Update job detail page to track revisions and display document image next to transcription text.

README.md

@@ -10,6 +10,7 @@ The app lets you upload a document image/PDF, queues a background transcription
 - Persist document + job records in SQLite
 - Process jobs in a background worker (`queued -> processing -> transcribed/failed`)
 - Store transcript text (or failure detail)
+- Track transcript revisions (AI-generated and manual updates)
 - Show status and results in the NiceGUI interface
 
 ## Quick start
@@ -61,8 +62,11 @@ uv run uvicorn transcription.app:create_app --factory --reload
 
 - **Job detail page** (`/ui/jobs/{job_id}`)
   - Shows job metadata and status.
-  - Displays transcript text when successful.
-  - Displays failure detail when transcription fails.
+  - Shows transcript metadata, including provider and model.
+  - Shows a version table with `Created` and `Version`.
+  - Displays latest version text in an editable textbox.
+  - **Update** creates a new transcript version.
+  - Displays failure detail for failed revisions.
 
 ## Prompt artifacts
 

src/transcription/db/operations.py

@@ -58,3 +58,9 @@ def _ensure_sqlite_compat_columns(connection: Connection) -> None:
     if "retry_count" not in columns:
         connection.execute(text("ALTER TABLE job ADD COLUMN retry_count INTEGER NOT NULL DEFAULT 0"))
         logger.warning("Applied SQLite compatibility schema patch table=job column=retry_count default=0")
+
+    if "transcript" in table_names:
+        transcript_columns = {column["name"] for column in inspector.get_columns("transcript")}
+        if "model" not in transcript_columns:
+            connection.execute(text("ALTER TABLE transcript ADD COLUMN model VARCHAR"))
+            logger.warning("Applied SQLite compatibility schema patch table=transcript column=model")

src/transcription/models.py

@@ -1,7 +1,9 @@
 """SQLModel domain models for the transcription system.
 
-Three models capture the MVP lifecycle:
-  Document  ->  one-to-many  ->  Job  ->  one-to-one  ->  Transcript
+Core models capture the MVP lifecycle:
+Document (1) -> (many) Job
+Job (1) -> (1) Transcript
+Job (1) -> (many) TranscriptRevision
 """
 
 from datetime import UTC
@@ -11,6 +13,7 @@ from typing import Optional
 from uuid import UUID
 from uuid import uuid4
 
+from sqlalchemy import UniqueConstraint
 from sqlmodel import Field
 from sqlmodel import Relationship
 from sqlmodel import SQLModel
@@ -48,6 +51,7 @@ class Job(SQLModel, table=True):
     # --- relationships ---
     document: Document = Relationship(back_populates="jobs")
     transcript: Optional["Transcript"] = Relationship(back_populates="job")
+    transcript_revisions: list["TranscriptRevision"] = Relationship(back_populates="job")
 
     @property
     def filename(self) -> str:
@@ -65,6 +69,8 @@ class Transcript(SQLModel, table=True):
     """Name of the transcription provider used to generate this transcript."""
     prompt_name: str
     """Name of the prompt used to generate this transcript."""
+    model: str | None = None
+    """Provider model that generated the original AI transcript."""
     text: str | None = None
     """The transcribed text. This may be None if the job failed or is still in progress."""
     error_detail: str | None = None
@@ -73,3 +79,23 @@ class Transcript(SQLModel, table=True):
 
     # --- relationships ---
     job: Job = Relationship(back_populates="transcript")
+
+
+class TranscriptRevision(SQLModel, table=True):
+    """Version history entries for a transcription job."""
+
+    __table_args__ = (UniqueConstraint("job_id", "version_number", name="uq_transcript_revision_job_version"),)
+
+    id: UUID = Field(default_factory=uuid4, primary_key=True)
+    job_id: UUID = Field(foreign_key="job.id", index=True)
+    version_number: int = Field(ge=1)
+    provider: str
+    prompt_name: str
+    model: str | None = None
+    source: str = Field(default="ai")
+    text: str | None = None
+    error_detail: str | None = None
+    created_at: datetime = Field(default_factory=lambda: datetime.now(UTC))
+
+    # --- relationships ---
+    job: Job = Relationship(back_populates="transcript_revisions")

src/transcription/providers/base.py

@@ -34,6 +34,7 @@ class TranscriptionResult:
             job_id=job_id,
             provider=self.provider,
             prompt_name=self.prompt_name,
+            model=self.model,
             text=self.text,
         )
 

src/transcription/services/transcription.py

@@ -18,6 +18,7 @@ from transcription.config import get_settings
 from transcription.errors import AppError
 from transcription.errors import ErrorCategory
 from transcription.models import Transcript
+from transcription.models import TranscriptRevision
 from transcription.providers import ProviderAuthError
 from transcription.providers import ProviderError
 from transcription.providers import ProviderResponseError
@@ -63,6 +64,18 @@ class TranscriptionService(ServiceBase):
             await self._finalize(session=_session, caller_session=session, refresh=(transcript,))
             return transcript
 
+    async def create_transcript_revision(
+        self,
+        transcript_revision: TranscriptRevision,
+        *,
+        session: AsyncSession | None = None,
+    ) -> TranscriptRevision:
+        """Create a new transcript revision in the database."""
+        async with self._session_scope(session) as _session:
+            _session.add(transcript_revision)
+            await self._finalize(session=_session, caller_session=session, refresh=(transcript_revision,))
+            return transcript_revision
+
     async def read_transcript(self, transcript_id: UUID, *, session: AsyncSession | None = None) -> Transcript:
         """Read an existing transcript from the database."""
         async with self._session_scope(session) as _session:
@@ -80,6 +93,27 @@ class TranscriptionService(ServiceBase):
             )
         return transcript
 
+    async def read_transcript_revision(
+        self,
+        transcript_revision_id: UUID,
+        *,
+        session: AsyncSession | None = None,
+    ) -> TranscriptRevision:
+        """Read an existing transcript revision from the database."""
+        async with self._session_scope(session) as _session:
+            transcript_revision = await _session.get(
+                TranscriptRevision,
+                transcript_revision_id,
+                options=(selectinload(TranscriptRevision.job),),  # pyright: ignore[reportArgumentType]
+            )
+        if transcript_revision is None:
+            raise TranscriptionNotFoundError(
+                f"Transcript revision with id {transcript_revision_id} not found",
+                category=ErrorCategory.NOT_FOUND,
+                suggestion="Verify the transcript revision id and retry.",
+            )
+        return transcript_revision
+
     async def update_transcript(self, transcript: Transcript, *, session: AsyncSession | None = None) -> Transcript:
         """Update an existing transcript in the database."""
         async with self._session_scope(session) as _session:
@@ -87,12 +121,35 @@ class TranscriptionService(ServiceBase):
             await self._finalize(session=_session, caller_session=session, refresh=(merged,))
             return merged
 
+    async def update_transcript_revision(
+        self,
+        transcript_revision: TranscriptRevision,
+        *,
+        session: AsyncSession | None = None,
+    ) -> TranscriptRevision:
+        """Update an existing transcript revision in the database."""
+        async with self._session_scope(session) as _session:
+            merged = await _session.merge(transcript_revision)
+            await self._finalize(session=_session, caller_session=session, refresh=(merged,))
+            return merged
+
     async def delete_transcript(self, transcript: Transcript, *, session: AsyncSession | None = None) -> None:
         """Delete a transcript from the database."""
         async with self._session_scope(session) as _session:
             await _session.delete(transcript)
             await self._finalize(session=_session, caller_session=session)
 
+    async def delete_transcript_revision(
+        self,
+        transcript_revision: TranscriptRevision,
+        *,
+        session: AsyncSession | None = None,
+    ) -> None:
+        """Delete a transcript revision from the database."""
+        async with self._session_scope(session) as _session:
+            await _session.delete(transcript_revision)
+            await self._finalize(session=_session, caller_session=session)
+
     async def transcribe_document(
         self,
         image_path: str | Path,
@@ -118,6 +175,7 @@ class TranscriptionService(ServiceBase):
         error_detail: str | None,
         provider: str | None = None,
         prompt_name: str = DEFAULT_PROMPT_FILE,
+        model: str | None = None,
         session: AsyncSession | None = None,
     ) -> Transcript:
         """Create or update a transcript for a job id."""
@@ -135,11 +193,68 @@ class TranscriptionService(ServiceBase):
             if provider is not None:
                 transcript.provider = provider
             transcript.prompt_name = prompt_name
+            if model is not None:
+                transcript.model = model
 
             _session.add(transcript)
             await self._finalize(session=_session, caller_session=session, refresh=(transcript,))
             return transcript
 
+    async def list_transcript_revisions_by_job(
+        self,
+        *,
+        job_id: UUID,
+        session: AsyncSession | None = None,
+    ) -> list[TranscriptRevision]:
+        """Return transcript revisions for a job ordered by version number."""
+        async with self._session_scope(session) as _session:
+            revisions = (
+                await _session.exec(
+                    select(TranscriptRevision)
+                    .where(TranscriptRevision.job_id == job_id)
+                    .order_by(TranscriptRevision.version_number)
+                )
+            ).all()
+            return list(revisions)
+
+    async def append_transcript_revision(
+        self,
+        *,
+        job_id: UUID,
+        text: str | None,
+        error_detail: str | None,
+        provider: str,
+        prompt_name: str,
+        model: str | None,
+        source: str,
+        session: AsyncSession | None = None,
+    ) -> TranscriptRevision:
+        """Append a new transcript revision and allocate the next version number."""
+        async with self._session_scope(session) as _session:
+            latest_version = (
+                await _session.exec(
+                    select(TranscriptRevision.version_number)
+                    .where(TranscriptRevision.job_id == job_id)
+                    .order_by(TranscriptRevision.version_number.desc())
+                    .limit(1)
+                )
+            ).first()
+            next_version = 1 if latest_version is None else latest_version + 1
+
+            revision = TranscriptRevision(
+                job_id=job_id,
+                version_number=next_version,
+                provider=provider,
+                prompt_name=prompt_name,
+                model=model,
+                source=source,
+                text=text,
+                error_detail=error_detail,
+            )
+            _session.add(revision)
+            await self._finalize(session=_session, caller_session=session, refresh=(revision,))
+            return revision
+
 
 async def transcribe_document_image(
     image_path: str | Path,

src/transcription/services/workflows.py

@@ -121,12 +121,24 @@ async def _finalize_transcribed(
     """Transaction B: transcript + TRANSCRIBED in one commit."""
     if session is None:
         async with services.jobs._session_scope() as local_session:
+            prompt_name = result.prompt_name or DEFAULT_PROMPT_FILE
             await services.transcriptions.upsert_transcript_by_job(
                 job_id=job.id,
                 text=result.text,
                 error_detail=None,
                 provider=result.provider,
-                prompt_name=result.prompt_name,
+                prompt_name=prompt_name,
+                model=result.model,
+                session=local_session,
+            )
+            await services.transcriptions.append_transcript_revision(
+                job_id=job.id,
+                text=result.text,
+                error_detail=None,
+                provider=result.provider,
+                prompt_name=prompt_name,
+                model=result.model,
+                source="ai",
                 session=local_session,
             )
             updated_job = await services.jobs.mark_job_status(
@@ -137,12 +149,24 @@ async def _finalize_transcribed(
             await local_session.commit()
             return updated_job
 
+    prompt_name = result.prompt_name or DEFAULT_PROMPT_FILE
     await services.transcriptions.upsert_transcript_by_job(
         job_id=job.id,
         text=result.text,
         error_detail=None,
         provider=result.provider,
-        prompt_name=result.prompt_name,
+        prompt_name=prompt_name,
+        model=result.model,
+        session=session,
+    )
+    await services.transcriptions.append_transcript_revision(
+        job_id=job.id,
+        text=result.text,
+        error_detail=None,
+        provider=result.provider,
+        prompt_name=prompt_name,
+        model=result.model,
+        source="ai",
         session=session,
     )
     updated_job = await services.jobs.mark_job_status(
@@ -165,11 +189,24 @@ async def _finalize_retry(
     """Transaction C: transcript error + QUEUED + retry increment in one commit."""
     if session is None:
         async with services.jobs._session_scope() as local_session:
+            provider_name = services.transcriptions.settings.provider.value
             await services.transcriptions.upsert_transcript_by_job(
                 job_id=job.id,
                 text=None,
                 error_detail=format_error_detail(error),
+                provider=provider_name,
                 prompt_name=DEFAULT_PROMPT_FILE,
+                model=services.transcriptions.settings.provider_model,
+                session=local_session,
+            )
+            await services.transcriptions.append_transcript_revision(
+                job_id=job.id,
+                text=None,
+                error_detail=format_error_detail(error),
+                provider=provider_name,
+                prompt_name=DEFAULT_PROMPT_FILE,
+                model=services.transcriptions.settings.provider_model,
+                source="ai",
                 session=local_session,
             )
             updated_job = await services.jobs.update_job_state(
@@ -180,11 +217,24 @@ async def _finalize_retry(
             )
             await local_session.commit()
     else:
+        provider_name = services.transcriptions.settings.provider.value
         await services.transcriptions.upsert_transcript_by_job(
             job_id=job.id,
             text=None,
             error_detail=format_error_detail(error),
+            provider=provider_name,
             prompt_name=DEFAULT_PROMPT_FILE,
+            model=services.transcriptions.settings.provider_model,
+            session=session,
+        )
+        await services.transcriptions.append_transcript_revision(
+            job_id=job.id,
+            text=None,
+            error_detail=format_error_detail(error),
+            provider=provider_name,
+            prompt_name=DEFAULT_PROMPT_FILE,
+            model=services.transcriptions.settings.provider_model,
+            source="ai",
             session=session,
         )
         updated_job = await services.jobs.update_job_state(
@@ -210,11 +260,24 @@ async def _finalize_failed(
     """Transaction B: transcript error + FAILED in one commit."""
     if session is None:
         async with services.jobs._session_scope() as local_session:
+            provider_name = services.transcriptions.settings.provider.value
             await services.transcriptions.upsert_transcript_by_job(
                 job_id=job.id,
                 text=None,
                 error_detail=format_error_detail(error),
+                provider=provider_name,
                 prompt_name=DEFAULT_PROMPT_FILE,
+                model=services.transcriptions.settings.provider_model,
+                session=local_session,
+            )
+            await services.transcriptions.append_transcript_revision(
+                job_id=job.id,
+                text=None,
+                error_detail=format_error_detail(error),
+                provider=provider_name,
+                prompt_name=DEFAULT_PROMPT_FILE,
+                model=services.transcriptions.settings.provider_model,
+                source="ai",
                 session=local_session,
             )
             updated_job = await services.jobs.mark_job_status(
@@ -225,11 +288,24 @@ async def _finalize_failed(
             await local_session.commit()
             return updated_job
 
+    provider_name = services.transcriptions.settings.provider.value
     await services.transcriptions.upsert_transcript_by_job(
         job_id=job.id,
         text=None,
         error_detail=format_error_detail(error),
+        provider=provider_name,
         prompt_name=DEFAULT_PROMPT_FILE,
+        model=services.transcriptions.settings.provider_model,
+        session=session,
+    )
+    await services.transcriptions.append_transcript_revision(
+        job_id=job.id,
+        text=None,
+        error_detail=format_error_detail(error),
+        provider=provider_name,
+        prompt_name=DEFAULT_PROMPT_FILE,
+        model=services.transcriptions.settings.provider_model,
+        source="ai",
         session=session,
     )
     updated_job = await services.jobs.mark_job_status(

src/transcription/ui/components/job_detail.py

@@ -2,11 +2,46 @@
 
 from __future__ import annotations
 
+import base64
+import mimetypes
+from collections.abc import Awaitable
+from collections.abc import Callable
+from dataclasses import dataclass
+from pathlib import Path
+
 from nicegui import ui
 
 from transcription.models import Document
 from transcription.models import Job
 from transcription.models import Transcript
+from transcription.models import TranscriptRevision
+
+
+@dataclass(frozen=True)
+class RevisionDisplayRow:
+    id: str
+    created: str
+    version: str
+    text: str
+    error_detail: str | None
+
+
+def _extract_row_id(args: object) -> str | None:
+    if isinstance(args, dict):
+        if isinstance(args.get("row"), dict):
+            row_id = args["row"].get("id")
+            return str(row_id) if row_id is not None else None
+        row_id = args.get("id")
+        return str(row_id) if row_id is not None else None
+
+    if isinstance(args, list):
+        for value in args:
+            if isinstance(value, dict):
+                row_id = value.get("id")
+                if row_id is not None:
+                    return str(row_id)
+
+    return None
 
 
 def _status_chip_classes(status: str) -> str:
@@ -36,7 +71,119 @@ def _render_document_section(document: Document) -> None:
             _metadata_row("File path", document.file_path)
 
 
-def _render_transcript_section(transcript: Transcript | None) -> None:
+def _document_data_url(document: Document) -> tuple[str | None, str | None]:
+    path = Path(document.file_path)
+    if not path.exists() or not path.is_file():
+        return None, "Document preview unavailable: file not found"
+
+    suffix = path.suffix.lower()
+    mime_type, _ = mimetypes.guess_type(path.name)
+    if suffix in {".tif", ".tiff"}:
+        mime_type = "image/tiff"
+    if mime_type is None:
+        return None, "Document preview unavailable: unsupported MIME type"
+
+    encoded = base64.b64encode(path.read_bytes()).decode("ascii")
+    return f"data:{mime_type};base64,{encoded}", None
+
+
+def _render_document_preview(document: Document) -> None:
+    source, error = _document_data_url(document)
+    if error is not None or source is None:
+        ui.label(error or "Document preview unavailable").classes("text-caption text-grey-7")
+        return
+
+    suffix = Path(document.file_path).suffix.lower()
+    if suffix == ".pdf":
+        ui.html(
+            (
+                '<iframe title="Document preview" '
+                f'src="{source}" '
+                'style="width:100%;height:520px;border:1px solid #ddd;border-radius:8px;"></iframe>'
+            )
+        )
+        ui.label("Zoom controls are currently available for image files.").classes("text-caption text-grey-7 q-mt-sm")
+        return
+
+    zoom_percent = {"value": 100}
+
+    with ui.element("div").style(
+        "width:100%;height:520px;overflow:auto;border:1px solid #ddd;border-radius:8px;padding:8px;background:#fafafa;"
+    ):
+        image = ui.image(source).classes("rounded-borders").style("width:100%;max-width:none;")
+
+    zoom_label = ui.label("Zoom: 100%").classes("text-caption text-grey-7 q-mt-sm")
+
+    def _apply_zoom() -> None:
+        image.style(f"width:{zoom_percent['value']}%;max-width:none;")
+        image.update()
+        zoom_label.text = f"Zoom: {zoom_percent['value']}%"
+        zoom_label.update()
+
+    def _zoom_in() -> None:
+        zoom_percent["value"] = min(300, zoom_percent["value"] + 25)
+        _apply_zoom()
+
+    def _zoom_out() -> None:
+        zoom_percent["value"] = max(50, zoom_percent["value"] - 25)
+        _apply_zoom()
+
+    def _zoom_reset() -> None:
+        zoom_percent["value"] = 100
+        _apply_zoom()
+
+    with ui.row().classes("q-gutter-sm q-mt-xs"):
+        ui.button("-", on_click=_zoom_out)
+        ui.button("+", on_click=_zoom_in)
+        ui.button("Reset", on_click=_zoom_reset)
+
+
+def _build_display_rows(transcript: Transcript, revisions: list[TranscriptRevision]) -> list[RevisionDisplayRow]:
+    ordered = sorted(revisions, key=lambda revision: revision.version_number)
+    rows: list[RevisionDisplayRow] = []
+
+    if ordered:
+        first = ordered[0]
+        rows.append(
+            RevisionDisplayRow(
+                id="original",
+                created=first.created_at.isoformat(),
+                version="original",
+                text=first.text or "",
+                error_detail=first.error_detail,
+            )
+        )
+        for revision in ordered[1:]:
+            rows.append(
+                RevisionDisplayRow(
+                    id=str(revision.version_number),
+                    created=revision.created_at.isoformat(),
+                    version=str(revision.version_number),
+                    text=revision.text or "",
+                    error_detail=revision.error_detail,
+                )
+            )
+    else:
+        rows.append(
+            RevisionDisplayRow(
+                id="original",
+                created=transcript.created_at.isoformat(),
+                version="original",
+                text=transcript.text or "",
+                error_detail=transcript.error_detail,
+            )
+        )
+
+    return rows
+
+
+def _render_transcript_versioned_section(
+    *,
+    document: Document | None,
+    transcript: Transcript | None,
+    revisions: list[TranscriptRevision],
+    on_update: Callable[[str], Awaitable[None]] | None,
+) -> None:
     with ui.card().classes("w-full q-pa-md"):
         ui.label("Transcript").classes("text-subtitle1 text-weight-medium")
         ui.separator().classes("q-my-sm")
@@ -46,24 +193,85 @@ def _render_transcript_section(transcript: Transcript | None) -> None:
             return
 
         with ui.column().classes("w-full q-gutter-y-xs"):
+            model_name = transcript.model
+            if model_name is None and revisions:
+                model_name = revisions[0].model
             _metadata_row("Provider", transcript.provider)
+            _metadata_row("Model", model_name or "unknown")
             _metadata_row("Prompt", transcript.prompt_name)
-            _metadata_row("Created", transcript.created_at.isoformat())
 
-        if transcript.text:
-            ui.separator().classes("q-my-sm")
-            with ui.card().classes("w-full bg-grey-1 q-pa-sm"):
-                ui.markdown(transcript.text)
-            return
+        display_rows = _build_display_rows(transcript, revisions)
+        rows_by_id = {row.id: row for row in display_rows}
 
-        if transcript.error_detail:
-            ui.separator().classes("q-my-sm")
-            with ui.card().classes("w-full bg-red-1 text-red-10 q-pa-sm"):
-                ui.label("Failure detail").classes("text-caption text-uppercase")
-                ui.label(transcript.error_detail).classes("text-body2")
+        ui.separator().classes("q-my-sm")
+        ui.label("Versions").classes("text-subtitle2 text-weight-medium")
+        table = ui.table(
+            columns=[
+                {"name": "created", "label": "Created", "field": "created", "align": "left"},
+                {"name": "version", "label": "Version", "field": "version", "align": "left"},
+            ],
+            rows=[
+                {
+                    "id": row.id,
+                    "created": row.created,
+                    "version": row.version,
+                }
+                for row in display_rows
+            ],
+            row_key="id",
+        ).classes("w-full")
+
+        default_selected = display_rows[-1].id
+        selected_label = ui.label(f"Selected version: {rows_by_id[default_selected].version}").classes(
+            "text-caption text-grey-7"
+        )
+
+        ui.separator().classes("q-my-sm")
+        with ui.row().classes("w-full no-wrap items-start q-gutter-md"):
+            if document is not None:
+                with ui.column().classes("w-1/2"):
+                    ui.label("Document Preview").classes("text-subtitle2 text-weight-medium")
+                    _render_document_preview(document)
+
+            with ui.column().classes("w-1/2"):
+                editor = (
+                    ui.textarea(label="Transcript text", value=rows_by_id[default_selected].text)
+                    .props("autogrow outlined")
+                    .classes("w-full")
+                )
+                error_label = ui.label("").classes("text-body2 text-red-10")
+
+        def _set_selected(version_id: str) -> None:
+            selected = rows_by_id.get(version_id)
+            if selected is None:
+                return
+            selected_label.text = f"Selected version: {selected.version}"
+            editor.value = selected.text
+            editor.update()
+            error_label.text = selected.error_detail or ""
+            error_label.update()
+
+        def _on_row_click(event) -> None:  # noqa: ANN001
+            row_id = _extract_row_id(event.args)
+            if row_id is None:
+                return
+            _set_selected(row_id)
+
+        table.on("rowClick", _on_row_click)
+        _set_selected(default_selected)
+
+        if on_update is not None:
+            ui.button("Update", on_click=lambda: on_update(editor.value or ""))
 
 
-def render_job_detail(*, job: Job, document: Document | None, transcript: Transcript | None) -> None:
+def render_job_detail(
+    *,
+    job: Job,
+    document: Document | None,
+    transcript: Transcript | None,
+    revisions: list[TranscriptRevision],
+    on_update: Callable[[str], Awaitable[None]] | None = None,
+) -> None:
     """Render all sections for the job detail page."""
     status_text = job.status.value
     with ui.column().classes("w-full max-w-4xl q-gutter-md"):
@@ -88,4 +296,9 @@ def render_job_detail(*, job: Job, document: Document | None, transcript: Transc
         if document is not None:
             _render_document_section(document)
 
-        _render_transcript_section(transcript)
+        _render_transcript_versioned_section(
+            document=document,
+            transcript=transcript,
+            revisions=revisions,
+            on_update=on_update,
+        )

src/transcription/ui/pages/jobs_page.py

@@ -13,6 +13,8 @@ from transcription.db import get_session
 from transcription.models import Document
 from transcription.models import Job
 from transcription.models import Transcript
+from transcription.models import TranscriptRevision
+from transcription.services import ServiceBundle
 from transcription.ui.components.error_presenter import show_error
 from transcription.ui.components.error_presenter import summarize_error
 from transcription.ui.components.job_detail import render_job_detail
@@ -43,15 +45,22 @@ async def fetch_job_rows() -> list[JobTableRow]:
         ]
 
 
-async def fetch_job_detail(job_id: UUID) -> tuple[Job | None, Document | None, Transcript | None]:
-    """Return job, document, and transcript for detail view."""
+async def fetch_job_detail(job_id: UUID) -> tuple[Job | None, Document | None, Transcript | None, list[TranscriptRevision]]:
+    """Return job, document, transcript snapshot, and revisions for detail view."""
     async with get_session() as session:
         job = await session.get(Job, job_id)
         if job is None:
-            return None, None, None
+            return None, None, None, []
         document = await session.get(Document, job.document_id)
         transcript = (await session.exec(select(Transcript).where(Transcript.job_id == job.id))).first()
-        return job, document, transcript
+        revisions = (
+            await session.exec(
+                select(TranscriptRevision)
+                .where(TranscriptRevision.job_id == job.id)
+                .order_by(TranscriptRevision.version_number)
+            )
+        ).all()
+        return job, document, transcript, list(revisions)
 
 
 def register_page() -> None:
@@ -83,6 +92,7 @@ def register_page() -> None:
     @ui.page("/jobs/{job_id}")
     async def job_detail_page(job_id: str) -> None:
         ui.label("Job Detail")
+        content = ui.column().classes("w-full")
         try:
             parsed_id = UUID(job_id)
         except ValueError:
@@ -90,12 +100,64 @@ def register_page() -> None:
             ui.link("Back to jobs", "/jobs")
             return
 
-        job, document, transcript = await fetch_job_detail(parsed_id)
-        if job is None:
-            ui.label("Job not found")
-            ui.link("Back to jobs", "/jobs")
-            return
+        async def refresh_content() -> None:
+            content.clear()
+            job, document, transcript, revisions = await fetch_job_detail(parsed_id)
+            if job is None:
+                with content:
+                    ui.label("Job not found")
+                return
 
-        render_job_detail(job=job, document=document, transcript=transcript)
+            services = ServiceBundle()
+
+            async def update_transcript_text(value: str) -> None:
+                try:
+                    update_text = value.strip()
+                    async with get_session() as session:
+                        current_transcript = (
+                            await session.exec(select(Transcript).where(Transcript.job_id == parsed_id))
+                        ).first()
+                        provider_name = current_transcript.provider if current_transcript is not None else "openrouter"
+                        prompt_name = (
+                            current_transcript.prompt_name if current_transcript is not None else "transcribe_document.md"
+                        )
+                        model_name = current_transcript.model if current_transcript is not None else None
+
+                        await services.transcriptions.upsert_transcript_by_job(
+                            job_id=parsed_id,
+                            text=update_text,
+                            error_detail=None,
+                            provider=provider_name,
+                            prompt_name=prompt_name,
+                            model=model_name,
+                            session=session,
+                        )
+                        await services.transcriptions.append_transcript_revision(
+                            job_id=parsed_id,
+                            text=update_text,
+                            error_detail=None,
+                            provider=provider_name,
+                            prompt_name=prompt_name,
+                            model=model_name,
+                            source="user",
+                            session=session,
+                        )
+                        await session.commit()
+
+                    ui.notify("Transcript updated", type="positive")
+                    await refresh_content()
+                except Exception as exc:  # noqa: BLE001
+                    show_error(exc, title="Transcript update failed", operation="jobs.detail.update")
+
+            with content:
+                render_job_detail(
+                    job=job,
+                    document=document,
+                    transcript=transcript,
+                    revisions=revisions,
+                    on_update=update_transcript_text,
+                )
+
+        await refresh_content()
 
         ui.link("Back to jobs", "/jobs")

tests/integration/test_pipeline_flow.py

@@ -6,9 +6,9 @@ import pytest
 from sqlmodel import select
 
 from transcription.config import Settings
-from transcription.models import Job, JobStatus, Transcript
+from transcription.models import Job, JobStatus, Transcript, TranscriptRevision
 from transcription.providers.base import TranscriptionResult
-from transcription.services.upload import create_upload_job
+from transcription.services.store import create_upload_job
 from transcription.worker import process_next_queued_job
 
 
@@ -16,24 +16,37 @@ from transcription.worker import process_next_queued_job
 class TestPipelineSuccessFlow:
     """Verify end-to-end success lifecycle behavior."""
 
-    def test_upload_then_worker_persists_transcribed_terminal_state(self, session, tmp_path: Path, monkeypatch):
+    @pytest.mark.asyncio
+    async def test_upload_then_worker_persists_transcribed_terminal_state(self, async_session, tmp_path: Path, monkeypatch):
         """Upload followed by worker processing persists transcript and transcribed status."""
         settings = Settings(openrouter_api_key="test-key", upload_dir=tmp_path)
-        upload_result = create_upload_job(
+        upload_result = await create_upload_job(
             filename="pipeline.jpg",
             file_bytes=b"pipeline-bytes",
-            session=session,
+            session=async_session,
             settings=settings,
         )
 
-        def _fake_transcribe(_path: str) -> TranscriptionResult:
-            return TranscriptionResult(text="Pipeline transcript", provider="openrouter", model="test-model")
+        async def _fake_transcribe(_path: str) -> TranscriptionResult:
+            return TranscriptionResult(
+                text="Pipeline transcript",
+                provider="openrouter",
+                prompt_name="transcribe_document.md",
+                model="test-model",
+            )
 
-        monkeypatch.setattr("transcription.worker.transcribe_document_image", _fake_transcribe)
+        monkeypatch.setattr("transcription.services.workflows.transcribe_document_image", _fake_transcribe)
 
-        processed = process_next_queued_job(session=session)
-        job = session.get(Job, upload_result.job_id)
-        transcript = session.exec(select(Transcript).where(Transcript.job_id == upload_result.job_id)).first()
+        processed = await process_next_queued_job(session=async_session)
+        job = await async_session.get(Job, upload_result.job_id)
+        transcript = (await async_session.exec(select(Transcript).where(Transcript.job_id == upload_result.job_id))).first()
+        revisions = (
+            await async_session.exec(
+                select(TranscriptRevision)
+                .where(TranscriptRevision.job_id == upload_result.job_id)
+                .order_by(TranscriptRevision.version_number)
+            )
+        ).all()
 
         assert processed is True
         assert job is not None
@@ -41,30 +54,43 @@ class TestPipelineSuccessFlow:
         assert transcript is not None
         assert transcript.text == "Pipeline transcript"
         assert transcript.error_detail is None
+        assert transcript.model == "test-model"
+        assert len(revisions) == 1
+        assert revisions[0].version_number == 1
+        assert revisions[0].source == "ai"
+        assert revisions[0].text == "Pipeline transcript"
 
 
 @pytest.mark.integration
 class TestPipelineFailureFlow:
     """Verify end-to-end failure lifecycle behavior."""
 
-    def test_upload_then_worker_persists_failed_terminal_state(self, session, tmp_path: Path, monkeypatch):
+    @pytest.mark.asyncio
+    async def test_upload_then_worker_persists_failed_terminal_state(self, async_session, tmp_path: Path, monkeypatch):
         """Upload followed by worker processing persists error detail and failed status."""
         settings = Settings(openrouter_api_key="test-key", upload_dir=tmp_path)
-        upload_result = create_upload_job(
+        upload_result = await create_upload_job(
             filename="pipeline.jpg",
             file_bytes=b"pipeline-bytes",
-            session=session,
+            session=async_session,
             settings=settings,
         )
 
-        def _fake_transcribe(_path: str) -> TranscriptionResult:
+        async def _fake_transcribe(_path: str) -> TranscriptionResult:
             raise RuntimeError("pipeline provider failure")
 
-        monkeypatch.setattr("transcription.worker.transcribe_document_image", _fake_transcribe)
+        monkeypatch.setattr("transcription.services.workflows.transcribe_document_image", _fake_transcribe)
 
-        processed = process_next_queued_job(session=session)
-        job = session.get(Job, upload_result.job_id)
-        transcript = session.exec(select(Transcript).where(Transcript.job_id == upload_result.job_id)).first()
+        processed = await process_next_queued_job(session=async_session)
+        job = await async_session.get(Job, upload_result.job_id)
+        transcript = (await async_session.exec(select(Transcript).where(Transcript.job_id == upload_result.job_id))).first()
+        revisions = (
+            await async_session.exec(
+                select(TranscriptRevision)
+                .where(TranscriptRevision.job_id == upload_result.job_id)
+                .order_by(TranscriptRevision.version_number)
+            )
+        ).all()
 
         assert processed is True
         assert job is not None
@@ -74,3 +100,8 @@ class TestPipelineFailureFlow:
         assert "pipeline provider failure" in transcript.error_detail
         assert "[internal_unexpected_error]" in transcript.error_detail
         assert "error_id=" in transcript.error_detail
+        assert len(revisions) == 1
+        assert revisions[0].version_number == 1
+        assert revisions[0].source == "ai"
+        assert revisions[0].text is None
+        assert "pipeline provider failure" in (revisions[0].error_detail or "")

tests/test_db.py

@@ -1,96 +1,71 @@
-"""Tests for transcription.db — schema bootstrap and session factory."""
+"""Tests for transcription.db — async schema bootstrap/runtime behavior."""
 
-from sqlalchemy import inspect, text
-from sqlmodel import Session, SQLModel, create_engine
-from sqlmodel.pool import StaticPool
-
-
-def _in_memory_engine():
-    """Create a fresh in-memory SQLite engine for isolated db tests."""
-    return create_engine(
-        "sqlite://",
-        connect_args={"check_same_thread": False},
-        poolclass=StaticPool,
-    )
+from sqlalchemy import inspect
+from sqlalchemy import text
+import pytest
 
 
 class TestSchemaBootstrap:
-    """Verify create_all produces the expected table set."""
+    """Verify async create_all produces the expected table set."""
 
-    def test_create_all_creates_expected_tables(self):
-        """After create_all(), document, job, and transcript tables exist."""
-        engine = _in_memory_engine()
-        # Ensure models are imported so metadata is populated
-        from transcription.models import Document, Job, Transcript  # noqa: F401
+    @pytest.mark.asyncio
+    async def test_create_all_creates_expected_tables(self, default_settings):
+        """After async create_all(), document/job/transcript/revision tables exist."""
+        # Ensure models are imported so metadata is populated.
+        from transcription.models import Document, Job, Transcript, TranscriptRevision  # noqa: F401
 
-        import transcription.db as db_module
+        from transcription.db.operations import create_all
+        from transcription.db.runtime import get_engine
 
-        db_module.create_all(engine=engine)
+        engine = get_engine(settings=default_settings)
+        await create_all(engine=engine)
+
+        async with engine.begin() as connection:
+            table_names = set(await connection.run_sync(lambda sync_conn: inspect(sync_conn).get_table_names()))
 
-        inspector = inspect(engine)
-        table_names = set(inspector.get_table_names())
         assert "document" in table_names
         assert "job" in table_names
         assert "transcript" in table_names
+        assert "transcriptrevision" in table_names
 
 
 class TestSessionFactory:
-    """Verify get_session yields and cleans up sessions."""
+    """Verify async get_session yields a usable AsyncSession."""
 
-    def test_get_session_yields_session(self):
-        """get_session() yields a usable Session object."""
-        engine = _in_memory_engine()
-        SQLModel.metadata.create_all(engine)
+    @pytest.mark.asyncio
+    async def test_get_session_yields_session(self, default_settings):
+        """get_session() yields an AsyncSession with a live connection."""
+        from transcription.db.runtime import get_session
 
-        import transcription.db as db_module
-
-        with db_module.get_session(engine=engine) as session:
-            assert isinstance(session, Session)
-
-    def test_session_is_closed_after_generator_exit(self):
-        """After the context manager exits, the session is closed."""
-        engine = _in_memory_engine()
-        SQLModel.metadata.create_all(engine)
-
-        import transcription.db as db_module
-
-        with db_module.get_session(engine=engine) as session:
-            # Session is usable inside the context
-            session.execute(text("SELECT 1"))
-            captured = session
-
-        # After exiting, the session's internal connection is released
-        # (no active transaction bound to the session)
-        assert captured._transaction is None
+        async with get_session(settings=default_settings) as session:
+            result = await session.exec(text("SELECT 1"))
+            assert result.first()[0] == 1
 
 
 class TestBootstrapPolicy:
-    """Verify schema bootstrap policy defaults and overrides."""
+    """Verify startup schema bootstrap policy via Settings property."""
 
     def test_production_defaults_to_no_bootstrap(self):
         """Production defaults to explicit non-bootstrap startup behavior."""
         from transcription.config import Settings
-        from transcription.db import should_bootstrap_schema
 
         settings = Settings(openrouter_api_key="test-key", environment="production")
-        assert should_bootstrap_schema(settings) is False
+        assert settings.should_bootstrap_schema is False
 
     def test_development_defaults_to_bootstrap(self):
         """Development defaults to schema bootstrap for local workflows."""
         from transcription.config import Settings
-        from transcription.db import should_bootstrap_schema
 
         settings = Settings(openrouter_api_key="test-key", environment="development")
-        assert should_bootstrap_schema(settings) is True
+        assert settings.should_bootstrap_schema is True
 
     def test_explicit_override_wins(self):
         """Explicit bootstrap_schema_on_startup overrides environment default."""
         from transcription.config import Settings
-        from transcription.db import should_bootstrap_schema
 
         settings = Settings(
             openrouter_api_key="test-key",
             environment="production",
             bootstrap_schema_on_startup=True,
         )
-        assert should_bootstrap_schema(settings) is True
+        assert settings.should_bootstrap_schema is True

tests/test_models.py

@@ -1,11 +1,11 @@
-"""Tests for transcription.models — Document, Job, Transcript persistence and relationships."""
+"""Tests for transcription.models — Document, Job, Transcript, TranscriptRevision models."""
 
 from uuid import UUID
 
 import pytest
 from sqlalchemy.exc import IntegrityError
 
-from transcription.models import Document, Job, JobStatus, Transcript
+from transcription.models import Document, Job, JobStatus, Transcript, TranscriptRevision
 
 
 def _make_document(**overrides) -> Document:
@@ -113,7 +113,7 @@ class TestTranscriptModel:
         """A Transcript with text set and error_detail None persists correctly."""
         doc = _persist_document(session)
         job = _persist_job(session, doc)
-        transcript = Transcript(job_id=job.id, text="Dear Sir, ...")
+        transcript = Transcript(job_id=job.id, provider="openrouter", prompt_name="transcribe_document.md", text="Dear Sir, ...")
         session.add(transcript)
         session.commit()
         session.refresh(transcript)
@@ -127,7 +127,12 @@ class TestTranscriptModel:
         """A Transcript with text None and error_detail set persists correctly."""
         doc = _persist_document(session)
         job = _persist_job(session, doc)
-        transcript = Transcript(job_id=job.id, error_detail="Provider timeout")
+        transcript = Transcript(
+            job_id=job.id,
+            provider="openrouter",
+            prompt_name="transcribe_document.md",
+            error_detail="Provider timeout",
+        )
         session.add(transcript)
         session.commit()
         session.refresh(transcript)
@@ -142,16 +147,101 @@ class TestTranscriptModel:
         doc = _persist_document(session)
         job = _persist_job(session, doc)
 
-        t1 = Transcript(job_id=job.id, text="First")
+        t1 = Transcript(job_id=job.id, provider="openrouter", prompt_name="transcribe_document.md", text="First")
         session.add(t1)
         session.commit()
 
-        t2 = Transcript(job_id=job.id, text="Duplicate")
+        t2 = Transcript(job_id=job.id, provider="openrouter", prompt_name="transcribe_document.md", text="Duplicate")
         session.add(t2)
         with pytest.raises(IntegrityError):
             session.commit()
 
 
+class TestTranscriptRevisionModel:
+    """Verify TranscriptRevision persistence and version uniqueness constraints."""
+
+    def test_revision_record_persists(self, session):
+        """A TranscriptRevision with version metadata persists correctly."""
+        doc = _persist_document(session)
+        job = _persist_job(session, doc)
+        revision = TranscriptRevision(
+            job_id=job.id,
+            version_number=1,
+            provider="openrouter",
+            prompt_name="transcribe_document.md",
+            model="google/gemini-2.5-flash",
+            source="ai",
+            text="Initial text",
+        )
+        session.add(revision)
+        session.commit()
+        session.refresh(revision)
+
+        fetched = session.get(TranscriptRevision, revision.id)
+        assert fetched is not None
+        assert fetched.version_number == 1
+        assert fetched.text == "Initial text"
+        assert fetched.source == "ai"
+
+    def test_job_version_pair_is_unique(self, session):
+        """Duplicate version_number for same job raises integrity error."""
+        doc = _persist_document(session)
+        job = _persist_job(session, doc)
+
+        first = TranscriptRevision(
+            job_id=job.id,
+            version_number=1,
+            provider="openrouter",
+            prompt_name="transcribe_document.md",
+            source="ai",
+            text="Initial",
+        )
+        duplicate = TranscriptRevision(
+            job_id=job.id,
+            version_number=1,
+            provider="openrouter",
+            prompt_name="transcribe_document.md",
+            source="user",
+            text="Edited",
+        )
+        session.add(first)
+        session.commit()
+
+        session.add(duplicate)
+        with pytest.raises(IntegrityError):
+            session.commit()
+
+    def test_same_version_number_allowed_for_different_jobs(self, session):
+        """Version numbers are scoped per job, not globally."""
+        doc1 = _persist_document(session)
+        job1 = _persist_job(session, doc1)
+        doc2 = _make_document(filename="letter2.jpg", file_path="/uploads/letter2.jpg")
+        session.add(doc2)
+        session.commit()
+        session.refresh(doc2)
+        job2 = _persist_job(session, doc2)
+
+        r1 = TranscriptRevision(
+            job_id=job1.id,
+            version_number=1,
+            provider="openrouter",
+            prompt_name="transcribe_document.md",
+            source="ai",
+            text="Job1 v1",
+        )
+        r2 = TranscriptRevision(
+            job_id=job2.id,
+            version_number=1,
+            provider="openrouter",
+            prompt_name="transcribe_document.md",
+            source="ai",
+            text="Job2 v1",
+        )
+        session.add(r1)
+        session.add(r2)
+        session.commit()
+
+
 class TestRelationships:
     """Verify SQLModel relationship navigation between models."""
 
@@ -169,7 +259,12 @@ class TestRelationships:
         """job.transcript returns the linked Transcript."""
         doc = _persist_document(session)
         job = _persist_job(session, doc)
-        transcript = Transcript(job_id=job.id, text="Transcribed text")
+        transcript = Transcript(
+            job_id=job.id,
+            provider="openrouter",
+            prompt_name="transcribe_document.md",
+            text="Transcribed text",
+        )
         session.add(transcript)
         session.commit()
 
@@ -177,3 +272,32 @@ class TestRelationships:
         assert job.transcript is not None
         assert isinstance(job.transcript, Transcript)
         assert job.transcript.text == "Transcribed text"
+
+    def test_job_exposes_transcript_revisions(self, session):
+        """job.transcript_revisions returns all linked revisions."""
+        doc = _persist_document(session)
+        job = _persist_job(session, doc)
+        session.add(
+            TranscriptRevision(
+                job_id=job.id,
+                version_number=1,
+                provider="openrouter",
+                prompt_name="transcribe_document.md",
+                source="ai",
+                text="v1",
+            )
+        )
+        session.add(
+            TranscriptRevision(
+                job_id=job.id,
+                version_number=2,
+                provider="openrouter",
+                prompt_name="transcribe_document.md",
+                source="user",
+                text="v2",
+            )
+        )
+        session.commit()
+
+        session.refresh(job)
+        assert len(job.transcript_revisions) == 2