The Pipe Operator, Demystified

Every operator in Python — +, -, ==, and yes, | — is syntax sugar for a method call on the left-hand object. a | b is literally a.__or__(b). Python's standard library uses | for bitwise OR on integers by default, but any class is free to redefine what | means for its own instances by implementing __or__ — which is exactly what LangChain's Runnable class does.

Once you know prompt | model | parser is just three chained __or__ calls, LCEL stops looking like special framework magic and starts looking like ordinary Python composition — because that's exactly what it is. This understanding transfers directly: any time you see | chaining LangChain components, you now know precisely what's happening underneath.

Modern LangChain code rarely uses SimpleSequentialChain directly — it uses LCEL (LangChain Expression Language), which composes steps with Python's | pipe operator: prompt | model | parser. It looks like magic. It's actually just Python operator overloading, and you're about to build it yourself.

Python lets any class define what the | operator means for its instances, by implementing __or__. LangChain's Runnable base class implements __or__ to mean exactly one thing: 'compose these two steps so the first one's output feeds the second one's input.'

Implement the Pipe Operator Yourself. Finish __or__: it should return a new Runnable whose function, when called, runs self first, then feeds that result into other. Once this one method works, uppercase | add_exclamation composes automatically — that's the entire trick behind LCEL's pipe syntax.

When Python evaluates pipeline = uppercase | add_exclamation, what method does it actually call?

• →It calls uppercase.__or__(add_exclamation) — the | operator is just syntax sugar for the __or__ dunder method, exactly like + calls __add__.
• →It's special Python syntax that only works inside the LangChain library itself.

Module 2 complete: you've built a Chain three different ways — the classic LLMChain, SimpleSequentialChain, and now the pipe-operator mechanism behind LCEL. Module 3 gives your chains something they've been missing entirely so far: memory across turns.