📝 Prompt Engineering

Description

• Prompting is the interface between human intent and model behavior — how you "program" an LLM using natural language. Well-structured prompts can dramatically improve output quality.
• Application prompting requires more rigor than personal prompting: the system must work reliably across varied inputs, without a human in the loop to correct it.

Key points

• Effective Communication: Approach prompts as explaining a task clearly to a competent colleague.
• Systematic Experimentation: Successful prompt engineering requires systematic experimentation and reliable evaluation.
• Clarity: Use precise, unambiguous language that can't be misinterpreted
• Completeness: Include all necessary information. Models only work with what you provide
• Organization: Structure prompts logically with clear sections. Position critical info at the start or at the end.
• Consistency: Maintain uniform terminology, structure, and formatting throughout.
• Positive: Frame instructions positively ("do this") rather than negatively ("don't do that")

Implementation

• https://github.com/yujie-hao/yujie-hao-prompt-engineering-with-langchain

Reference

• Prompt Engineering with LangChain | DataCamp