LangChain
Compose real applications around language models — chains, retrievers, memory, tools and agents — with the trade-offs that come with them.
Course Duration: 6h
What You'll Learn
- Decide when LangChain earns its keep and when direct API calls stay simpler
- Compose runnables with the LCEL operator, stream and batch them at will
- Load documents, split them cleanly and build a working retrieval chain
- Give a chat model memory that fits inside a shrinking context budget
- Declare typed tools, hand them to an agent and keep the loop from running away
- Trace, evaluate and debug a chain with LangSmith or an open tracer
Prerequisites
- Course 16 — Large language models
- Course 18 — Retrieval-augmented generation
- Working Python and a running open-weight model (Ollama, an API key, or the setup from course 29)
Course Modules
- What LangChain adds and what it complicates
- Models, prompts and output parsers
- Composing chains
- Document loaders and splitting
- Vector stores and retrievers
- Conversation memory
- Tools and function calling
- Agents: the reasoning and acting loop
- Tracing, evaluation and debugging
- Project: an end-to-end tool-using assistant
The running project
Across the ten modules you build the same object: an expense-reports assistant for a small team. It reads receipts, answers questions about the reimbursement policy, remembers the conversation, calls a currency converter and a per-diem ceiling check, and finishes as a tool-using agent that drafts a spreadsheet line. Each module adds one capability to that same assistant — no throwaway toy examples.
Assessment and certificate
The course ends with a 40-question exam covering every module. On success, a certificate of completion is issued; its number is verifiable on the platform.
Free courses, by contrast, end with a 5-question quiz and a preview of the certificate, without certification.