Meeting Summarizer 🎙️📝

Objective

The primary goal of the Meeting Summarizer is to automate the process of converting meeting audio files into action-oriented summaries that highlight key decisions and actionable tasks.

---

Features & Scope of Work

This project is designed to ingest meeting audio, process it, and output structured text data.

Input: Meeting audio files. Output: Text transcript, a summary, and extracted action items. Optional Frontend: An optional user interface (frontend) is included to allow users to easily upload audio files and view the generated summary.

---

Technical Expectations

Core Technologies

The system integrates various technologies to handle transcription and summarization:

ASR API Integration: Utilizes an Automated Speech Recognition (ASR) API (e.g., Google, Azure, OpenAI Whisper) to convert audio input into a text transcript. LLM for Summary Generation: A Large Language Model (LLM) is employed to process the transcript and generate the structured summary. Prompt Guidance: The LLM is guided to summarize the transcript, highlight key decisions, and generate tasks.A prompt example is: "Summarize this meeting transcript into key decisions and action items". Backend: A backend is required to store and process the data.

Directory Structure & Development Stack

The project uses a modern web development stack, evident from the included configuration and source files:

File/Directory	Purpose
src/	Contains the main application source code.
supabase/	Likely configuration/code for a Supabase integration (used for backend/database needs).
vite.config.ts, index.html	Files indicating a Vite-based modern frontend setup.
tailwind.config.js, postcss.config.js	Configuration for Tailwind CSS and PostCSS (used for styling the optional frontend).
tsconfig.json files	Configuration files for TypeScript.
package.json files	Defines project dependencies (packages) for Node.js/JavaScript.
.gitignore, eslint.config.js	Standard files for code management and linting.

---

Deliverables

The successful completion of this project includes:

1. This GitHub Repository and comprehensive README.md file.
2. A Demo Video.

---

Evaluation Focus

The system will be evaluated based on the following key metrics:

• Transcription Accuracy
• Summary Quality
• LLM Prompt Effectiveness
• Code Structure

---

Deployment Configuration

Required Environment Variables

To deploy and run the Meeting Summarizer, you need to configure the following environment variables in your Supabase Edge Functions:

Required Secrets:

• OPENAI_API_KEY: Your OpenAI API key for accessing transcription (Whisper) and summarization (GPT-4o-mini) services
• SUPABASE_URL: Your Supabase project URL
• SUPABASE_SERVICE_ROLE_KEY: Your Supabase service role key with admin privileges

Optional Environment Variables:

• TRANSCRIPTION_MODEL: The OpenAI transcription model to use (default: whisper-1)
• TRANSCRIPTION_LANGUAGE: The language code for transcription (default: en). See OpenAI's language support for available options.

Setting Secrets in Supabase

Use the Supabase CLI to set your environment secrets:

# Install Supabase CLI if you haven't already
npm install -g supabase

# Login to Supabase
supabase login

# Link your project
supabase link --project-ref your-project-ref

# Set required secrets
supabase secrets set OPENAI_API_KEY=your_openai_api_key_here
supabase secrets set SUPABASE_URL=your_supabase_url_here
supabase secrets set SUPABASE_SERVICE_ROLE_KEY=your_service_role_key_here

# Optional: Set custom transcription model (if different from whisper-1)
supabase secrets set TRANSCRIPTION_MODEL=whisper-1

# Optional: Set transcription language (if different from English)
supabase secrets set TRANSCRIPTION_LANGUAGE=en

Deploying the Edge Function

After setting the secrets, deploy the process-meeting function:

# Deploy the function
supabase functions deploy process-meeting

# Verify deployment
supabase functions list

Testing the Deployment

You can test the deployed function using curl:

curl -X POST https://your-project-ref.supabase.co/functions/v1/process-meeting \
  -H "Authorization: Bearer YOUR_ANON_KEY" \
  -H "Content-Type: application/json" \
  -d '{"meetingId": "test-meeting-id", "audioUrl": "path/to/audio.mp3"}'

---