Power Up Your App with AI Integrations & Optimizing for Scale - Bolt.new Workshop
In Part 3 of the Bolt.new Workshop Series, we take a working customer feedback portal and supercharge it with AI integrations and performance optimizations that make your app scalable, intelligent, and cost-effective. Getting performance right early means lower hosting costs and faster load times, while integrating AI unlocks powerful capabilities like semantic search and automated insight generation. If you've ever wondered how to handle image-heavy pages efficiently, build vector-based search, or connect your database to an LLM for automatic analysis, this session walks you through each implementation prompt by prompt, showing you exactly how to build these features in a practical, hands-on way. By the end, you'll have an optimized app with smart image compression, paginated data loading, semantic search powered by vector embeddings, and an AI-driven feedback summarization feature that analyzes sentiment and generates actionable recommendations, all built live. Note: While these 4 workshops build on the same project, each one covers stand-alone skills. You can attend one or all four and still follow along. *In this video, we cover* - Smart image compression and thumbnail generation (cutting page loads from 2MB to 18KB) - Pagination strategies for handling large datasets without blowing through bandwidth - Vector embeddings and semantic search with OpenAI and Supabase - Connecting your database to LLMs via server functions - Building an admin "Summarize Feedback" feature that analyzes sentiment and generates recommendations automatically - Importing the latest API docs (like GPT-5.2) to keep your AI integrations current - Debugging auth issues, JWT errors, and prompting strategies for better results *Timestamps:* 00:00 Welcome & workshop roadmap 01:02 Recap: portal, roles, real-time upvotes 03:22 Add screenshot upload to feature requests 05:32 Bandwidth theory & Bolt Cloud limits 08:41 Test upload & duplicate project 12:09 DevTools network analysis 15:24 Thumbnail compression fix 24:28 Verify compression & intro to pagination 29:30 Semantic search, vectors & RAG explained 34:44 Implement embeddings & debug nulls 44:00 Fix 401 embedding error 44:49 Import new OpenAI model docs (GPT-5.2) 47:31 Debug missing embeddings & JWT errors 54:29 Workshop Q&A & support options 56:34 Embeddings working & vector explanation 57:25 Build semantic search modal 01:00:15 Demo: admin summarize feedback feature 01:02:17 Plan mode for deeper reasoning 01:04:35 Semantic search troubleshooting 01:10:12 Wrap-up & documentation as context 01:13:50 API key usage tracking 01:15:27 Debug summarization JWT failure 01:23:39 Final fix & feedback summary demo *About Bolt* Bolt turns ideas into production software in minutes, entirely in your browser. No setup, no servers. Powered by WebContainers. *Connect with us:* X (Twitter): https://x.com/boltdotnew LinkedIn: https://www.linkedin.com/company/boltdotnew/ Questions? Check out our Support page: https://support.bolt.new #BoltNew #AIDevelopment #Prompting #WebDevelopment #NoCode #BuildWithAI #WebApps