---
title: Ask The Right Question
emoji: 🎤
colorFrom: purple
colorTo: indigo
sdk: gradio
sdk_version: "5.50.0"
app_file: app.py
pinned: false
tags:
  - mcp-in-action-track-consumer
  - agents
  - question-generation
  - q-and-a
  - critical-thinking
---

# Ask The Right Question

> AI-powered assistant that helps you ask insightful questions during Q&A sessions at conferences, classes, and talks.

## The Problem

You're at a conference, a class, or a talk. The speaker finishes and opens the floor for questions. Your mind goes blank. You want to:
- Ask something meaningful that shows you were engaged
- Make a connection with the speaker
- Actually learn something valuable from the Q&A

But coming up with good questions on the spot is hard.

## The Solution

**Ask The Right Question** listens to the talk and helps you generate thoughtful, well-researched questions. It:

1. **Captures Context** - Record audio snippets or paste notes from the talk
2. **Researches** - Automatically researches the speaker's background, fact-checks claims, and finds relevant trends
3. **Generates Questions** - Creates insightful questions with explanations of *why* they're good
4. **Teaches Critical Thinking** - Helps you understand what makes a question valuable

## Features

### Core Features (MVP)
- Live microphone capture with OpenAI Whisper transcription
- Context accumulation as the talk progresses
- AI-powered question generation with Claude
- Web research via Tavily (speaker background, fact-checking, trends)
- Question categorization (Clarification, Depth, Connection, Challenge, Practical, Forward)
- Explanation of why each question is valuable

### Stretch Goals
- Fill-in-the-blank questions for learning
- Question ranking exercises
- Gamification to build critical thinking skills

## How It Works

```
+----------------+     +------------------+     +-----------------+
|  Microphone    | --> |  Transcription   | --> |  Context Store  |
|  (Gradio)      |     |  (OpenAI Whisper)|     |  (Accumulating) |
+----------------+     +------------------+     +--------+--------+
                                                        |
                       +--------------------------------+--------+
                       |                                v        |
                       |  +-------------+    +----------------+  |
                       |  | Tavily API  |    | Claude Agent   |  |
                       |  | - Web Search| -> | Question Gen   |  |
                       |  | - News      |    +-------+--------+  |
                       |  | - Fact Check|            |           |
                       |  +-------------+            |           |
                       |        Agent Loop           |           |
                       +-----------------------------+-----------+
                                                     |
                                                     v
                                           +-----------------+
                                           |   Gradio UI     |
                                           |   - Questions   |
                                           |   - Explanations|
                                           +-----------------+
```

## Question Categories

| Category | Description | Example |
|----------|-------------|---------|
| **CLARIFICATION** | Seeks to understand better | "Could you elaborate on what you meant by...?" |
| **DEPTH** | Explores a topic more deeply | "What's the underlying mechanism behind...?" |
| **CONNECTION** | Connects to other fields | "How does this relate to developments in...?" |
| **CHALLENGE** | Probes assumptions respectfully | "Have you considered the alternative view that...?" |
| **PRACTICAL** | Asks about real-world application | "How would this work in practice for...?" |
| **FORWARD** | Explores future implications | "Where do you see this heading in the next...?" |

## Tech Stack

- **Frontend**: Gradio 6
- **Transcription**: OpenAI Whisper API
- **Agent LLM**: Claude (Anthropic)
- **Research**: Tavily API (web search, news, fact-checking)
- **Hosting**: HuggingFace Spaces

## Setup

### Environment Variables

Create a `.env` file with:

```bash
OPENAI_API_KEY=your_openai_api_key
ANTHROPIC_API_KEY=your_anthropic_api_key
TAVILY_API_KEY=your_tavily_api_key
```

### Local Development

```bash
# Clone the repo
git clone https://huggingface.co/spaces/YOUR_USERNAME/ask-the-right-question

# Install dependencies
uv sync

# Run the app
uv run python app.py
```

## Usage

1. **Record Audio**: Click the microphone button to record snippets of the talk
2. **Or Paste Text**: Manually paste notes or transcript
3. **Add Speaker Info**: (Optional) Enter the speaker's name for background research
4. **Generate Questions**: Click the button to get AI-generated questions
5. **Learn**: Read the explanations to understand why each question is valuable

## Demo Video

[Coming soon]

## Social Media Post

[Link to social media post]

## Team

- [Your Name](https://huggingface.co/YOUR_USERNAME)

## Acknowledgments

Built for the MCP's 1st Birthday Hackathon by Hugging Face (November 2025)

Powered by:
- [Anthropic Claude](https://anthropic.com) - AI reasoning and question generation
- [OpenAI Whisper](https://openai.com) - Audio transcription
- [Tavily](https://tavily.com) - AI-optimized web search
- [Gradio](https://gradio.app) - UI framework
- [HuggingFace](https://huggingface.co) - Hosting

## Challenges & Solutions

### Audio Transcription Timing Issues (Solved ✅)

**Original problem:** Short audio chunk intervals caused the transcription API to fall behind, resulting in fragmented audio and information loss.

**Solutions implemented:**
- ✅ **20-second audio chunks** - Longer intervals give Whisper more context and reduce API calls
- ✅ **2-second overlap** - Prevents word cutoffs at chunk boundaries
- ✅ **Background threading** - Transcription and question generation run in separate threads
- ✅ **Sequential queue** - Ensures transcriptions complete in order while audio keeps buffering

## To-Do

### Future Improvements

- [ ] **Add useful MCPs** - Integrate additional MCP servers for enhanced capabilities:
  - Academic paper search (for research talks)
  - Company/organization lookup (for business presentations)
  - Technical documentation search (for developer conferences)
- [ ] **Implement voice activity detection (VAD)** - Only transcribe when speech is detected
- [ ] **Add local transcription fallback** - Use local Whisper model when API latency is too high

## License

MIT