Optimization/story telling

[BurakKTopal]

Summary

• Pattern detection: get_local_line_context now detects common code patterns (commented code with active alternatives, TODOs, stale fixes) and provides quick_answer suggestions
• Nearby context: New include_nearby_context parameter shows code before/after selection to detect active implementations
• Confidence scoring: Returns confidence level with specific signals to help LLMs decide when they have enough context
• Updated agent prompts: SYSTEM_PROMPT now guides LLMs to lead with answers ("TL;DR first") instead of archaeology

Test plan

• Run pytest tests/test_nearby_context.py -v to verify pattern detection logic
• Test get_local_line_context on commented code with active implementation below — should return patterns_detected and quick_answer
• Test VS Code extension on commented code — should show faster, more direct answers
• Verify confidence scores appear in tool output