How meetjariwala10/insightsage works
Project type and purpose
• InsightSage is a Python-based, single‐page web application (Streamlit frontend) that orchestrates an audio processing pipeline and AI services to transcribe meeting recordings, generate summaries and tasks, and send email notifications. It’s essentially a full-stack AI-powered meeting assistant using local file storage and external AI and SMTP services.
File structure
• Top-level scripts:
• new_app.py (main Streamlit app)
• audio_transcription.py (transcription logic)
• mp3_to_wav.py (FFmpeg wrapper)
• process_transcripts.py (orchestration & email)
• transcripts_vdb.py (Whisper integration)
• transcripts_ollama.py (Ollama & LangChain integration)
• speaker_recognition.ipynb (speaker diarization exploration)
• Data directories (per README):
• Audio/ (raw uploads)
• WAV_Audio/ (converted)
• TEXT_Transcripts/ (TXT output)
• JSON_Transcripts/ (JSON output)
• Summaries/ (AI summaries)
• Config/build files:
• .env(.example), requirements.txt, ffmpeg-master-…/
README for additional insights
• Technology stack: Streamlit (UI), FFmpeg (conversion), Whisper & LangChain + Ollama (AI), SMTP (email).
• Features emphasize a staged pipeline: conversion → transcription → diarization → summarization & task extraction → Q&A → email.