Skip to content

Speakr is a personal, self-hosted web application designed for transcribing audio recordings

Python

Record where a title came from (#412): typed titles are kept, generated ones are regenerated on reprocess; a merge without a typed title is titled like an upload (falls back to "<first> (merged)" without an LLM); merge dialog hint updated in 7 languages

Murtaza Nasirc5810c58caa8
diff --git a/src/api/api_v1.py b/src/api/api_v1.py
index abb91390..52b4e4a1 100644
--- a/src/api/api_v1.py
+++ b/src/api/api_v1.py
@@ -1237,6 +1237,8 @@ def update_recording(recording_id):
# Update fields if provided
if 'title' in data:
+ if data['title'] != recording.title:
+ recording.title_source = 'user'
recording.title = data['title']
changed_fields.append('title')
if 'participants' in data:
diff --git a/src/api/recordings.py b/src/api/recordings.py
index f97b9ce3..69b02fcd 100644
--- a/src/api/recordings.py
+++ b/src/api/recordings.py
@@ -41,7 +41,7 @@ from src.services.transcription_defaults import (
from src.tasks.processing import format_transcription_for_llm, _resolve_timestamp_template_format
from src.utils.dates import to_utc_naive
from src.utils.ffmpeg_utils import FFmpegError, FFmpegNotFoundError
-from src.utils.titles import resolve_upload_title
+from src.utils.titles import resolve_upload_title, upload_title_source
from src.services.speaker import (
update_speaker_usage, apply_speaker_map, participants_from_segments, update_voice_profiles,
)
@@ -1346,6 +1346,7 @@ def regenerate_title(recording_id):
return jsonify({'error': 'Failed to generate a title'}), 500
recording.title = new_title
+ recording.title_source = 'auto'
db.session.commit()
# Keep the exported file in step with the new title, as /save does.
try:
@@ -2083,6 +2084,8 @@ def save_metadata():
# Update fields requiring edit permission
if requires_edit:
if 'title' in data:
+ if data['title'] != recording.title:
+ recording.title_source = 'user'
recording.title = data['title']
if 'participants' in data:
recording.participants = data['participants']
@@ -2341,7 +2344,7 @@ def share_target():
# (resolve_upload_title) so the AI title task recognises it and generates
# a title — previously the filename stem was used, which the title task
# treated as a user-chosen title and skipped, leaving shared files untitled.
- from src.utils.titles import resolve_upload_title
+ from src.utils.titles import resolve_upload_title, upload_title_source
share_title = resolve_upload_title(request.form.get('title'), original_filename)
notes_parts = []
@@ -2355,6 +2358,7 @@ def share_target():
audio_path=filepath,
original_filename=original_filename,
title=share_title,
+ title_source=upload_title_source(request.form.get('title')),
status='PENDING',
user_id=current_user.id,
notes=share_notes,
@@ -2970,6 +2974,7 @@ def ingest_uploaded_recording(
audio_path=None,
original_filename=original_filename,
title=resolve_upload_title(user_title, original_filename),
+ title_source=upload_title_source(user_title),
file_size=final_file_size,
status='PENDING',
meeting_date=meeting_date,
diff --git a/src/init_db.py b/src/init_db.py
index c1fe7681..f527fad8 100644
--- a/src/init_db.py
+++ b/src/init_db.py
@@ -749,6 +749,11 @@ def _run_migrations(app, engine):
except Exception as e:
app.logger.warning(f"Could not create unique index on user.sso_subject: {e}")
+ # Where a recording's title came from (#412): user-typed titles are
+ # kept on reprocess, generated ones are regenerated.
+ if add_column_if_not_exists(engine, 'recording', 'title_source', 'VARCHAR(16)'):
+ app.logger.info("Added title_source column to recording table")
+
# Add file_hash column for duplicate detection
if add_column_if_not_exists(engine, 'recording', 'file_hash', 'VARCHAR(64)'):
app.logger.info("Added file_hash column to recording table")
diff --git a/src/models/recording.py b/src/models/recording.py
index 4a049b41..c34c6632 100644
--- a/src/models/recording.py
+++ b/src/models/recording.py
@@ -22,6 +22,10 @@ class Recording(db.Model):
id = db.Column(db.Integer, primary_key=True)
# Title will now often be AI-generated, maybe start with filename?
title = db.Column(db.String(200), nullable=True) # Allow Null initially
+ # Where the title came from (#412): 'user' (typed in a form, edit or API),
+ # 'auto' (title step: AI, naming template or filename). NULL for recordings
+ # from before #412 and for placeholders; see utils.titles.title_is_user_chosen.
+ title_source = db.Column(db.String(16), nullable=True)
participants = db.Column(db.String(500))
notes = db.Column(db.Text)
transcription = db.Column(db.Text, nullable=True)
diff --git a/src/services/recording_merge.py b/src/services/recording_merge.py
index 23dd009b..a9cfa9a6 100644
--- a/src/services/recording_merge.py
+++ b/src/services/recording_merge.py
@@ -180,7 +180,13 @@ def create_merge_recording(user, recording_ids, title=None, delete_originals=Fal
ordered_ids = [r.id for r in recordings]
first = recordings[0]
- merged_title = (title or '').strip() or f"{first.title or 'Recording'} (merged)"
+ # A typed title is kept. A blank title gets the upload placeholder, so the
+ # title step names the merged recording like an upload (AI title or naming
+ # template); without an LLM its fallback is the "<first title> (merged)"
+ # filename, the name merges had before (#412).
+ from src.utils.titles import resolve_upload_title, upload_title_source
+ user_title = (title or '').strip()
+ merged_name = user_title or f"{first.title or 'Recording'} (merged)"
now = datetime.utcnow()
@@ -223,8 +229,9 @@ def create_merge_recording(user, recording_ids, title=None, delete_originals=Fal
recording = Recording(
audio_path=None,
- original_filename=f"{merged_title}.m4a",
- title=merged_title,
+ original_filename=f"{merged_name}.m4a",
+ title=resolve_upload_title(user_title, f"{merged_name}.m4a"),
+ title_source=upload_title_source(user_title),
# 'QUEUED' matches what job_queue.enqueue() sets for a transcription-side
# job; set it here too so the row is consistent even before enqueue.
status='QUEUED',
diff --git a/src/services/recording_stitch.py b/src/services/recording_stitch.py
index 7a89c4ec..ebc4cc41 100644
--- a/src/services/recording_stitch.py
+++ b/src/services/recording_stitch.py
@@ -380,9 +380,10 @@ def stitch_recording_session(session_id: str) -> Tuple[int, str, dict]:
# title — instead of a fabricated default being mistaken for a user
# choice and AI titling getting skipped (the same class of bug fixed for
# the share target in 482614c).
- from src.utils.titles import resolve_upload_title
+ from src.utils.titles import resolve_upload_title, upload_title_source
user_title = (metadata.get('title') or '').strip()
recording.title = resolve_upload_title(user_title, final_filename)
+ recording.title_source = upload_title_source(user_title)
recording.audio_path = final_path
recording.original_filename = final_filename
recording.file_size = file_size
diff --git a/src/services/upload_join.py b/src/services/upload_join.py
index 2a918702..81ffbb99 100644
--- a/src/services/upload_join.py
+++ b/src/services/upload_join.py
@@ -144,7 +144,7 @@ def create_joined_recording(owner, parts, *, title=None, notes=None, folder=None
prompt_variables=None, transcribe_params=None):
"""Create the recording for a claimed group and queue the merge job. Commits."""
from src.services.job_queue import job_queue
- from src.utils.titles import resolve_upload_title
+ from src.utils.titles import resolve_upload_title, upload_title_source
now = datetime.utcnow()
original_filename = _joined_filename(parts[0].original_filename)
@@ -154,6 +154,7 @@ def create_joined_recording(owner, parts, *, title=None, notes=None, folder=None
audio_path=None,
original_filename=original_filename,
title=resolve_upload_title(title, original_filename),
+ title_source=upload_title_source(title),
file_size=sum(p.file_size or 0 for p in parts),
status='QUEUED',
meeting_date=min(meeting_dates) if meeting_dates else now,
diff --git a/src/tasks/processing.py b/src/tasks/processing.py
index 2b40c590..680b43c7 100644
--- a/src/tasks/processing.py
+++ b/src/tasks/processing.py
@@ -404,8 +404,8 @@ def generate_title_task(app_context, recording_id, will_auto_summarize=False):
# is_placeholder_title is the shared source of truth (also used by the
# upload + share-target routes) so every entry point's title is
# recognised here and gets an AI title unless the user chose one.
- from src.utils.titles import is_placeholder_title
- if not is_placeholder_title(recording.title, recording.original_filename):
+ from src.utils.titles import title_is_user_chosen
+ if title_is_user_chosen(recording):
current_app.logger.info(f"Recording {recording_id} has user-provided title '{recording.title}', skipping AI title generation")
db.session.commit()
if not will_auto_summarize:
@@ -419,6 +419,7 @@ def generate_title_task(app_context, recording_id, will_auto_summarize=False):
if final_title:
recording.title = final_title
+ recording.title_source = 'auto'
current_app.logger.info(f"Title set for recording {recording_id}: {final_title}")
else:
current_app.logger.warning(f"Could not generate title for recording {recording_id}")
diff --git a/src/utils/titles.py b/src/utils/titles.py
index 1c7dc069..18f81c6d 100644
--- a/src/utils/titles.py
+++ b/src/utils/titles.py
@@ -40,6 +40,27 @@ def resolve_upload_title(user_title, original_filename):
return placeholder_title(original_filename)
+def upload_title_source(user_title):
+ """title_source for a recording created with this form value: 'user' when a
+ title was typed, else None (a placeholder the title step replaces)."""
+ return 'user' if user_title and user_title.strip() else None
+
+
+def title_is_user_chosen(recording):
+ """True when the title step must keep the recording's title (#412).
+
+ 'user' titles are kept and 'auto' titles are regenerated, for example on a
+ transcription reprocess. Recordings from before title_source existed keep
+ the earlier rule: a placeholder gets a title, anything else is kept.
+ """
+ source = getattr(recording, 'title_source', None)
+ if source == 'user':
+ return True
+ if source == 'auto':
+ return False
+ return not is_placeholder_title(recording.title, recording.original_filename)
+
+
def is_placeholder_title(title, original_filename):
"""True if `title` is empty or an auto-generated placeholder.
diff --git a/static/locales/de.json b/static/locales/de.json
index acc0210a..360a04ce 100644
--- a/static/locales/de.json
+++ b/static/locales/de.json
@@ -1343,7 +1343,7 @@
"moveUp": "Nach oben",
"moveDown": "Nach unten",
"titleLabel": "Titel (optional)",
- "titlePlaceholder": "Leer lassen, um den Titel der ersten Aufnahme zu verwenden",
+ "titlePlaceholder": "Leer lassen für einen automatischen Titel wie beim Hochladen",
"deleteOriginals": "Ursprüngliche Aufnahmen löschen",
"deleteOriginalsHint": "Entfernt die Quellaufnahmen, nachdem die zusammengeführte Aufnahme erstellt wurde. Standardmäßig deaktiviert.",
"reprocessNote": "Das kombinierte Audio wird erneut transkribiert und zusammengefasst, um eine durchgehende Transkription und Zusammenfassung zu erstellen.",
diff --git a/static/locales/en.json b/static/locales/en.json
index 4dcb87a8..fefa62b3 100644
--- a/static/locales/en.json
+++ b/static/locales/en.json
@@ -1343,7 +1343,7 @@
"moveUp": "Move up",
"moveDown": "Move down",
"titleLabel": "Title (optional)",
- "titlePlaceholder": "Leave blank to use the first recording's title",
+ "titlePlaceholder": "Leave blank for an automatic title, as for an upload",
"deleteOriginals": "Delete original recordings",
"deleteOriginalsHint": "Remove the source recordings after the merged one is created. Off by default.",
"reprocessNote": "The combined audio will be transcribed and summarized again to produce one continuous transcript and summary.",
diff --git a/static/locales/es.json b/static/locales/es.json
index 4aeca8fe..62e65be8 100644
--- a/static/locales/es.json
+++ b/static/locales/es.json
@@ -1343,7 +1343,7 @@
"moveUp": "Subir",
"moveDown": "Bajar",
"titleLabel": "Título (opcional)",
- "titlePlaceholder": "Déjalo en blanco para usar el título de la primera grabación",
+ "titlePlaceholder": "Déjalo en blanco para un título automático, como en una subida",
"deleteOriginals": "Eliminar las grabaciones originales",
"deleteOriginalsHint": "Elimina las grabaciones de origen después de crear la combinada. Desactivado de forma predeterminada.",
"reprocessNote": "El audio combinado se transcribirá y resumirá de nuevo para producir una transcripción y un resumen continuos.",
diff --git a/static/locales/fr.json b/static/locales/fr.json
index a74f087c..a5427d0a 100644
--- a/static/locales/fr.json
+++ b/static/locales/fr.json
@@ -1343,7 +1343,7 @@
"moveUp": "Monter",
"moveDown": "Descendre",
"titleLabel": "Titre (facultatif)",
- "titlePlaceholder": "Laissez vide pour utiliser le titre du premier enregistrement",
+ "titlePlaceholder": "Laissez vide pour un titre automatique, comme pour un import",
"deleteOriginals": "Supprimer les enregistrements d'origine",
"deleteOriginalsHint": "Supprime les enregistrements sources après la création de l'enregistrement fusionné. Désactivé par défaut.",
"reprocessNote": "L'audio combiné sera de nouveau transcrit et résumé afin de produire une transcription et un résumé continus.",
diff --git a/static/locales/pt-BR.json b/static/locales/pt-BR.json
index 8aa28350..1762dc8f 100644
--- a/static/locales/pt-BR.json
+++ b/static/locales/pt-BR.json
@@ -1343,7 +1343,7 @@
"moveUp": "Mover para cima",
"moveDown": "Mover para baixo",
"titleLabel": "Título (opcional)",
- "titlePlaceholder": "Deixe em branco para usar o título da primeira gravação",
+ "titlePlaceholder": "Deixe em branco para um título automático, como em um envio",
"deleteOriginals": "Excluir gravações originais",
"deleteOriginalsHint": "Remove as gravações de origem após a criação da mesclada. Desativado por padrão.",
"reprocessNote": "O áudio combinado será transcrito e resumido novamente para produzir uma transcrição e um resumo contínuos.",
diff --git a/static/locales/ru.json b/static/locales/ru.json
index 986c563e..28159725 100644
--- a/static/locales/ru.json
+++ b/static/locales/ru.json
@@ -1343,7 +1343,7 @@
"moveUp": "Вверх",
"moveDown": "Вниз",
"titleLabel": "Название (необязательно)",
- "titlePlaceholder": "Оставьте пустым, чтобы использовать название первой записи",
+ "titlePlaceholder": "Оставьте пустым, чтобы название создалось автоматически, как при загрузке",
"deleteOriginals": "Удалить исходные записи",
"deleteOriginalsHint": "Удаляет исходные записи после создания объединённой. По умолчанию отключено.",
"reprocessNote": "Объединённый звук будет заново расшифрован и обобщён, чтобы создать непрерывную расшифровку и резюме.",
diff --git a/static/locales/zh.json b/static/locales/zh.json
index 6f3f0d90..720e3b3a 100644
--- a/static/locales/zh.json
+++ b/static/locales/zh.json
@@ -1343,7 +1343,7 @@
"moveUp": "上移",
"moveDown": "下移",
"titleLabel": "标题(可选)",
- "titlePlaceholder": "留空则使用第一条录音的标题",
+ "titlePlaceholder": "留空则像上传一样自动生成标题",
"deleteOriginals": "删除原始录音",
"deleteOriginalsHint": "创建合并录音后删除源录音。默认关闭。",
"reprocessNote": "合并后的音频将被重新转录和总结,以生成连续的转录文本和摘要。",
diff --git a/tests/test_title_source.py b/tests/test_title_source.py
new file mode 100644
index 00000000..0a5cff41
--- /dev/null
+++ b/tests/test_title_source.py
@@ -0,0 +1,185 @@
+"""Where a title came from decides whether processing may replace it (#412).
+
+- 'user': typed in an upload form, an edit or the API; never replaced.
+- 'auto': produced by the title step; replaced when the recording is
+ processed again, for example on a transcription reprocess (decision Da).
+- NULL: recordings from before title_source existed keep the earlier rule
+ (a placeholder gets a title, anything else is kept), so an upgrade changes
+ no existing title.
+A merge without a typed title is titled like an upload (decision Dg); without
+an LLM it falls back to the "<first title> (merged)" name merges had before.
+
+SHARED-DB: everything created here is removed afterwards.
+"""
+
+import os
+import sys
+import uuid
+from types import SimpleNamespace
+from unittest.mock import patch
+
+import pytest
+from flask import g
+from flask.testing import FlaskClient
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from src.app import app, db
+from src.models import Recording, User
+from src.services import recording_merge
+from src.tasks import processing
+from src.utils.titles import title_is_user_chosen, upload_title_source
+
+app.config["WTF_CSRF_ENABLED"] = False
+TRANSCRIPT = "Speaker 1: We agreed to move the launch to next quarter and review the budget on Monday."
+
+
+class _Client(FlaskClient):
+ def open(self, *args, **kwargs):
+ g.pop('_login_user', None)
+ return super().open(*args, **kwargs)
+
+
+def made():
+ ids = []
+ yield ids
+ with app.app_context():
+ db.session.rollback()
+ for model, oid in reversed(ids):
+ obj = db.session.get(model, oid)
+ if obj is not None:
+ db.session.delete(obj)
+ db.session.commit()
+
+
+def _user(made):
+ s = uuid.uuid4().hex[:8]
+ u = User(username=f"tsrc_{s}", email=f"tsrc_{s}@local.test", password="x")
+ db.session.add(u)
+ db.session.commit()
+ made.append((User, u.id))
+ return u
+
+
+def _rec(made, user_id, **kw):
+ kw.setdefault("original_filename", "call.wav")
+ kw.setdefault("transcription", TRANSCRIPT)
+ kw.setdefault("status", "PROCESSING")
+ r = Recording(user_id=user_id, **kw)
+ db.session.add(r)
+ db.session.commit()
+ made.append((Recording, r.id))
+ return r
+
+
+# ------------------------------------------------------------------ the rule
+
[email protected]("title,source,expected", [
+ ("Board meeting", "user", True),
+ ("Launch moved", "auto", False),
+ ("Recording - call.wav", None, False), # placeholder, any era
+ ("Board meeting", None, True), # legacy title: kept
+ (None, None, False),
+])
+def test_title_is_user_chosen(title, source, expected):
+ rec = SimpleNamespace(title=title, title_source=source, original_filename="call.wav")
+ assert title_is_user_chosen(rec) is expected
+
+
[email protected]("value,expected", [("Board meeting", "user"), (" ", None), ("", None), (None, None)])
+def test_upload_title_source(value, expected):
+ assert upload_title_source(value) == expected
+
+
+# ------------------------------------------- reprocess re-titles generated titles
+
+def _run_title_step(rid, ai="Launch moved"):
+ with patch.object(processing, "_generate_ai_title", return_value=ai), \
+ patch.object(processing, "client", object()):
+ processing.generate_title_task(app.app_context(), rid, will_auto_summarize=True)
+
+
[email protected]("source,old,expected", [
+ ("auto", "Old AI title", "Launch moved"),
+ ("user", "Board meeting", "Board meeting"),
+ (None, "Board meeting", "Board meeting"),
+ (None, "Recording - call.wav", "Launch moved"),
+])
+def test_processing_again_replaces_only_generated_titles(made, source, old, expected):
+ with app.app_context():
+ u = _user(made)
+ r = _rec(made, u.id, title=old, title_source=source)
+ _run_title_step(r.id)
+ db.session.expire_all()
+ out = db.session.get(Recording, r.id)
+ assert out.title == expected
+ if expected == "Launch moved":
+ assert out.title_source == "auto"
+
+
+def test_an_edited_title_is_kept_on_reprocess(made):
+ with app.app_context():
+ u = _user(made)
+ r = _rec(made, u.id, title="Launch moved", title_source="auto", status="COMPLETED")
+ c = _Client(app, app.response_class, use_cookies=True)
+ with c.session_transaction() as sess:
+ sess["_user_id"] = str(u.id)
+ with patch("src.api.recordings.export_recording"):
+ resp = c.post("/save", json={"id": r.id, "title": "Q3 planning"})
+ assert resp.status_code == 200, resp.get_data(as_text=True)[:300]
+ db.session.expire_all()
+ assert db.session.get(Recording, r.id).title_source == "user"
+ _run_title_step(r.id)
+ db.session.expire_all()
+ assert db.session.get(Recording, r.id).title == "Q3 planning"
+
+
+# ------------------------------------------------------------- merge (Dg)
+
+def _merge(made, user, sources, title=None):
+ with patch.object(recording_merge, "_validate_sources", return_value=sources):
+ merged = recording_merge.create_merge_recording(user, [s.id for s in sources], title=title)
+ made.append((Recording, merged.id))
+ return merged
+
+
+def test_a_merge_without_a_title_is_titled_like_an_upload(made):
+ with app.app_context():
+ u = _user(made)
+ a = _rec(made, u.id, title="Standup", status="COMPLETED")
+ b = _rec(made, u.id, title="Standup part 2", status="COMPLETED")
+ m = _merge(made, u, [a, b])
+ assert m.original_filename == "Standup (merged).m4a"
+ assert m.title_source is None and not title_is_user_chosen(m)
+ m.transcription = TRANSCRIPT
+ db.session.commit()
+ _run_title_step(m.id)
+ db.session.expire_all()
+ assert db.session.get(Recording, m.id).title == "Launch moved"
+
+
+def test_a_merge_without_a_title_or_llm_keeps_the_old_name(made):
+ with app.app_context():
+ u = _user(made)
+ a = _rec(made, u.id, title="Standup", status="COMPLETED")
+ b = _rec(made, u.id, title="Standup part 2", status="COMPLETED")
+ m = _merge(made, u, [a, b])
+ with patch.object(processing, "client", None):
+ processing.generate_title_task(app.app_context(), m.id, will_auto_summarize=True)
+ db.session.expire_all()
+ assert db.session.get(Recording, m.id).title == "Standup (merged)"
+
+
+def test_a_typed_merge_title_is_kept(made):
+ with app.app_context():
+ u = _user(made)
+ a = _rec(made, u.id, title="Standup", status="COMPLETED")
+ b = _rec(made, u.id, title="Standup part 2", status="COMPLETED")
+ m = _merge(made, u, [a, b], title="Weekly sync")
+ assert (m.title, m.title_source) == ("Weekly sync", "user")
+ m.transcription = TRANSCRIPT
+ db.session.commit()
+ _run_title_step(m.id)
+ db.session.expire_all()
+ assert db.session.get(Recording, m.id).title == "Weekly sync"