/Ice-breaker-powered-by-LLM

Ice Breaker is comprehensive fullstack app leveraging generative AI and LangChain to find LinkedIn profiles and generate engaging ice breakers. LangChain ReAct agents ensure accurate URL retrieval and JSON cleaning, identifying a summary, facts, topics, and ice breakers. The frontend is built with HTML/CSS, and Flask powers the backend development.

Primary LanguagePython

Ice Breaker

NOTE: Currently under progress Estimated timeline: To be finished this weekend (19/7/2024 - 20/7/2024)

Overview

Ice Breaker is a comprehensive fullstack application powered by generative AI. It crawls the web to find a person's LinkedIn profile and utilizes the LangChain library to craft engaging ice breakers based on their history. LangChain ReAct (Reasoning Action) agents are employed to ensure the accurate return of the LinkedIn profile URL and to clean tokens in the resulting JSON data. The output from the LLM is parsed using LangChain's output parser to identify four key components from the JSON file: a summary, two intriguing facts, topics of interest, and two crafted ice breakers. The frontend is created with static HTML/CSS, while Flask is used for the backend development.

Topics covered and Learned

Python | Langchain | LLMs | Langsmith | Geneartive-AI | Prompt Enginerring | Chain of thoughts | Agents | ReAct Agents | Tools | Web scrapping and Crawling | Flask | HTML/CSS | Output Parser | Proxycurl | Tavily Web API |

Components

  • Flask Backend: Serves as the backbone of the application, handling HTTP requests, API interactions, and serving dynamic content using Jinja2 templating.

  • LangChain with OpenAI LLM: Utilizes advanced language models for generating personalized and contextually relevant IceBreaker suggestions.

  • Social Media Integration: Leverages APIs from Google Search, Twitter, and LinkedIn to gather insights and tailor conversations effectively.

  • Frontend: Built with HTML and CSS, providing a user-friendly interface for interaction.

Future Scope

  • Resolve the Frontend delivery issue
  • I have a front-end application built with HTML and CSS that I want to migrate to Streamlit.
  • Finally Deploy it on huggingFace and make it public
  • Problem: I've ran out of proxycurl credits and I have no money to buy extra credits LOL!, that's one reson why I'm nor looking forward to deploy the app

Environment Variables

To run this project, you will need to add the following environment variables to your .env file

PYTHONPATH={YOUR_PATH_TO_PROJECT}/ice_breaker OPENAI_API_KEY PROXYCURL_API_KEY TAVILY_API_KEY