Prompt templates

A structured blueprint for Large Language Model (LLM) input: it combines fixed instructions with dynamic input variables for consistent, repeatable prompt execution.

Prompt templates are the standard for professional LLM interaction, acting like an email template with merge fields. They separate static, core instructions (e.g., 'Act as a Senior Python Developer') from dynamic data points, which are represented by placeholders like `{user_query}` or `{context_data}`. This structure is critical: it ensures consistency across thousands of API calls, drastically reduces prompt engineering overhead, and facilitates efficient testing and version control. For example, a single template can generate 50 unique product descriptions by simply swapping the product name and key features variables.

https://docs.langchain.com/