

Type
GitHub repo
Category
AI
Topic
Research Agents
Open source
Yes
Platforms
macOS, Windows, Linux, Self-hosted, Command line
For
Developers and privacy-focused researchers running local AI models
Alternative to
Perplexity Pro, OpenAI Deep Research
Tags
agents, langchain, langgraph, ollama, deepseek, local ai, web search

Why it was saved

  • @RLanceMartin, 25 January 2025: Give it a research topic and it searches the web, summarizes sources, reflects on knowledge gaps to generate follow-up queries, and loops as long as wanted, ending with a report with sources — all on local models.
  • @ihteshamali, 20 March 2026: A free, 100% open-source, fully local alternative to paid deep research tools: generates its own queries, hunts sources, self-critiques for gaps and re-searches until done, then outputs a full markdown report with citations.

Related

Updated