ai

# Glide AI Glide has powerful AI columns that run inference on your data. These can add significant value to apps with minimal effort. ## AI Column Types | Column | Input | Output | Use Case | |--------|-------|--------|----------| | **Generate Text** | Text/template | AI-generated text | Summaries, descriptions, recommendations | | **Image to Text** | Image URL | Extracted text | Receipt scanning, document OCR, product analysis | | **Document to Text** | Document URL | Extracted/summarized text | PDF parsing, document processing | | **Audio to Text** | Audio URL | Transcription | Voice notes, meeting recordings | | **Text to Boolean** | Text | true/false | Sentiment analysis, spam detection | | **Text to Choice** | Text + options | Selected option | Auto-categorization, priority assignment | | **Text to JSON** | Text | Structured JSON | Entity extraction, form parsing | ## AI Techniques ### Detail Screen Summary with Hint Component A powerful pattern: show an AI-generated summary at the top of detail screens. How it works: 1. **Aggregate data** - Create a Template column or JSON Object column that combines relevant fields - Template: `Name: {Name}, Status: {Status}, Due: {Due Date}, Notes: {Notes}` - Or use JSON Object to structure the data 2. **Generate summary** - Create a Generate Text column that takes the aggregated data - Prompt: "Summarize this task in 1-2 sentences, highlighting priority and next steps" 3. **Display with Hint** - Add a Hint component at the top of the detail screen - Bind it to the Generate Text column - Users see an instant AI summary of the item Example prompts: - Task: "Summarize this task, noting urgency and blockers" - Customer: "Provide a brief profile of this customer based on their activity" - Order: "Summarize this order status and any issues" - Project: "Give a quick health check of this project" This gives users instant context without reading through all fields. ### Auto-Categorization with Text to Choice Automatically categorize items as they're added: 1. Create a Text to Choice column 2. Define the categories (e.g., "Bug", "Feature", "Question", "Other") 3. Point it at the text field to analyze (e.g., ticket description) 4. New items get auto-categorized Great for: - Support tickets → category - Feedback → sentiment (Positive, Negative, Neutral) - Expenses → expense type - Leads → qualification level ### Smart Descriptions from Basic Info Generate rich descriptions from minimal input: 1. User enters basic info (name, a few keywords) 2. Template column combines the inputs 3. Generate Text creates a full description Example for products: - Input: "Blue widget, small, outdoor use" - Generate Text prompt: "Write a compelling 2-sentence product description" - Output: "Introducing our compact Blue Widget, perfectly sized for any outdoor adventure. Built to withstand the elements while delivering reliable performance wherever you go." ### Receipt/Document Scanning Extract structured data from images: 1. User uploads a receipt photo 2. Image to Text extracts the content 3. Text to JSON parses into structured fields (vendor, amount, date) 4. Display parsed data in the UI Works for: - Receipts → expense tracking - Business cards → contact creation - Documents → data entry - Product photos → inventory details ### Sentiment Analysis Detect sentiment in text: 1. Create a Text to Boolean column 2. Prompt: "Is this feedback positive?" 3. Use result to show 😊 or 😞 emoji via If-Then-Else 4. Or use Text to Choice for Positive/Neutral/Negative Apply to: - Customer feedback - Support ticket tone - Review analysis - Survey responses ## Best Practices - **Aggregate before generating** - Combine relevant data into one input for better AI results - **Be specific in prompts** - Tell the AI exactly what format and length you want - **Use for summaries** - AI excels at condensing information - **Cache results** - AI columns compute once and store the result - **Consider cost** - AI columns use Glide AI credits; use strategically ## When to Add AI Look for opportunities where: - Users need to quickly understand complex data - Manual categorization is tedious - Text input could be richer/more helpful - Images contain data that should be structured - Summaries would save users time