feat: add custom dictionary with words and misspelling corrections#5
Open
nicremo wants to merge 3 commits into
Open
feat: add custom dictionary with words and misspelling corrections#5nicremo wants to merge 3 commits into
nicremo wants to merge 3 commits into
Conversation
- Cloud transcription via Groq/OpenAI-compatible APIs (whisper-large-v3) - Auto/Cloud/Local transcription mode with automatic offline fallback - API key encrypted via macOS Keychain (Electron safeStorage) - Default text model changed from gemma4:e4b (9.6GB) to qwen3.5:2b (2.7GB) - Configurable API base URL (Groq, OpenAI, Lemonfox, any compatible provider) - Language selector (German default, 11 languages available) - Stronger same-language prompt to prevent LLM translation - Built-in microphone preferred over external devices (AirPods fix) - New TranscriptionCard UI with source selector, API key management - Setup wizard with cloud/local transcription choice - Relaxed hotkey validation: Ollama not required when enhancement is off
Dictionary & Corrections: - Custom vocabulary tab with words and misspelling corrections - Words sent as Whisper prompt hints for better transcription - Corrections injected into LLM system prompt for auto-replacement - Two-column layout: Words (left) + Misspellings (right) - Async file lock prevents race conditions on concurrent writes - Whisper prompt truncated at ~800 chars (224 token limit) - IPC handlers with runtime input validation LLM Performance: - Disable thinking mode (think: false) for qwen3.5 models - Reduces rewrite time from ~14s to ~0.3s - Strip <think> tags from output as safety fallback Pipeline logging: - Log transcription settings, raw text, and final text for debugging
This was referenced Apr 13, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds a custom dictionary system with two features: vocabulary words for better Whisper transcription accuracy, and misspelling correction rules for the LLM rewrite pass. Also fixes a major performance issue with thinking-mode LLMs.
Custom Dictionary
promptparameter (max ~224 tokens). Helps Whisper correctly spell domain-specific terms, names, and technical jargon.LLM Performance Fix
think: falsesent to Ollama for models like Qwen 3.5 and DeepSeek: reduces rewrite time from ~14 seconds to ~0.3 seconds<think>tags from output as a safety fallbackWhat changed
dictionary.ts(CRUD operations, file persistence, Whisper prompt builder, LLM context builder)dictation.ts(dictionary/corrections pipeline integration),ollama.ts(think: false, tag stripping),prompts.ts(dictionary context injection),App.tsx(dictionary UI page),ipc.ts(dictionary/correction handlers)Test plan