Advanced RAG Course: Build Text to SQL Agentic AI System
Build production agentic AI systems, from architecture to deployment.
AdvancedProHot
View course
Loading...
Loading...
LangGraph is where most production AI teams land when a simple ReAct loop stops scaling. It gives you stateful graphs instead of ad-hoc function calls, checkpointing so a conversation survives a crash, and first-class streaming that plays nicely with SSE backends. If you've ever hit the wall with plain function-calling agents, LangGraph is the step up.
Curated by Param Harrison
These courses cover LangGraph across the whole surface you actually use: stateful agents, multi-agent supervisors, agentic RAG, document analyzers, and conversational state machines. Each one ships real code you can run locally, with explicit nodes and edges you can read instead of guess at.
Showing 15 of 15 courses
Common questions
Plain function calling is great for a single turn with a handful of tools. LangGraph earns its complexity once you need branching logic, retries, human-in-the-loop pauses, or multi-agent handoffs. Start with function calling, graduate to LangGraph when the loop stops fitting in your head.
Same team, different library. LangChain is the broader toolkit (prompts, loaders, wrappers). LangGraph is the graph runtime. You can use LangGraph without touching LangChain, and most of these courses do.
Yes. It is provider-agnostic, so you can swap OpenAI, Anthropic, Gemini, OpenRouter, or a local Ollama model with the same graph code. Several courses here use LiteLLM as the provider layer to keep things portable.
You declare a TypedDict as your graph state, and each node receives and returns that state. Checkpointing saves it to SQLite or Postgres so you can pause, resume, or branch a run. The stateful agent workflows course walks through this setup end to end.
Yes, out of the box. Each node emits events you can forward as SSE to a browser. The fullstack agents course pairs LangGraph with the Vercel AI SDK so you see streaming with graph nodes on the wire.