The Lang Universe: LangChain, LangGraph, LangSmith & Langfuse

The complete map of the Lang ecosystem: LangChain and LCEL for composable LLM pipelines, LangGraph for stateful agents, LangSmith and Langfuse for observability and evaluation, LangServe and the LangGraph Platform for deployment — plus when to use each, how they work together, and when to swap them for alternatives.

intermediate 90 min 11 lessons

Lessons

  • The Lang Universe Map

    The four-layer mental model — build, orchestrate, observe, serve — and who makes what in the Lang ecosystem.

    8 min
  • LangChain Core — The Building Blocks

    Models, prompts, parsers, retrievers, and tools: the standardized abstractions that make LLM apps composable — and where LangChain stops.

    8 min
  • LCEL — The Expression Language

    The Runnable interface, the | pipe operator, composition primitives, and built-in resilience: retries, fallbacks, streaming, and batching — and where LCEL hits its limits.

    8 min
  • LangGraph — Orchestration & State

    StateGraph, nodes, edges, conditional edges, checkpointers, human-in-the-loop interrupts, time travel, and streaming — the engine for agents and stateful workflows.

    9 min
  • LangSmith — Observability & Evaluation

    LangChain Inc.'s LLMOps platform: tracing, datasets and evals, the Prompt Hub, monitoring, and the eval loop — plus how to hook it up with zero code changes.

    8 min
  • Langfuse — Open-Source Observability

    The independent, self-hostable LLMOps platform: tracing, sessions, prompt management, evals, cost tracking — and how the OpenTelemetry GenAI standard makes the observe layer swappable.

    8 min
  • LangSmith vs. Langfuse — The Big Decision

    Head-to-head comparison of the two observability platforms: licensing, self-hosting, ecosystem ties, cost tracking, lock-in — plus a decision framework and the other players (Phoenix, Weave, Helicone).

    8 min
  • Serving & Deploying — LangServe, Platform, Studio

    How Lang apps become APIs: LangServe for stateless chains, the LangGraph Platform for stateful production agents, LangGraph Studio for local debugging — and a decision table for which to use.

    9 min
  • Memory — LangMem & the Persistence Stack

    Context vs. short-term (checkpointer/threads) vs. long-term (LangMem) memory: semantic, episodic, and procedural stores, and when each layer is worth adding.

    8 min
  • The Wider Ecosystem & When to Swap

    LlamaIndex, CrewAI, AutoGen, OpenAI Agents SDK, Semantic Kernel, Haystack — and a practical playbook for swapping models, observability, vector stores, and orchestrators.

    9 min
  • Grand Quiz — The Lang Universe

    Twenty questions, two per lesson, graded A–D. Test your map of the whole ecosystem: LangChain, LCEL, LangGraph, LangSmith, Langfuse, serving, memory, and swapping.

    15 min