Skip to content

ProdSense: Multi-agent product recommendation system powered by CrewAI. Scrapes product details from any product website, discovers alternatives, analyzes community sentiments, and curates YouTube reviews. Outputs tailored Markdown recommendations. Experimental, API-driven, and open for tinkering.

Notifications You must be signed in to change notification settings

Sanjith-3/ProdSense

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 

Repository files navigation

ProdSense: Multi-Agent Product Recommendation System

ProdSense is a Streamlit app powered by CrewAI that scrapes product details from any product website, hunts for alternatives, pulls community sentiments, and finds YouTube reviews. It delivers a Markdown recommendation tailored to your needs. Built for tinkerers who want to dive into AI-driven product research.

Disclaimer: This is an experimental learning project. APIs can be flaky, and data quality varies. Expect to debug and tweak.

Features

  • Scrapes product specs from any product website (optimized for e-commerce).
  • Finds up to three alternative products based on your preferences.
  • Analyzes user feedback from forums and Reddit.
  • Curates 1-2 relevant YouTube review videos.
  • Outputs a concise Markdown recommendation.

The ProdSense Crew

Agents

  • URL Identifier: Detects the product URL from inputs.
  • Input Analyst: Parses preferences and scrapes product data.
  • Market Researcher: Discovers alternative products.
  • Community Analyst: Gathers user sentiments.
  • Video Specialist: Finds YouTube review videos.
  • Recommendation Expert: Crafts the final Markdown output.

Tasks

  • Identify URL: Locates the target product URL.
  • Parse Input: Extracts user needs and product details.
  • Find Alternatives: Scouts alternative products.
  • Gather Sentiment: Collects community pros and cons.
  • Find Videos: Retrieves relevant review videos.
  • Synthesize Recommendation: Builds the Markdown recommendation.

Tools

  • FirecrawlScrapeWebsiteTool: Scrapes product pages.
  • SerperDevTool: Powers web searches for alternatives, sentiments, and videos.
  • ComposioToolSet (Reddit): Fetches Reddit feedback.

Disclaimer: Results depend on API reliability and data availability. It’s a learning experiment, not a production tool.

Example Usage

  • Product URL: https://www.example.com/product
  • Preferences: "Smartphone under 25k, gaming, long battery life."
  • Click "Get Recommendations" for a Markdown report.
  • Watch ProdSense in Action
    Screen-recorded demo showing the app’s flow—raw and unpolished.

Troubleshooting

  • API Errors: Verify SERPER_API_KEY, GOOGLE_API_KEY, FIRECRAWL_API_KEY, and COMPOSIO_API_KEY in .env.
  • Scraping Issues: Ensure the URL points to a valid product page.
  • No Results: Refine preferences or check API quotas.

Why ProdSense?

ProdSense is a playground for exploring multi-agent AI, web scraping, and recommendation systems. It’s rough around the edges but offers a solid starting point for enthusiasts who enjoy dissecting code and experimenting with data.

About

ProdSense: Multi-agent product recommendation system powered by CrewAI. Scrapes product details from any product website, discovers alternatives, analyzes community sentiments, and curates YouTube reviews. Outputs tailored Markdown recommendations. Experimental, API-driven, and open for tinkering.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages