More stuff

- fix unicode output from python terminal
- fix cpu inference
- add filters
- add beam search params to UI
- DRY up config definition in UI

4 files changed, 125 insertions, 56 deletions

diff --git a/ui/config-schema.js b/ui/config-schema.js
new file mode 100644
index 0000000..b1108ff
--- /dev/null
+++ b/ui/config-schema.js
@@ -0,0 +1,49 @@
+// Shared configuration schema with types and defaults
+const CONFIG_SCHEMA = {
+    // String fields
+    compute_type: { type: 'select', default: 'float16' },
+    language: { type: 'select', default: 'english' },
+    model: { type: 'select', default: 'turbo' },
+    microphone: { type: 'number', default: 0 },
+    user_prompt: { type: 'text', default: 'Use proper punctuation and grammar. Prefer spelled out numbers like one, eleven, twenty, etc. Mm.' },
+    
+    // Number fields
+    gpu_idx: { type: 'number', default: 0 },
+    max_speech_duration_s: { type: 'number', default: 10 },
+    min_silence_duration_ms: { type: 'number', default: 250 },
+    reset_after_silence_s: { type: 'number', default: 15 },
+    transcription_loop_delay_ms: { type: 'number', default: 100 },
+    block_width: { type: 'number', default: 2 },
+    num_blocks: { type: 'number', default: 40 },
+    rows: { type: 'number', default: 10 },
+    cols: { type: 'number', default: 24 },
+    beam_size: { type: 'number', default: 5 },
+    best_of: { type: 'number', default: 5 },
+    
+    // Boolean fields (stored as 1/0)
+    enable_debug_mode: { type: 'boolean', default: 0 },
+    enable_previews: { type: 'boolean', default: 1 },
+    save_audio: { type: 'boolean', default: 0 },
+    use_cpu: { type: 'boolean', default: 0 },
+    enable_lowercase_filter: { type: 'boolean', default: 0 },
+    enable_uppercase_filter: { type: 'boolean', default: 0 },
+    enable_profanity_filter: { type: 'boolean', default: 0 },
+    remove_trailing_period: { type: 'boolean', default: 0 }
+};
+
+// Helper to extract just the default values
+function getDefaultConfig() {
+    const defaults = {};
+    for (const [key, schema] of Object.entries(CONFIG_SCHEMA)) {
+        defaults[key] = schema.default;
+    }
+    return defaults;
+}
+
+// Export for both CommonJS (main process) and ES modules (renderer)
+if (typeof module !== 'undefined' && module.exports) {
+    module.exports = { CONFIG_SCHEMA, getDefaultConfig };
+} else {
+    window.CONFIG_SCHEMA = CONFIG_SCHEMA;
+    window.getDefaultConfig = getDefaultConfig;
+} 
\ No newline at end of file
diff --git a/ui/index.html b/ui/index.html
index 90f78c1..97da3d2 100644
--- a/ui/index.html
+++ b/ui/index.html
@@ -10,9 +10,9 @@
     <div class="container-fluid px-6 py-6 h-screen flex flex-col">
         <div class="flex flex-1 gap-6 overflow-hidden">
             <!-- Left Panel: Configuration Form -->
-            <div class="max-w-96 relative flex flex-col">
+            <div class="max-w-96 relative flex flex-col overflow-hidden rounded-lg">
                 <!-- Loading Overlay -->
-                <div id="loading-overlay" class="absolute inset-0 bg-white bg-opacity-75 backdrop-blur-sm z-50 hidden flex items-center justify-center rounded-lg">
+                <div id="loading-overlay" class="absolute inset-0 bg-white bg-opacity-75 backdrop-blur-sm z-50 hidden flex items-center justify-center">
                     <div class="text-center p-6">
                         <div class="animate-spin rounded-full h-12 w-12 border-b-2 border-blue-600 mx-auto mb-4"></div>
                         <p class="text-gray-700 font-medium"></p>
@@ -126,7 +126,7 @@
                                 <h2 class="section-title">Transcription Settings</h2>
                                 <div>
                                     <label for="user_prompt" class="form-label">
-                                        Custom Prompt
+                                        Prompt
                                         <span class="text-gray-500 text-xs block mt-1"
                                               title="Whisper is given this prompt before transcribing. It helps guide the transcription style. For example, you could improve the spelling of your friends' names with: 'My friends' names are Saoirse, Azariah, and Caoimhe.'">
                                             (Hover for details)
@@ -136,6 +136,28 @@
                                               class="form-input h-20 resize-none" 
                                               placeholder="My friends' names are Saoirse, Azariah, and Caoimhe."></textarea>
                                 </div>
+                                <div class="grid grid-cols-2 gap-4 mt-4">
+                                    <div>
+                                        <label for="beam_size" class="form-label">
+                                            Beam size
+                                            <span class="text-gray-500 text-xs block mt-1"
+                                                  title="Number of beams for beam search. Higher values may improve accuracy but increase compute time.">
+                                                (Search width)
+                                            </span>
+                                        </label>
+                                        <input type="number" id="beam_size" min="1" max="20" value="5" class="form-input">
+                                    </div>
+                                    <div>
+                                        <label for="best_of" class="form-label">
+                                            Best of
+                                            <span class="text-gray-500 text-xs block mt-1"
+                                                  title="Number of candidates to generate when sampling. The best one will be selected.">
+                                                (Sampling candidates)
+                                            </span>
+                                        </label>
+                                        <input type="number" id="best_of" min="1" max="20" value="5" class="form-input">
+                                    </div>
+                                </div>
                             </section>
 
                             <!-- Performance Settings -->
@@ -166,6 +188,29 @@
                                 </div>
                             </section>
 
+                            <!-- Text Filters -->
+                            <section class="config-section">
+                                <h2 class="section-title">Text Filters</h2>
+                                <div class="space-y-3">
+                                    <label for="enable_lowercase_filter" class="checkbox-label">
+                                        <input type="checkbox" id="enable_lowercase_filter" class="mr-2">
+                                        <span class="checkbox-text">Convert to lowercase</span>
+                                    </label>
+                                    <label for="enable_uppercase_filter" class="checkbox-label">
+                                        <input type="checkbox" id="enable_uppercase_filter" class="mr-2">
+                                        <span class="checkbox-text">Convert to uppercase</span>
+                                    </label>
+                                    <label for="enable_profanity_filter" class="checkbox-label">
+                                        <input type="checkbox" id="enable_profanity_filter" class="mr-2">
+                                        <span class="checkbox-text">Filter profanity</span>
+                                    </label>
+                                    <label for="remove_trailing_period" class="checkbox-label">
+                                        <input type="checkbox" id="remove_trailing_period" class="mr-2">
+                                        <span class="checkbox-text">Remove trailing period</span>
+                                    </label>
+                                </div>
+                            </section>
+
                             <!-- Display Settings -->
                             <section class="config-section">
                                 <h2 class="section-title">Custom Chatbox Settings</h2>
@@ -240,6 +285,7 @@
         </div>
     </div>
 
+    <script src="config-schema.js"></script>
     <script src="renderer.js"></script>
 </body>
 </html>
diff --git a/ui/index.js b/ui/index.js
index 2420ece..7717c92 100644
--- a/ui/index.js
+++ b/ui/index.js
@@ -4,6 +4,7 @@ const fs = require('node:fs').promises;
 const yaml = require('js-yaml');
 const { spawn } = require('child_process');
 const https = require('https');
+const { CONFIG_SCHEMA, getDefaultConfig } = require('./config-schema.js');
 
 const APP_ROOT = path.join(__dirname, '..');
 const CONFIG_PATH = path.join(APP_ROOT, 'config.yaml');
@@ -82,6 +83,14 @@ function downloadFile(url, outputPath) {
   });
 }
 
+function shouldFilterMessage(message) {
+  // Filter out pydub ffmpeg/avconv warning. It does not actually matter.
+  if (message.includes("Couldn't find ffmpeg or avconv - defaulting to ffmpeg, but may not work")) {
+    return true;
+  }
+  return false;
+}
+
 // Helper function to setup process event handlers
 function setupProcessHandlers(process) {
   process.stdout.on('data', (data) => {
@@ -91,7 +100,9 @@ function setupProcessHandlers(process) {
   
   process.stderr.on('data', (data) => {
     const text = data.toString();
-    sendPythonOutput(text.trimEnd(), 'stderr');
+    if (!shouldFilterMessage(text)) {
+      sendPythonOutput(text.trimEnd(), 'stderr');
+    }
   });
   
   process.on('error', (error) => {
@@ -137,7 +148,10 @@ function executePythonCommand(args, options = {}) {
     pythonProcess.stderr.on('data', (data) => {
       const text = data.toString();
       stderr += text;
-      sendPythonOutput(text.trimEnd(), 'stderr');
+      // Filter out specific warning messages
+      if (!shouldFilterMessage(text)) {
+        sendPythonOutput(text.trimEnd(), 'stderr');
+      }
     });
     
     pythonProcess.on('error', (error) => {
@@ -171,27 +185,8 @@ function createWindow () {
   mainWindow.loadFile('index.html');
 }
 
-// Default configuration based on user's current config.yaml
-const DEFAULT_CONFIG = {
-  compute_type: 'float16',
-  enable_debug_mode: 0,
-  enable_previews: 1,
-  user_prompt: 'Use proper punctuation and grammar. Prefer spelled out numbers like one, eleven, twenty, etc.',
-  save_audio: 0,
-  language: 'english',
-  gpu_idx: 0,
-  max_speech_duration_s: 10,
-  min_silence_duration_ms: 250,
-  microphone: 0,
-  model: 'turbo',
-  reset_after_silence_s: 15,
-  transcription_loop_delay_ms: 100,
-  use_cpu: 0,
-  block_width: 2,
-  num_blocks: 40,
-  rows: 10,
-  cols: 24
-};
+// Replace the DEFAULT_CONFIG constant with:
+const DEFAULT_CONFIG = getDefaultConfig();
 
 // IPC handlers
 ipcMain.handle('load-config', async () => {
@@ -521,12 +516,12 @@ ipcMain.handle('start-process', async () => {
 });
 
 ipcMain.handle('stop-process', async () => {
-  if (!runningProcess) {
-    throw new Error('No process is running');
-  }
-  
   return new Promise((resolve) => {
     let forcefullyKilled = false;
+
+    if (!runningProcess) {
+      resolve({ success: true, forcefullyKilled });
+    }
     
     // Set up a timeout to force kill after 10 seconds
     const killTimeout = setTimeout(() => {
diff --git a/ui/renderer.js b/ui/renderer.js
index 201eef6..133a79b 100644
--- a/ui/renderer.js
+++ b/ui/renderer.js
@@ -1,29 +1,5 @@
-// Configuration and form field mappings
-const CONFIG_FIELDS = {
-    // String fields
-    compute_type: { type: 'select', default: 'float16' },
-    language: { type: 'select', default: 'english' },
-    model: { type: 'select', default: 'turbo' },
-    microphone: { type: 'number', default: 0 },
-    user_prompt: { type: 'text', default: '' },
-    
-    // Number fields
-    gpu_idx: { type: 'number', default: 0 },
-    max_speech_duration_s: { type: 'number', default: 10 },
-    min_silence_duration_ms: { type: 'number', default: 250 },
-    reset_after_silence_s: { type: 'number', default: 15 },
-    transcription_loop_delay_ms: { type: 'number', default: 100 },
-    block_width: { type: 'number', default: 2 },
-    num_blocks: { type: 'number', default: 40 },
-    rows: { type: 'number', default: 10 },
-    cols: { type: 'number', default: 24 },
-    
-    // Boolean fields (stored as 1/0)
-    enable_debug_mode: { type: 'boolean', default: 0 },
-    enable_previews: { type: 'boolean', default: 1 },
-    save_audio: { type: 'boolean', default: 0 },
-    use_cpu: { type: 'boolean', default: 0 }
-};
+// Import configuration schema
+const CONFIG_FIELDS = window.CONFIG_SCHEMA;
 
 // Button management system
 class ButtonManager {
@@ -35,6 +11,9 @@ class ButtonManager {
             resetVenv: document.getElementById('reset-venv'),
             refreshMicrophones: document.getElementById('refresh-microphones')
         };
+        
+        // Initialize button states on construction
+        this.setProcessStopped();
     }
     
     setState(buttonName, disabled) {