What is a «Chain» (Link) in LangChain?
It is a specific instruction:
llm_chain = prompt | llm | output_parser
An example perfect and modern of what means a Chain in the architecture of LangChain (specifically using the LangChain Expression Language or LCEL).
A Chain (A Chain) in LangChain is the structured sequence of steps or components that takes an input, processes it through one or more modules, and produces output.
Think of a chain as a assembly line for complex tasks of LLM. Instead of executing each component (prompt, model call, processing) separately, the chains to link the output of one step automatically to become the input of the next.
The example in the code snippet llm_chain = prompt | llm | output_parser illustrates the fundamental chain, replacing the old LLMChain:
| Step (Component) | Purpose in the Chain |
1. prompt (Input) |
Takes user input (e.g., {"concepto": "Machine Learning"}). It is the first link. |
| **2. ` | ` (Pipe) |
3. llm (Processing) |
The language model (your Ollama remote). It receives the final prompt and executes it, generating raw text output. |
| **4. ` | ` (Pipe) |
5. output_parser (Output) |
The output is processed by the LLM. In your case, StrOutputParser takes the model’s response and ensures it is a clean string, preparing it for the final output. |
When you call llm_chain.invoke({"concepto": "Machine Learning"}), this happens:
