Preset sweeps (4 enterprise use cases)

Presets are opinionated defaults designed to be a starting point. Each preset includes:

• default parameter settings
• a recommended prompt pattern
• evaluation metrics aligned to the use case

Presets in this repo:

• summarization
• rag_qa
• classification
• code_generation

Run with:

python scripts/run_sweep_rest.py --preset rag_qa --trials 5

Then analyze:

python scripts/analyze_results.py --input runs/latest.jsonl