From cf4d8a550c796b38d6ab3e224a87ee7bdd757e63 Mon Sep 17 00:00:00 2001
From: zj1123581321 <agy-executor@invalid>
Date: Sat, 10 Oct 2026 22:49:18 +0800
Subject: [PATCH] =?UTF-8?q?chapters=20segments=20=E7=A7=8D=E5=AD=90?=
 =?UTF-8?q?=E6=94=B9=E7=94=A8=20load=5Fsegments=20=E5=90=8C=E6=BA=90?=
 =?UTF-8?q?=E5=8A=A0=E8=BD=BD=EF=BC=88=E4=BF=AE=20notes=20=E6=8C=87?=
 =?UTF-8?q?=E7=BA=B9=E9=94=99=E9=85=8D=EF=BC=8C#227=EF=BC=89?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

transcription.py 分层缓存补种逻辑的 segments 分支不再取运行态
cache_data["segments"]，改用与 notes reader 同源的
transcriber.segments.load_segments(file_path)，使 chapters 存储指纹
永远可从磁盘复现，修复 notes 生成因指纹错配而确定性失败的问题。

Agent-Executor: agy
Agent-Model: gemini-3.8-flash-high
Agent-Effort: unknown
Dispatch-Id: dlg-20261010-142849-1cebf3
Task-Id: VideoTranscriptAPI-20261015-05
Fixes-Issue: #227
---
 .../api/services/transcription.py             |  2 -
 tests/integration/test_layered_cache.py       | 56 +++++++++++++++++++
 2 files changed, 56 insertions(+), 2 deletions(-)

diff --git a/src/video_transcript_api/api/services/transcription.py b/src/video_transcript_api/api/services/transcription.py
index b9052d15..42c30a7e 100644
--- a/src/video_transcript_api/api/services/transcription.py
+++ b/src/video_transcript_api/api/services/transcription.py
@@ -1705,8 +1705,6 @@ def _fail_task_and_notify(
                     processed.get("dialogs"), list
                 ) and processed["dialogs"]:
                     timeline_segments_seed = processed["dialogs"]
-                elif isinstance(cache_data.get("segments"), list) and cache_data["segments"]:
-                    timeline_segments_seed = cache_data["segments"]
                 else:
                     file_path = cache_data.get("file_path")
                     if file_path:
diff --git a/tests/integration/test_layered_cache.py b/tests/integration/test_layered_cache.py
index c07b8b5a..5d4e7a00 100644
--- a/tests/integration/test_layered_cache.py
+++ b/tests/integration/test_layered_cache.py
@@ -35,6 +35,10 @@
     ChaptersStatus,
     SummaryStatus,
 )
+from video_transcript_api.llm.processors.notes_processor import (
+    compute_notes_anchor_fingerprint,
+)
+from video_transcript_api.transcriber.segments import load_segments
 
 
 class DummyQueue:
@@ -943,6 +947,58 @@ def test_missing_llm_status_does_not_trust_existing_calibrated_file(
         # 重新校对必须喂原始转录文本，而不是那份未经确认的 calibrated 产物。
         assert task["transcript"] == "RAW uncalibrated transcript"
 
+    def test_chapters_seed_uses_load_segments_isomorphic_with_notes_reader(
+        self, monkeypatch, patch_runtime, tmp_path
+    ):
+        """Issue #227: chapters timeline seed for segments backfill must use
+        transcriber.segments.load_segments(cache_dir) rather than in-memory
+        cache_data['segments'], ensuring anchor fingerprint isomorphism."""
+        cache_dir = tmp_path / "cache_entry"
+        cache_dir.mkdir(parents=True)
+        disk_raw = {
+            "segments": [
+                {"start_time": 0.0, "end_time": 1.5, "text": "seg 1", "speaker": "0"},
+                {"start_time": 1.5, "end_time": 3.0, "text": "seg 2", "speaker": "1"},
+            ]
+        }
+        (cache_dir / "transcript_funasr.json").write_text(json.dumps(disk_raw), encoding="utf-8")
+        expected_disk_segments = load_segments(cache_dir)
+        expected_fingerprint = compute_notes_anchor_fingerprint(expected_disk_segments)
+
+        divergent_segments = [
+            {"start_time": 0.0, "end_time": 1.5, "text": "seg 1", "speaker": "Speaker 0"},
+            {"start_time": 1.5, "end_time": 3.0, "text": "seg 2", "speaker": "Speaker 1"},
+        ]
+        assert compute_notes_anchor_fingerprint(divergent_segments) != expected_fingerprint
+
+        cache_data = {
+            **BASE_CACHE_DATA,
+            "file_path": str(cache_dir),
+            "use_speaker_recognition": True,
+            "transcript_type": "funasr",
+            "transcript_data": disk_raw,
+            "segments": divergent_segments,
+            "llm_calibrated": "calibrated text",
+            "llm_summary": "summary text",
+            "llm_status": {
+                "calibration_status": CalibrationStatus.FULL,
+                "summary_status": SummaryStatus.GENERATED,
+                "chapters_status": ChaptersStatus.FAILED,
+            },
+        }
+
+        result, queued = _run(
+            monkeypatch, patch_runtime, cache_data,
+            {"calibrate": False, "summarize": False, "chapters": True, "infer_speaker_names": False},
+        )
+
+        assert result["status"] == "success"
+        assert len(queued) == 1
+        seed_segments = queued[0].get("timeline_segments")
+        assert seed_segments == expected_disk_segments
+        assert compute_notes_anchor_fingerprint(seed_segments) == expected_fingerprint
+
+
 
 class TestFullHitMirrorsCacheStatusOnTaskRow:
     """Regression for codex-review R2 item 2: a full cache hit takes no