fix

yujonglee · yujonglee · commit 05341a02d26d · 2025-08-13T16:22:24.000-07:00
diff --git a/apps/desktop/src/components/right-panel/hooks/useTranscript.ts b/apps/desktop/src/components/right-panel/hooks/useTranscript.ts
@@ -56,9 +56,9 @@ export function useTranscript(sessionId: string | null) {
 
     listenerEvents.sessionEvent.listen(({ payload }) => {
       if (payload.type === "finalWords") {
-        setFinalWords(payload.words as Word[]);
+        setFinalWords((existing) => [...existing, ...payload.words]);
       } else if (payload.type === "partialWords") {
-        setPartialWords((payload.words as Word[]).map(w => ({ ...w, confidence: 0 })));
+        setPartialWords((payload.words as Word[]).map(w => ({ ...w, confidence: -1 })));
       }
     }).then((fn) => {
       unlisten = fn;
diff --git a/apps/desktop/src/routes/app.tsx b/apps/desktop/src/routes/app.tsx
@@ -166,9 +166,11 @@ function RestartSTT() {
     const sttPath = await localSttCommands.modelsDir();
 
     return watch(sttPath, (_event) => {
-      localSttCommands.stopServer("internal").then((stopped) => {
+      localSttCommands.stopServer(null).then((stopped) => {
         if (stopped) {
-          localSttCommands.startServer("internal");
+          localSttCommands.getCurrentModel().then((model) => {
+            localSttCommands.startServer(model);
+          });
         }
       });
     }, { delayMs: 1000 });
diff --git a/packages/tiptap/src/styles/transcript.css b/packages/tiptap/src/styles/transcript.css
@@ -45,7 +45,7 @@
   outline: 2px solid #4a9af4;
 }
 
-span[data-confidence^="0"] {
+span[data-interim] {
   opacity: 0.3;
   color: #666;
 }
diff --git a/packages/tiptap/src/transcript/index.tsx b/packages/tiptap/src/transcript/index.tsx
@@ -10,7 +10,7 @@ import { EditorContent, useEditor } from "@tiptap/react";
 import { forwardRef, useEffect, useRef } from "react";
 
 import { SpeakerSplit } from "./extensions";
-import { ConfidenceMark } from "./marks";
+import { InterimMark } from "./marks";
 import { SpeakerNode } from "./nodes";
 import { fromEditorToWords, fromWordsToEditor, getSpeakerLabel, type SpeakerAttributes, type Word2 } from "./utils";
 import type { SpeakerChangeRange, SpeakerViewInnerComponent, SpeakerViewInnerProps } from "./views";
@@ -52,7 +52,7 @@ const TranscriptEditor = forwardRef<TranscriptEditorRef, TranscriptEditorProps>(
       Document.configure({ content: "speaker+" }),
       History,
       Text,
-      ConfidenceMark,
+      InterimMark,
       SpeakerNode(c),
       SpeakerSplit,
       SearchAndReplace.configure({
diff --git a/packages/tiptap/src/transcript/marks.ts b/packages/tiptap/src/transcript/marks.ts
@@ -1,22 +1,22 @@
 import { Mark, mergeAttributes } from "@tiptap/core";
 
-export const ConfidenceMark = Mark.create({
-  name: "confidence",
+export const InterimMark = Mark.create({
+  name: "interim",
 
   addAttributes() {
     return {
-      confidence: {
+      interim: {
         default: null,
         parseHTML: element => {
-          const value = element.getAttribute("data-confidence");
+          const value = element.getAttribute("data-interim");
           return value ? parseFloat(value) : null;
         },
         renderHTML: attributes => {
-          if (attributes.confidence === null) {
+          if (attributes.interim === null) {
             return {};
           }
           return {
-            "data-confidence": attributes.confidence,
+            "data-interim": attributes.interim,
           };
         },
       },
@@ -26,7 +26,7 @@ export const ConfidenceMark = Mark.create({
   parseHTML() {
     return [
       {
-        tag: "span[data-confidence]",
+        tag: "span[data-interim]",
       },
     ];
   },
diff --git a/packages/tiptap/src/transcript/utils.ts b/packages/tiptap/src/transcript/utils.ts
@@ -67,15 +67,15 @@ export const fromWordsToEditor = (words: Word2[]): DocContent => {
         lastSpeaker.content.push({ type: "text", text: " " });
       }
 
-      if (word.confidence !== null) {
+      if (word.confidence !== null && word.confidence < 0) {
         lastSpeaker.content.push({
           type: "text",
           text: word.text,
           marks: [
             {
-              type: "confidence",
+              type: "interim",
               attrs: {
-                confidence: word.confidence,
+                interim: true,
               },
             },
           ],
@@ -126,16 +126,13 @@ export const fromEditorToWords = (content: DocContent | JSONContent): Word2[] =>
         continue;
       }
 
-      const confidenceMark = node.marks?.find((mark: any) => mark.type === "confidence");
-      const confidence = confidenceMark?.attrs?.confidence ?? null;
-
       const wordTexts = node.text.split(/\s+/).filter(Boolean);
 
       for (const wordText of wordTexts) {
         words.push({
           text: wordText,
           speaker,
-          confidence,
+          confidence: null,
           start_ms: null,
           end_ms: null,
         });
diff --git a/plugins/listener/src/manager.rs b/plugins/listener/src/manager.rs
@@ -2,7 +2,6 @@
 pub struct TranscriptManager {
     id: uuid::Uuid,
     partial_words: Vec<owhisper_interface::Word>,
-    final_words: Vec<owhisper_interface::Word>,
 }
 
 #[derive(Debug, Default, Clone)]
@@ -36,8 +35,6 @@ impl TranscriptManager {
         #[cfg(debug_assertions)]
         Self::log(self.id, &response);
 
-        let mut diff = Diff::default();
-
         if let owhisper_interface::StreamResponse::TranscriptResponse {
             is_final, channel, ..
         } = response
@@ -57,36 +54,31 @@ impl TranscriptManager {
                 .collect::<Vec<_>>();
 
             if is_final {
-                self.process_final_words(&mut diff, words);
+                let last_final_word_end = words.last().unwrap().end;
+                let partial_words = self
+                    .partial_words
+                    .iter()
+                    .filter(|w| w.start > last_final_word_end)
+                    .cloned()
+                    .collect::<Vec<_>>();
+
+                return Diff {
+                    final_words: words.clone(),
+                    partial_words,
+                };
             } else {
-                self.process_partial_words(&mut diff, words);
+                self.partial_words = words.clone();
+
+                return Diff {
+                    final_words: vec![],
+                    partial_words: words.clone(),
+                };
             }
         }
 
-        diff
-    }
-
-    fn process_final_words(&mut self, diff: &mut Diff, words: &Vec<owhisper_interface::Word>) {
-        diff.final_words = words.clone();
-        self.final_words.extend(words.clone());
-        self.partial_words.clear();
+        Diff::default()
     }
 
-    fn process_partial_words(&mut self, diff: &mut Diff, words: &Vec<owhisper_interface::Word>) {
-        let last_final_end = self.get_last_final_end_time();
-
-        self.partial_words = words
-            .into_iter()
-            .filter(|w| w.end > last_final_end)
-            .cloned()
-            .collect();
-
-        diff.partial_words = self.partial_words.clone();
-    }
-
-    fn get_last_final_end_time(&self) -> f64 {
-        self.final_words.last().map(|w| w.end).unwrap_or(0.0)
-    }
     fn log(id: uuid::Uuid, response: &owhisper_interface::StreamResponse) {
         use std::fs::OpenOptions;
         use std::io::Write;

Original file line number	Diff line number	Diff line change
`@@ -45,7 +45,7 @@`
`45`	`45`	`outline: 2px solid #4a9af4;`
`46`	`46`	`}`
`47`	`47`
`48`		`-span[data-confidence^="0"] {`
	`48`	`+span[data-interim] {`
`49`	`49`	`opacity: 0.3;`
`50`	`50`	`color: #666;`
`51`	`51`	`}`