Feat/add confidence scoring rules

[AAdewunmi]

Summary

Adds deterministic confidence scoring rules for study insights.

Changes

• Added rule-based confidence scoring in apps/insights/nlp/confidence.py.
• Added Low, Medium, and High confidence label mapping.
• Added focused tests in apps/insights/tests/test_confidence.py.
• Clarified confidence wording in docs/ai-nlp-contract.md.

Behaviour

• Scores confidence from 0 to 100.
• Uses meaningful token count, unique meaningful token count, keyword count, and usable summary presence.
• Returns 0 for empty or non-meaningful note content.
• Does not reward fallback low-information summaries.
• Produces repeatable scores for the same inputs.
• Labels scores as:
  • Low below 45
  • Medium from 45 to 74
  • High from 75 upward

Why

Confidence should be a transparent quality signal for generated study insights. It should indicate whether the source notes contain enough meaningful content to support useful keywords and an extractive summary, without implying model intelligence, probability, or factual correctness.

Validation

• pytest apps/insights/tests/test_confidence.py -q
• ruff check apps/insights/nlp/confidence.py apps/insights/tests/test_confidence.py
• black --check apps/insights/nlp/confidence.py apps/insights/tests/test_confidence.py

Result

All targeted confidence tests, lint checks, and format checks pass.

Notes

This is deterministic heuristic scoring only. It does not call an AI model and does not claim factual correctness.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!