Building AI Enabled Web Apps
summary
Built two demos: ask the docs, a semantic-search QA over Laravel docs, and discount GPT, a chat client exercising function-calling tools. Uses Laravel, Inertia, Vue, Tailwind, Postgres with PGVector, OpenAI JS SDK, and Sequelize-style bindings. Key numbers: LangChain ~6,000 downloads/week vs React ~22M, embedding vectors of 1,536 floats, max_tokens ~1,024 (~800 words), temperature 0 for determinism. Implements RAG with embeddings, multi-query generation, and reciprocal-rank fusion to combine results; function calls run via tool messages and return JSON to the model.
AI-written summary · about
chapters (8)
- 0:00Introduction and speaker background
- 5:47Demo apps overview
- 8:13Prompting fundamentals
- 17:30Vector embeddings explained
- 24:08Retrieval augmented generation
- 30:31RAG fusion with multiquery
- 37:17AI assistant and function calling
- 45:02Summary and Q&A
related talks
24mEmpowering Vue Apps with AI
48mGoing Beyond Prompt Engineering
37mBest AI Tools for Web Developers / How to Use AI in Web Development
40m4 pillars for building a modern AI application in JavaScript
58mDemystifying GenAI: Build a ChatGPT App with Vector Store
28mAPIs and GenAI : Your Perfect Combo with LLM-Powered Frameworks.
45mWeb Development in the Age of AI
42mAdd Useful AI to Your Web App (Not Just Chatbots)