feat: search backend (#537)

* docs: transient docs * chore: cleanup * webvtt WIP * webvtt field * chore: webvtt tests comments * chore: remove useless tests * feat: search TASK.md * feat: full text search by title/webvtt * chore: search api task * feat: search api * feat: search API * chore: rm task md * chore: roll back unnecessary validators * chore: pr review WIP * chore: pr review WIP * chore: pr review * chore: top imports * feat: better lint + ci * feat: better lint + ci * feat: better lint + ci * feat: better lint + ci * chore: lint * chore: lint * fix: db datetime definitions * fix: flush() params * fix: update transcript mutability expectation / test * fix: update transcript mutability expectation / test * chore: auto review * chore: new controller extraction * chore: new controller extraction * chore: cleanup * chore: review WIP * chore: pr WIP * chore: remove ci lint * chore: openapi regeneration * chore: openapi regeneration * chore: postgres test doc * fix: .dockerignore for arm binaries * fix: .dockerignore for arm binaries * fix: cap test loops * fix: cap test loops * fix: cap test loops * fix: get_transcript_topics * chore: remove flow.md docs and claude guidance * chore: remove claude.md db doc * chore: remove claude.md db doc * chore: remove claude.md db doc * chore: remove claude.md db doc
2025-12-22 13:19:05 +00:00 · 2025-08-13 10:03:38 -04:00
parent a42ed12982
commit 6fb5cb21c2
29 changed files with 3213 additions and 1493 deletions
--- a/server/reflector/utils/webvtt.py
+++ b/server/reflector/utils/webvtt.py
@@ -0,0 +1,63 @@
+"""WebVTT utilities for generating subtitle files from transcript data."""
+
+from typing import TYPE_CHECKING, Annotated
+
+import webvtt
+
+from reflector.processors.types import Seconds, Word, words_to_segments
+
+if TYPE_CHECKING:
+    from reflector.db.transcripts import TranscriptTopic
+
+VttTimestamp = Annotated[str, "vtt_timestamp"]
+WebVTTStr = Annotated[str, "webvtt_str"]
+
+
+def _seconds_to_timestamp(seconds: Seconds) -> VttTimestamp:
+    # lib doesn't do that
+    hours = int(seconds // 3600)
+    minutes = int((seconds % 3600) // 60)
+    secs = int(seconds % 60)
+    milliseconds = int((seconds % 1) * 1000)
+
+    return f"{hours:02d}:{minutes:02d}:{secs:02d}.{milliseconds:03d}"
+
+
+def words_to_webvtt(words: list[Word]) -> WebVTTStr:
+    """Convert words to WebVTT using existing segmentation logic."""
+    vtt = webvtt.WebVTT()
+    if not words:
+        return vtt.content
+
+    segments = words_to_segments(words)
+
+    for segment in segments:
+        text = segment.text.strip()
+        # lib doesn't do that
+        text = f"<v Speaker{segment.speaker}>{text}"
+
+        caption = webvtt.Caption(
+            start=_seconds_to_timestamp(segment.start),
+            end=_seconds_to_timestamp(segment.end),
+            text=text,
+        )
+        vtt.captions.append(caption)
+
+    return vtt.content
+
+
+def topics_to_webvtt(topics: list["TranscriptTopic"]) -> WebVTTStr:
+    if not topics:
+        return webvtt.WebVTT().content
+
+    all_words: list[Word] = []
+    for topic in topics:
+        all_words.extend(topic.words)
+
+    # assert it's in sequence
+    for i in range(len(all_words) - 1):
+        assert (
+            all_words[i].start <= all_words[i + 1].start
+        ), f"Words are not in sequence: {all_words[i].text} and {all_words[i + 1].text} are not consecutive: {all_words[i].start} > {all_words[i + 1].start}"
+
+    return words_to_webvtt(all_words)