ConvoDigest - an app designed to summarize and query unstructured group chats.
Purpose: Helps users in large communities stay on top of long, unstructured group chats by generating summaries and answering queries based on chat content.
Key Technologies:
AI/LLM Integration: Utilizes large language models (LLMs), retrieval-augmented generation (RAG), and embeddings.
Vector Database: Employs Vectra for storing chat embeddings.
LLaMA 3.1 (7B parameters) via Ollama for AI-driven analysis.
Key Problem:
Managing and retrieving important information from overwhelming group chat volumes (e.g., WhatsApp chats).
Solution:
Summarizes chat content and allows users to query chats for key points and specific questions.
Local-first approach: Entire processing is done locally on the user’s device to maintain privacy and security.
How It Works:
Users upload their WhatsApp chat data.
App processes the chats using Node.js, Express, and React, generating searchable summaries.
Privacy-Focused: No offloading to external servers; all data is processed locally to maintain full user privacy.
Built with