explaingit

aashish2998/langchainmultiagentresearchsystem_project

★ 0PythonAudience · developerComplexity · 2/5ActiveLicenseSetup · moderate

TLDR

LangChain pipeline of four agents that searches the web, scrapes the best page, drafts a report with Groq Llama 3.3, and has a critic agent review it.

Mindmap

mindmap
  root((research-mind))
    Inputs
      Topic string
      Tavily API key
      Groq API key
    Outputs
      Structured report
      Critic feedback
      Streamlit page
    Use Cases
      Auto research on a topic
      Learn LangChain agents
      Render free-tier deploy
    Tech Stack
      Python
      LangChain
      Groq
      Streamlit
      Tavily

Things people build with this

USE CASE 1

Generate a short research report on any topic with one prompt

USE CASE 2

Study a 4-agent LangChain pipeline with search, read, write, and critic roles

USE CASE 3

Deploy a free Streamlit research tool to Render

USE CASE 4

Swap the Groq Llama 3.3 model for Gemini as a working migration example

Tech stack

PythonLangChainGroqStreamlitTavilyBeautifulSoup4Trafilatura

Getting it running

Difficulty · moderate Time to first run · 30min

Needs three API keys (Tavily, Groq, Gemini) and a Python 3.11 environment before the first run.

MIT license, free to use and modify in any project including commercial as long as the copyright notice is kept.

In plain English

ResearchMind AI is a small Python project that chains together a few language model agents into a research pipeline. You type in a topic, and the system searches the web, picks the most useful page, scrapes it, writes a structured report, and then has a second agent critique that report. The whole flow is automated, so the user only has to provide the topic and read the final output. The code is organised into four roles. A Search Agent uses an API called Tavily to find recent web results for the topic. A Reader Agent looks at those results, picks the best link, and pulls the full text of the page using BeautifulSoup4 and Trafilatura. A Writer chain feeds everything into a language model to produce a structured report, and a Critic chain reviews that report and suggests improvements. All four pieces are wired together using LangChain. The language model is Groq's hosted Llama 3.3 70B, chosen because it is fast and has a free tier. The author originally used Google's Gemini, but kept hitting daily quota limits, so they added Groq as an automatic fallback. The current code uses Groq directly; the project history table in the README walks through the migration from gemini-1.5-pro to gemini-2.0-flash to Groq Llama 3.3. To run it locally you clone the repo, create a Python 3.11 environment with conda or venv, install requirements, then set three API keys in a .env file: GEMINI_API_KEY, GROQ_API_KEY, and TAVILY_API_KEY. You can launch a Streamlit web interface with streamlit run app.py and use it in a browser, or use a command line entry point with python main.py. The README also gives step-by-step instructions to deploy the Streamlit app to Render's free tier, including the build and start commands and a list of known errors with their fixes. The project is MIT licensed, has 0 stars, and credits an existing LangChain Multi-Agent Research tutorial as the starting point.

Copy-paste prompts

Prompt 1
Walk me through running the Streamlit app locally with my own Tavily and Groq keys
Prompt 2
Explain how the Reader Agent picks the best link and combines BeautifulSoup4 with Trafilatura
Prompt 3
Replace the Critic chain with a stricter rubric that scores accuracy and citation quality
Prompt 4
Deploy this app to Render's free tier and resolve the known build errors listed in the README
Prompt 5
Swap the Llama 3.3 model for a local Ollama model and keep the agent wiring intact
Open on GitHub → Explain another repo

Generated 2026-05-22 · Model: sonnet-4-6 · Verify against the repo before relying on details.