Research Rabbit: Web research and report writing using native LLM, automatically drilling down into user-specified topics and generating summaries.

General Introduction

Research Rabbit is a native LLM (Large Language Model) based web research and summarization assistant. After the user provides a research topic, Research Rabbit generates a search query, obtains relevant web results, and summarizes those results. It iterates this process, filling in knowledge gaps, and ultimately generates a Markdown summary that includes all sources. The tool runs entirely locally, ensuring data privacy and security.

Function List

• Generate Search Queries: Generate search queries based on user-supplied topics.
• Web Search: Use a configured search engine (e.g. Tavily) to find relevant resources.
• Summarizing the results: Summarize web search results using local LLM.
• reflective summary: Identify knowledge gaps and generate new search queries.
• iterative updating: Repeatedly conduct searches and summaries to gradually refine the research.
• local operation: All operations are performed locally to ensure data privacy.
• Markdown Summary: Generate a final Markdown summary that includes all sources.

Using Help

Installation process

1. Pull local LLM: Pull the required local LLM from Ollama, for exampleollama pull llama3.2The
2. Get Tavily API key: Register Tavily and get API key, set environment variablesexport TAVILY_API_KEY=<your_tavily_api_key>The
3. clone warehouse: Rungit clone https://github.com/langchain-ai/research-rabbit.gitCloning Warehouse.
4. Installation of dependencies: Go to the project directory and runuvx --refresh --from "langgraph-cli[inmem]" --with-editable . --python 3.11 langgraph devInstall the dependencies.
5. Startup Assistant: Start the LangGraph server and access thehttp://127.0.0.1:2024View the API documentation and Web UI.

Usage Process

1. Configuring LLM: Set the name of the local LLM to be used in the LangGraph Studio Web UI (default)llama3.2).
2. Setting the depth of research: Configure the depth of the study iterations (default 3).
3. Enter a research topic: Start the Research Assistant by entering the research topic in the Configuration tab.
4. View Process: The assistant generates search queries, performs web searches, summarizes the results using LLM, and iterates the process.
5. Getting the summary: Once the research is complete, the assistant generates a Markdown summary with all the sources, which the user can view and edit.

Main Functions

• Generate Search Queries: After entering a research topic, the assistant will automatically generate a search query.
• Web Search: The assistant uses the configured search engine to find relevant resources.
• Summarizing the results: The assistant uses the local LLM to summarize the search results and generate a preliminary report.
• reflective summary: The assistant identifies knowledge gaps in the summary and generates new search queries to continue the research.
• iterative updating: The assistant will iteratively conduct searches and summaries to gradually refine the research.
• Generating Markdown Summaries: Once the research is complete, the assistant generates a final Markdown summary with all the sources, which the user can view and edit.