Interactive architecture advice
RAG vs Prompting vs Fine-Tuning
Answer a few questions to explore a suggested pattern and its trade-offs. Each answer explains why it fits, what you gain and what to check before putting it into practice.
Prompting supplies instructions and context. Retrieval-augmented generation (RAG) finds relevant documents. Fine-tuning changes model behaviour using examples. Each solves a different problem.
01Your requirements
Choose the main problem: fitting a small context, finding knowledge, or making behaviour more consistent.
Examples should reflect real cases and have reviewed target outputs. Keep a separate evaluation set.
Suggested pattern
Answer the questions to explore a starting pattern.
Architecture depends on workload, risk, team capability and constraints. Validate assumptions and evaluate a small proof of concept; this is a starting point for your design.
Turn this into a practical decision
- Create an evaluation set before changing the approach.
- Try a prompt baseline and measure where it fails.
- For retrieval, test source access and relevance; for fine-tuning, test behaviour on unseen examples.