Lab 6 — Pattern Recommendation¶
Goal¶
Prototype a system that recommends a pattern or refactoring based on code/design context.
Possible approaches:
- hand-authored rules;
- classifier/ranker;
- retrieval over examples;
- LLM with structured context;
- hybrid static-analysis + LLM pipeline.
Required evaluation¶
- a baseline;
- top-k recommendation metric or human relevance labels;
- error analysis;
- at least one example where the technically recognized pattern should not be recommended.