Question

A Generative AI Engineer is building a customer support summarization feature. They need to construct prompts programmatically so that each request dynamically injects the customer's name, the product category, and the raw ticket text into a fixed instruction template before sending it to a Foundation Model endpoint. The team wants a maintainable, reusable approach that validates required inputs at runtime and integrates cleanly into a LangChain chain. Which approach best meets these requirements?

Accepted Answer

Use a LangChain PromptTemplate (or ChatPromptTemplate) with defined input variables, then pipe it into the chat model so variables are substituted at invocation time.

Answer

Concatenate the instruction string with the variable values using Python f-strings inside the chain's invoke call each time a request comes in.

Answer

Store the full prompt including all customer data as a static string in a Delta table and read the matching row at query time.

Answer

Fine-tune the foundation model on examples containing the customer name and product category so the values no longer need to be passed at inference.