refactor: listener transcript and batch processing

apps/desktop/src/components/main/body/sessions/note-input/transcript/shared/index.tsx

@@ -2,7 +2,6 @@ import { type RefObject, useCallback, useMemo, useRef, useState } from "react";
 import { useHotkeys } from "react-hotkeys-hook";
 
 import type { DegradedError } from "@hypr/plugin-listener";
-import type { RuntimeSpeakerHint } from "@hypr/transcript";
 import { DancingSticks } from "@hypr/ui/components/ui/dancing-sticks";
 import { cn } from "@hypr/utils";
 
@@ -46,42 +45,12 @@ export function TranscriptContainer({
   const partialWordsByChannel = useListener(
     (state) => state.partialWordsByChannel,
   );
-  const partialHintsByChannel = useListener(
-    (state) => state.partialHintsByChannel,
-  );
 
   const partialWords = useMemo(
     () => Object.values(partialWordsByChannel).flat(),
     [partialWordsByChannel],
   );
 
-  const partialHints = useMemo(() => {
-    const channelIndices = Object.keys(partialWordsByChannel)
-      .map(Number)
-      .sort((a, b) => a - b);
-
-    const offsetByChannel = new Map<number, number>();
-    let currentOffset = 0;
-    for (const channelIndex of channelIndices) {
-      offsetByChannel.set(channelIndex, currentOffset);
-      currentOffset += partialWordsByChannel[channelIndex]?.length ?? 0;
-    }
-
-    const reindexedHints: RuntimeSpeakerHint[] = [];
-    for (const channelIndex of channelIndices) {
-      const hints = partialHintsByChannel[channelIndex] ?? [];
-      const offset = offsetByChannel.get(channelIndex) ?? 0;
-      for (const hint of hints) {
-        reindexedHints.push({
-          ...hint,
-          wordIndex: hint.wordIndex + offset,
-        });
-      }
-    }
-
-    return reindexedHints;
-  }, [partialWordsByChannel, partialHintsByChannel]);
-
   const containerRef = useRef<HTMLDivElement>(null);
   const [scrollElement, setScrollElement] = useState<HTMLDivElement | null>(
     null,
@@ -169,11 +138,6 @@ export function TranscriptContainer({
                     ? partialWords
                     : []
                 }
-                partialHints={
-                  index === transcriptIds.length - 1 && currentActive
-                    ? partialHints
-                    : []
-                }
                 operations={operations}
               />
               {index < transcriptIds.length - 1 && <TranscriptSeparator />}

apps/desktop/src/components/main/body/sessions/note-input/transcript/shared/render-transcript.tsx

@@ -1,10 +1,6 @@
 import { memo, useEffect, useMemo } from "react";
 
-import type {
-  PartialWord,
-  RuntimeSpeakerHint,
-  Segment,
-} from "@hypr/transcript";
+import type { PartialWord, Segment } from "@hypr/transcript";
 import { cn } from "@hypr/utils";
 
 import * as main from "../../../../../../../store/tinybase/store/main";
@@ -31,7 +27,6 @@ export function RenderTranscript({
   editable,
   transcriptId,
   partialWords,
-  partialHints,
   operations,
 }: {
   scrollElement: HTMLDivElement | null;
@@ -40,7 +35,6 @@ export function RenderTranscript({
   editable: boolean;
   transcriptId: string;
   partialWords: PartialWord[];
-  partialHints: RuntimeSpeakerHint[];
   operations?: Operations;
 }) {
   const finalWords = useFinalWords(transcriptId);
@@ -54,14 +48,7 @@ export function RenderTranscript({
   ) as string | undefined;
   const numSpeakers = useSessionSpeakers(sessionId);
 
-  const allSpeakerHints = useMemo(() => {
-    const finalWordsCount = finalWords.length;
-    const adjustedPartialHints = partialHints.map((hint) => ({
-      ...hint,
-      wordIndex: finalWordsCount + hint.wordIndex,
-    }));
-    return [...finalSpeakerHints, ...adjustedPartialHints];
-  }, [finalWords.length, finalSpeakerHints, partialHints]);
+  const allSpeakerHints = useMemo(() => finalSpeakerHints, [finalSpeakerHints]);
 
   const segments = useStableSegments(
     finalWords,

apps/desktop/src/hooks/useRunBatch.ts

@@ -5,13 +5,7 @@ import type { BatchParams } from "@hypr/plugin-listener2";
 import { useConfigValue } from "../config/use-config";
 import { useListener } from "../contexts/listener";
 import * as main from "../store/tinybase/store/main";
-import type { SpeakerHintWithId, WordWithId } from "../store/transcript/types";
-import {
-  parseTranscriptHints,
-  parseTranscriptWords,
-  updateTranscriptHints,
-  updateTranscriptWords,
-} from "../store/transcript/utils";
+import { makePersistCallback } from "../store/transcript/utils";
 import type { HandlePersistCallback } from "../store/zustand/listener/transcript";
 import { type Tab, useTabs } from "../store/zustand/tabs";
 import { id } from "../utils";
@@ -99,71 +93,8 @@ export const useRunBatch = (sessionId: string) => {
         speaker_hints: "[]",
       });
 
-      const handlePersist: HandlePersistCallback | undefined =
-        options?.handlePersist;
-
-      const persist =
-        handlePersist ??
-        ((words, hints) => {
-          if (words.length === 0) {
-            return;
-          }
-
-          const existingWords = parseTranscriptWords(store, transcriptId);
-          const existingHints = parseTranscriptHints(store, transcriptId);
-
-          const newWords: WordWithId[] = [];
-          const newWordIds: string[] = [];
-
-          words.forEach((word) => {
-            const wordId = id();
-
-            newWords.push({
-              id: wordId,
-              text: word.text,
-              start_ms: word.start_ms,
-              end_ms: word.end_ms,
-              channel: word.channel,
-            });
-
-            newWordIds.push(wordId);
-          });
-
-          const newHints: SpeakerHintWithId[] = [];
-
-          hints.forEach((hint) => {
-            if (hint.data.type !== "provider_speaker_index") {
-              return;
-            }
-
-            const wordId = newWordIds[hint.wordIndex];
-            const word = words[hint.wordIndex];
-
-            if (!wordId || !word) {
-              return;
-            }
-
-            newHints.push({
-              id: id(),
-              word_id: wordId,
-              type: "provider_speaker_index",
-              value: JSON.stringify({
-                provider: hint.data.provider ?? conn.provider,
-                channel: hint.data.channel ?? word.channel,
-                speaker_index: hint.data.speaker_index,
-              }),
-            });
-          });
-
-          updateTranscriptWords(store, transcriptId, [
-            ...existingWords,
-            ...newWords,
-          ]);
-          updateTranscriptHints(store, transcriptId, [
-            ...existingHints,
-            ...newHints,
-          ]);
-        });
+      const persist: HandlePersistCallback =
+        options?.handlePersist ?? makePersistCallback(store, transcriptId);
 
       const params: BatchParams = {
         session_id: sessionId,

apps/desktop/src/hooks/useStartListening.ts

@@ -5,14 +5,7 @@ import { commands as analyticsCommands } from "@hypr/plugin-analytics";
 import { useConfigValue } from "../config/use-config";
 import { useListener } from "../contexts/listener";
 import * as main from "../store/tinybase/store/main";
-import type { SpeakerHintWithId, WordWithId } from "../store/transcript/types";
-import {
-  parseTranscriptHints,
-  parseTranscriptWords,
-  updateTranscriptHints,
-  updateTranscriptWords,
-} from "../store/transcript/utils";
-import type { HandlePersistCallback } from "../store/zustand/listener/transcript";
+import { makePersistCallback } from "../store/transcript/utils";
 import { id } from "../utils";
 import { getSessionEventById } from "../utils/session-event";
 import { useKeywords } from "./useKeywords";
@@ -55,70 +48,6 @@ export function useStartListening(sessionId: string) {
       stt_model: conn.model,
     });
 
-    const handlePersist: HandlePersistCallback = (words, hints) => {
-      if (words.length === 0) {
-        return;
-      }
-
-      store.transaction(() => {
-        const existingWords = parseTranscriptWords(store, transcriptId);
-        const existingHints = parseTranscriptHints(store, transcriptId);
-
-        const newWords: WordWithId[] = [];
-        const newWordIds: string[] = [];
-
-        words.forEach((word) => {
-          const wordId = id();
-
-          newWords.push({
-            id: wordId,
-            text: word.text,
-            start_ms: word.start_ms,
-            end_ms: word.end_ms,
-            channel: word.channel,
-          });
-
-          newWordIds.push(wordId);
-        });
-
-        const newHints: SpeakerHintWithId[] = [];
-
-        if (conn.provider === "deepgram") {
-          hints.forEach((hint) => {
-            if (hint.data.type !== "provider_speaker_index") {
-              return;
-            }
-
-            const wordId = newWordIds[hint.wordIndex];
-            const word = words[hint.wordIndex];
-            if (!wordId || !word) {
-              return;
-            }
-
-            newHints.push({
-              id: id(),
-              word_id: wordId,
-              type: "provider_speaker_index",
-              value: JSON.stringify({
-                provider: hint.data.provider ?? conn.provider,
-                channel: hint.data.channel ?? word.channel,
-                speaker_index: hint.data.speaker_index,
-              }),
-            });
-          });
-        }
-
-        updateTranscriptWords(store, transcriptId, [
-          ...existingWords,
-          ...newWords,
-        ]);
-        updateTranscriptHints(store, transcriptId, [
-          ...existingHints,
-          ...newHints,
-        ]);
-      });
-    };
-
     start(
       {
         session_id: sessionId,
@@ -131,7 +60,7 @@ export function useStartListening(sessionId: string) {
         keywords,
       },
       {
-        handlePersist,
+        handlePersist: makePersistCallback(store, transcriptId),
       },
     );
   }, [

apps/desktop/src/store/transcript/types.ts

@@ -1,4 +1,7 @@
 import type { SpeakerHintStorage, WordStorage } from "@hypr/store";
 
-export type WordWithId = WordStorage & { id: string };
+export type WordWithId = WordStorage & {
+  id: string;
+  state?: "final" | "pending";
+};
 export type SpeakerHintWithId = SpeakerHintStorage & { id: string };

apps/desktop/src/store/transcript/utils.ts

@@ -1,3 +1,7 @@
+import type { TranscriptDelta } from "@hypr/plugin-listener";
+
+import { id } from "../../utils";
+import type { HandlePersistCallback } from "../zustand/listener/transcript";
 import type { SpeakerHintWithId, WordWithId } from "./types";
 
 interface TranscriptStore {
@@ -12,6 +16,7 @@ interface TranscriptStore {
     cellId: "words" | "speaker_hints",
     value: string,
   ): void;
+  transaction<T>(fn: () => T): T;
 }
 
 export function parseTranscriptWords(
@@ -66,3 +71,68 @@ export function updateTranscriptHints(
     JSON.stringify(hints),
   );
 }
+
+export function replaceTranscriptWords(
+  store: TranscriptStore,
+  transcriptId: string,
+  replacedIds: Set<string>,
+  newWords: WordWithId[],
+): void {
+  const existing = parseTranscriptWords(store, transcriptId).filter(
+    (w) => !replacedIds.has(w.id),
+  );
+  const existingHints = parseTranscriptHints(store, transcriptId).filter(
+    (h) => h.word_id == null || !replacedIds.has(h.word_id),
+  );
+  updateTranscriptWords(store, transcriptId, [...existing, ...newWords]);
+  updateTranscriptHints(store, transcriptId, existingHints);
+}
+
+export function makePersistCallback(
+  store: TranscriptStore,
+  transcriptId: string,
+): HandlePersistCallback {
+  return (delta: TranscriptDelta) => {
+    if (delta.new_words.length === 0 && delta.replaced_ids.length === 0) {
+      return;
+    }
+
+    store.transaction(() => {
+      const newWords: WordWithId[] = delta.new_words.map((w) => ({
+        id: w.id,
+        text: w.text,
+        start_ms: w.start_ms,
+        end_ms: w.end_ms,
+        channel: w.channel,
+        state: w.state,
+      }));
+
+      const newHints: SpeakerHintWithId[] = delta.hints.map((h) => ({
+        id: id(),
+        word_id: h.word_id,
+        type: "provider_speaker_index" as const,
+        value: JSON.stringify({ speaker_index: h.speaker_index }),
+      }));
+
+      if (delta.replaced_ids.length > 0) {
+        replaceTranscriptWords(
+          store,
+          transcriptId,
+          new Set(delta.replaced_ids),
+          newWords,
+        );
+      } else {
+        const existing = parseTranscriptWords(store, transcriptId);
+        updateTranscriptWords(store, transcriptId, [...existing, ...newWords]);
+      }
+
+      if (newHints.length > 0) {
+        const existingHints = parseTranscriptHints(store, transcriptId);
+        updateTranscriptHints(store, transcriptId, [
+          ...existingHints,
+          ...newHints,
+        ]);
+      }
+    });
+  };
+}