LangChain Framework for Beginners – Build AI Systems + RAG

Set up a TypeScript workspace by installing node and Visual Studio Code, then install Lang Chain core and Lang Graph packages and configure your path for macOS and Windows.

Set up a new node project for LangChain with TypeScript by creating a folder, initializing npm, and installing LangChain core, TypeScript, and jot, plus optional OpenAI or cloud models.

Set up a LangChain agent in a TypeScript file and configure a model (anthropic or OpenAI). Invoke it with a user prompt in a messages array, then run with npx.

Explore how to set up Anthropic API keys and environment variables, install dependencies with npm, and extract the AI's core content from agent response messages for step by step logging.

Explore how LangChain tools function as the core framework component, enabling an AI agent to access your weather app code base and external systems in real time.

Implement a tool function with two arguments: the executable function and its metadata, defining a city input schema to fetch weather and guide the language model's tool selection.

Analyze how an agent uses internal tools to fetch weather and time, orchestrating tool calls and interpreting outputs to answer user queries.

Learn how runtime configuration lets agents personalize responses by using backend data, retrieving user location via a tool and user ID, and chaining tools to fetch weather.

Learn how to integrate a weather tool into a LangChain agent, derive the city from a user location at runtime, and fetch the weather via tool calls.

Learn how system prompts shape a LangChain based agent's behavior at run time, using prod/qa/dev configurations and tools like weather and user location to guide tasks.

Define a custom response schema with jarred to produce structured outputs. Learn to adjust agent behavior using qa config and system prompts for humanized, weather-like responses.

Customize the chat model in LangChain by setting temperature, timeout, and token limits to shape responses and support ai systems with rag workflows.

Master memory management in LangChain by using the memory saver class to persist conversations across a thread, with thread IDs and configurable settings guiding memory.

Explore middleware in the LangChain framework to tightly control agent behavior. Implement dynamic model selection to switch between cheaper and advanced models based on message count.

Configure middleware to dynamically select the model at run time, switching between base and advanced models based on message count, using a dynamic model selection key.

Build a LangChain agent with tools and a primary OpenAI GPT four model, then use a model fallback middleware to switch to a secondary model if the main fails.

Explore how to use summarization middleware and tool selector middleware in LangChain to reduce token costs, trigger concise replies near limits, and filter tools before the main model.

See how the LM tool selector shortlists tools for an agent, uses GPT-4 mini for tool selection, and explains when to use few tools to save agent model cost.

Learn to implement guardrails and PII protection by using a regex-based redaction middleware to filter and mask credit cards, SSNs, and phone numbers before AI processing.

Explore retrieval augmented generation (rag) and its architecture, including two forms, overcoming finite context by retrieving from documents and live sources to build an agentic rag application.

Explore the rag architecture, where embedding models power semantic search over a vector store to retrieve top documents, augment a query, and generate answers in an end-to-end pipeline.

Master the retrieval pipeline in LangChain by converting documents to embeddings, loading PDFs via a PDF loader, storing in a vector store, and building a RAG-enabled agent to answer queries.

Learn how to chunk documents using the recursive character text splitter, configure chunk size and overlap, and improve retrieval speed by indexing smaller chunks for embeddings and vector stores.

Learn to choose a vector store and load embeddings in a rag pipeline with LangChain, store documents, perform semantic search, and retrieve context for an llm.

Configure vector search retrieval with similarity or MMR and fetch k for top documents. Tune retriever parameters and monitor retrieval context to implement two-step rag and agent rag.

Build a custom middleware in LangChain to enable retrieval augmented generation. Customize the dynamic system prompt to pull vector store context and retrieved documents before invoking the agent.

learn how to load multiple pdf documents into a vector store and perform retrieval with LangChain rag integration, enabling multi-document queries with scalable embedding and retrieval.

Learn how to load word documents into a vector store and retrieve them with LangChain RAG integration using a doc loader and Mammoth, with embeddings and a unified prompt.

Explore the open source MCP standard and tools that enable LangChain AI agents to connect to external systems, with steps to install and integrate MCP tooling.

Set up a local MCP server and integrate it with LangChain for RAG applications by installing dependencies, configuring MySQL, and using CRUD tools to manage products.

Learn to integrate MCP tools with LangChain to power AI agents via stdio transport, expose MCP servers, fetch tools, and build end-to-end agentic rag workflows.

Showcases an end-to-end agentic rag workflow by integrating MCP tools with LangChain to perform retrieval, tool orchestration, and product data validation (Nike Air ID 28).

Install and set up Lang Graph Studio, install lang graph and cli, expose a rag agent as a graph node, configure graph json and env vars, and run dev server.

Add the Lang Smith API key to enable tracing, real-time monitoring, and observability of agent behavior in the UI, then explore rag agents and debugging with the observability dashboard.

Explore how to build ai agents with LangChain, from fundamentals and tool integration to retrieval augmented generation, dynamic prompts, memory management, and deployment as a graph node.