Year
2025
Category
Hackathon Project
Product Duration
1 - 2 Weeks
SocialPulse is a smart analytics tool that helps users understand and interpret datasets in a conversational, user-friendly manner. It leverages Langflow, Astra DB, and LLM APIs (Groq or OpenAI) to process uploaded CSV files and answer user queries based on the dataset. Instead of requiring technical skills to analyze data, users can simply ask questions in natural language and receive clear summaries powered by AI. The tool is especially useful for non-technical users looking to extract insights from data without complex visualization or manual analysis tools.
The backend is designed as a Langflow flow, called SocialPulse_flow, comprising several components:
Load File: Users upload a
.csvfile.Split Text into Chunks: Breaks the data into manageable parts.
Astra DB + NVIDIA Vector Search: Chunks are stored in Astra DB and indexed for fast, vector-based search.
User Query Input: The user inputs a natural language question via chat.
Search & Parse: Relevant chunks are retrieved, parsed, and prepared for response generation.
Custom Prompt + LLM (Groq/OpenAI): A prompt template guides the AI to answer based on retrieved results, focusing on clarity and simplicity.
Although initially built using the Groq API due to API constraints, it can be easily adapted to use OpenAI GPT models, which would likely improve output quality and depth.
A React-based website, developed by team member @dikjain, acts as the frontend for SocialPulse, allowing users to interact through a simple chat UI. The website connects to the Langflow backend via API, sends user queries, and displays results dynamically.
Deployment Requirements:
A Langflow environment with the flow JSON loaded.
An Astra DB instance with NVIDIA vector search enabled.
Groq or OpenAI API keys configured.
The goal is to make data querying as intuitive as chatting with an assistant.
