4 pillars for building a modern AI application in JavaScript
summary
Live-coding builds a TypeScript AI app using LangChain.js to call a local LLaMA-based LLM with streaming and system + few‑shot prompts, then adds 1024‑dim vector embeddings stored in an in‑memory retriever to implement RAG and pull PDFs for answers (it finds $1.90 a day per chicken). Tools are exposed as JSON‑schema endpoints, served by an MCP server on port 3000, and an agentic LangGraph workflow successfully routes to a donate call (LangChain.js).
AI-written summary · about
chapters (7)
- 0:00Introduction to AI pillars
- 1:05Large language models
- 10:30Vector embeddings and RAG
- 18:56Tools and function calling
- 25:28Model Context Protocol
- 29:59Agentic workflows
- 34:05Q&A and final thoughts
related talks
34mJavaScript-Powered AI: Crafting Intelligent Systems with LLMs and Vector Databases
23mBuild intelligent apps with JS: integrating RAG, Azure OpenAI and Langchain.js
01h 59mLLM applications with LangChain
25mLangChain with Azure AI: Building LLM powered Applications
02h 49mJava meets AI: Build LLM-Powered Apps with LangChain4j
55mAI in Java? Meet LangChain4j and Build Intelligent Services
48mBuilding AI Powered Applications with Java
49mBuilding AI Enabled Web Apps