AI Session Insights 🎯

A Streamlit app that analyzes conference session photos using Snowflake Cortex AISQL to automatically count attendees and detect raised hands, providing real-time insights into audience engagement and badge distribution rates.

✨ Features

• 🤖 AI-Powered Analysis: Leverages Snowflake Cortex AISQL for intelligent image processing
• 📊 Real-Time Analytics: Instant attendee counting and engagement metrics
• 🙋‍♀️ Hand Raise Detection: Automatically identifies potential badge recipients
• 📈 Visual Insights: Interactive charts and conversion rate analysis
• 🖼️ Image Display: View analyzed session photos with metadata
• ☁️ Cloud-Native: Built on Snowflake's native Streamlit platform

🚀 Demo

Upload conference session photos and watch as the AI analyzes:

• Total attendee count
• Number of raised hands
• Engagement conversion rates
• Session-by-session comparisons

Perfect for live demos showcasing AI capabilities in real-world scenarios!

🛠️ Technology Stack

• Snowflake Cortex AISQL - AI-powered SQL for image analysis
• Streamlit in Snowflake - Native app development platform
• Streamlit - Interactive web application framework
• Python - Core application logic
• Altair - Data visualization

📋 Prerequisites

• Snowflake account with Cortex AISQL enabled
• Database and schema with appropriate permissions
• Snowflake stage for file storage

Don't have Snowflake? Sign up for a free trial to get started!

⚙️ Setup

1. Create Snowflake Objects

    -- Create database and schema
    CREATE DATABASE IF NOT EXISTS KAMESH_DEMOS;
    CREATE SCHEMA IF NOT EXISTS KAMESH_DEMOS.CONFERENCES;
    
    -- Create stage for image uploads
    CREATE STAGE IF NOT EXISTS KAMESH_DEMOS.CONFERENCES.SNAPS
    ENCRYPTION = (TYPE = 'SNOWFLAKE_SSE')
    DIRECTORY = (ENABLE = true, AUTO_REFRESH = true);
    
    -- Create Git Integration the repo or its fork
    CREATE OR REPLACE API INTEGRATION kameshsampath_github
        API_PROVIDER = git_https_api
        API_ALLOWED_PREFIXES = ('https://github.com/kameshsampath')
    ENABLED = TRUE;
   

2. Create Core AI View

   -- This is the heart of the app - AI-powered image analysis
   CREATE OR REPLACE VIEW SMART_CROWD_COUNTER AS 
   WITH image_files AS (
       SELECT 
           relative_path as name,
           TO_FILE('@kamesh_demos.conferences.snaps',relative_path) as file,
           last_modified AS last_modified
       FROM DIRECTORY('@kamesh_demos.conferences.snaps')
       WHERE relative_path LIKE '%.jpg' OR relative_path LIKE '%.jpeg'
   ),
   processed_images AS (
       SELECT 
           name,
           file,
           last_modified,
           AI_COMPLETE(
               'claude-4-sonnet',
              'Analyze image: count people and raised hands. Return JSON only: {"total_attendees": N, "raised_hands": N, "percentage_with_hands_up": N.NN}. Calculate percentage as (raised_hands/attendees)*100, round to 2 decimals.',file
           ) AS attendees_count
       FROM image_files
   )
   SELECT 
       name,
       file as file_name,
       AI_COMPLETE('claude-4-sonnet',PROMPT('# Conference Photo Caption Generator Prompt
   
   Create a concise, descriptive caption for a conference/workshop photo with the filename: `{0}`
   
   ## Context clues from filename:
   - **NS** = Northstar
   - **SWT** = Snowflake World Tour  
   - **Location abbreviations** (e.g., PUNE, DELHI, MEL for Melbourne)
   - **Numbers** likely indicate session sequence or day
   
   ## Caption requirements:
   - Keep it simple and descriptive
   - Format: *Event Name + Location + Session Type/Number*
   - Use italics for styling
   - No hashtags or additional text
   - **Add "Workshop" if you see attendees with laptops or materials in front of them (hands-on learning environment)**
   - Focus on identifying the event, location, and session clearly
   
   ## Example format:
   *[Event Name] [City] [Session Type/Number]*
   
   **Generate a clean, minimal caption that clearly identifies the event and session.**',name)) as caption,
       attendees_count as raw,
       PARSE_JSON(attendees_count):total_attendees AS total_attendees,
       PARSE_JSON(attendees_count):raised_hands AS raised_hands,  
       PARSE_JSON(attendees_count):percentage_with_hands_up AS percentage_with_hands_up
   FROM processed_images
   ORDER BY name;

3. Deploy to Snowflake

   • Upload the Python file to your Snowflake environment
   • Create a Streamlit app using Snowflake's native Streamlit support
   • Configure database and schema connections

4. Configure Environment

   • Update database and schema names in the code if different
   • Ensure proper permissions for file uploads and stage access

📱 Usage

1. Select Database & Schema: Choose your target database and schema
2. Upload Images: Use the file uploader to add conference session photos (JPG, PNG, JPEG)
3. AI Analysis: Watch as AISQL processes images and extracts insights
4. View Results: Browse the data table with attendee metrics
5. Detailed Analytics: Click any row to see the original image with analytics charts

🎯 Use Cases

• Conference Analytics: Track session attendance and engagement
• Event Management: Monitor audience participation in real-time
• Badge Distribution: Identify participants for giveaways or prizes
• AI Demonstrations: Showcase computer vision capabilities
• Business Intelligence: Generate insights from event photography

📊 Output Metrics

• Total Attendees: AI-counted number of people in the session
• Raised Hands: Count of participants with hands up
• Conversion Rate: Percentage of attendees with raised hands
• File Metadata: Image details, timestamps, and storage info

🔧 Customization

• Database Configuration: Update cat_database and cat_schema defaults
• Stage Settings: Modify stage name and encryption settings
• UI Styling: Customize colors, charts, and layout
• Analysis Parameters: Adjust AI model settings and thresholds

🙏 Acknowledgments

• Snowflake for the powerful Cortex AISQL platform
• Streamlit for the intuitive app development framework
• Community for inspiration and feedback

📞 Support

For questions or issues:

• Open a GitHub issue
• Check Snowflake documentation for AISQL specifics
• Review Streamlit docs for UI customization

---

⭐ Star this repo if it helped you build something cool!