Rule Pipeline Sandbox, Whisper GGML Kernels & Text Normalization
Interactive rule preview sandbox with word-level diffs, customizable drag-and-drop rule ordering, high-performance Whisper GGML decode kernels and benchraw harness, SQLite-backed persistent dictation drafts, and comprehensive text normalization suite.
Rule Priority Ordering & Live Preview Sandbox
- Drag-and-drop custom execution ordering for enhancement rules stored in registry (`enhance.rule_order`).
- Interactive live preview sandbox in Settings (Output & Typing) with real-time character and word diffs running against production RuleEnhancer passes.
Whisper GGML Backend & Benchmarking Suite
- Integrated Whisper GGML backend kernels for direct native acceleration and decode execution.
- Standalone benchmarking harness (`benchraw`) for isolated latency profiling and memory throughput verification.
Persistent Dictation Drafts
- Multi-session draft accumulation mode backed by SQLite persistence across application restarts.
Comprehensive Text Normalization Suite
- Automated abbreviation expansion ('e.g.', 'i.e.', 'vs.') with user toggles.
- Spoken number, ordinal, and currency formatting ('twenty dollars' ➔ '$20', '1000' ➔ '1,000').
- Spoken URL and filepath canonicalization ('https colon slash slash github dot com' ➔ 'https://github.com').
- Code identifier casing pipeline (PascalCase, camelCase, snake_case) and opt-in profanity filter.
Rules Pipeline Caching
- Pre-cached lookup tables and match-probe early-outs eliminating redundant regex recompilations.
33b5027ae9f3d203594c5294a290d7a7b99fe4662f8654d479a2ff578902255e