LiveSuggest

LiveSuggest is a real-time, context-aware audio transcription and AI suggestion dashboard. It captures continuous audio, transcribes it, and feeds the context into language models (powered by Groq) to instantly generate conversational insights, fact-checks, and talking points. It also features a fully integrated Chat Window for manual deep dives into the transcribed context.

Visitors don't need a Groq API key of their own — every request is proxied through serverless functions that hold the key server-side, so only whoever deploys the app configures it.

Features

• Live Audio Transcription: Continuous batching and chunking of microphone input into readable transcripts.
• Dynamic AI Suggestions: Real-time generation of actionable cards (Fact Checks, Clarifying Info, Questions, Talking Points) based on the ongoing conversation.
• Context-Aware Chat: A conversational sandbox where you can query the AI about anything that was just spoken.
• No API key required for users: All Groq calls are proxied through server-side functions (/api/*) that hold the key in an environment variable — it's never shipped to the browser.
• Deep Customization (Settings Modal):
  • Pick the Groq model ID to use.
  • Tweak live suggestion generation prompts and chat system prompts.
  • Adjust hyper-parameters like Context Window size, Max Tokens, Temperature, and audio chunk interval.
• Export Functionality: One-click export of your full session (transcripts, suggestions, and chat history) to JSON for evaluation and review.

Tech Stack

• Framework: React 18 + Vite + TypeScript
• Styling: Tailwind CSS
• Backend: Vercel Edge Functions (/api) proxying to Groq, keeping the API key server-side
• AI Inference Engine: Groq API (Compatible with openai/gpt-oss-120b,Whisper LargeV3 )
• Icons & Markdown: lucide-react & react-markdown

Getting Started

Prerequisites

• Node.js installed on your machine.
• The Vercel CLI (npm i -g vercel) — used locally to run the API routes alongside the frontend.
• A free API key from Groq Console (only the deployer needs this).

Installation

1. Clone the repository:

   git clone <repository-url>
   cd LiveSuggest

2. Install dependencies:

   npm install

3. Configure your Groq API key locally:

   cp .env.example .env
   # then edit .env and set GROQ_API_KEY=gsk_...

4. Start the dev server (frontend + API routes):

   vercel dev

   (Plain npm run dev / vite only serves the frontend — the /api routes won't respond, since those are Vercel serverless functions.)

5. Open in browser: Navigate to the URL vercel dev prints (defaults to http://localhost:3000).

Deploying

When deploying on Vercel, set GROQ_API_KEY as an environment variable in the project settings (Project → Settings → Environment Variables) instead of committing it. It stays server-side and is never exposed to visitors.

Configuration & Prompts

All parameters and AI configurations are adjustable directly through the UI via the Options & Tweaks modal:

• System Prompts: You can modify the underlying system prompts that drive Live Suggestions, Standard Chat, and On-click Expanded Answers.
• Context & Rules: Choose the Groq model, set the audio chunk interval, define how many transcription chunks are sent to the AI, and control temperature/token generation directly from the interface.