Add optional PDF and Office document readers
Add --model flag and per-pass model configuration
Incremental AI re-investigation in watch mode
Add --security flag for security-focused analysis mode
Add Markdown and HTML output formats
Add structured JSONL investigation log
Add luminos.toml config file for tunable thresholds
Define superset output fields in synthesis submit_report schema
Update report formatter for dynamic field rendering
Use confidence scores to prioritize refinement targets
Add --refine-depth N for multiple refinement passes
Implement _run_refinement() with --refine flag
Update synthesis prompt for hypothesis formation
Enable think tool in synthesis pass
Implement grouping pass with submit_grouping tool
Implement two-level grouping for xlarge tier
Add tier-appropriate depth instructions to synthesis prompt
Implement tier classification (small/medium/large/xlarge)