diff --git a/docs/step4.md b/docs/step4.md
index 20810d2..45a0aa9 100644
--- a/docs/step4.md
+++ b/docs/step4.md
@@ -1,18 +1,18 @@
 ## Step 4: `services/upload.py` + `worker.py`
 
-## Objective
+### Objective
 
-Implement the MVP upload and background processing pipeline so the system can:
+Implement the MVP upload and background-processing pipeline so the system can:
 
 1. Save uploaded files into `UPLOAD_DIR`
 2. Create `Document` + `Job(status="queued")`
 3. Process queued jobs in a worker loop:
    - `queued -> processing`
-   - invoke Step 3 transcription
+   - call Step 3 transcription service
    - persist `Transcript`
    - finalize as `transcribed` or `failed`
 
-This covers MVP Feature 1 + 2 and supports REQ-1, REQ-2, REQ-3, REQ-4, REQ-6.
+This step advances MVP Feature 1 + Feature 2 and supports REQ-1, REQ-2, REQ-3, REQ-4, REQ-6.
 
 ---
 
@@ -21,15 +21,15 @@ This covers MVP Feature 1 + 2 and supports REQ-1, REQ-2, REQ-3, REQ-4, REQ-6.
 ### In scope
 - `src/transcription/services/upload.py`
 - `src/transcription/worker.py`
-- Upload persistence logic + job creation
-- Worker polling and single-job processing lifecycle
-- Tests for upload + worker behavior (deterministic, no live API)
+- Upload persistence logic and initial job creation
+- Worker polling and single-job lifecycle execution
+- Deterministic test coverage for upload + worker (default suite)
 
 ### Out of scope
-- UI wiring (Step 5)
-- Job/transcript pages (Step 5)
-- External queue infrastructure
-- Async DB/session architecture changes (post-MVP)
+- UI integration and pages (Step 5)
+- Queue infrastructure beyond in-process loop
+- Async DB/session architecture refactor
+- Broad production hardening beyond MVP needs
 
 ---
 
@@ -40,151 +40,223 @@ This covers MVP Feature 1 + 2 and supports REQ-1, REQ-2, REQ-3, REQ-4, REQ-6.
 - `src/transcription/worker.py`
 - `src/transcription/services/__init__.py` (export updates as needed)
 
-### Tests
+### Test files
 - `tests/services/test_upload.py`
-- `tests/services/test_worker.py` *(recommended for mirror consistency)*
+- `tests/services/test_worker.py`
+
+### Optional external lane (already present pattern)
+- reuse `external` marker for live-provider checks where appropriate
+- keep external out of default lane
+
+---
+
+## Required MCP Prompt References (for test workflow)
+
+Apply these resources directly during Step 4 test creation:
+
+1. `resource://catalog/prompts/pytest-scaffold`
+2. `resource://prompts/pytest-scaffold/document`
+3. `resource://catalog/prompts/pytest-fill-scaffold`
+4. `resource://prompts/pytest-fill-scaffold/document`
+
+And (as referenced by those prompts) apply relevant pytest skill references for:
+- naming/hierarchy
+- marker defaults
+- SQLAlchemy sync testing behavior where applicable
 
 ---
 
 ## Design Decisions
 
-1. **Upload service owns file persistence + initial DB records**
-   - Save file first, then create `Document` + `Job(queued)`.
+1. **Upload service owns initial file + record creation**
+   - Writes file, creates `Document`, creates queued `Job`, returns IDs/path.
 
-2. **Worker owns job lifecycle transitions**
-   - Transition only in worker:
-     - `queued -> processing`
-     - `processing -> transcribed|failed`
+2. **Worker owns lifecycle transitions**
+   - Worker is the single owner of `queued -> processing -> terminal` job state changes.
 
-3. **Worker calls Step 3 service, not provider SDK directly**
-   - Use `transcribe_document_image(...)` from `services/transcription.py`.
+3. **Worker uses Step 3 service boundary**
+   - Worker calls `transcribe_document_image(...)`; no provider-specific SDK logic in worker.
 
-4. **Failure details are always persisted**
-   - On failure, persist transcript row with `text=None`, `error_detail=...`.
+4. **Failure information is always persisted**
+   - On failure: store `Transcript(text=None, error_detail=...)` and set `Job.status=failed`.
 
-5. **Worker loop is stoppable**
-   - Simple in-process loop with stop signal/event and poll interval.
+5. **Loop remains simple and stoppable**
+   - In-process polling loop with stop event and poll interval for MVP simplicity and testability.
 
 ---
 
 ## Task-by-Task Execution Checklist
 
-## Phase A — Upload service (`services/upload.py`)
+## Phase A — Implement upload service (`src/transcription/services/upload.py`)
 
-- [ ] Create `src/transcription/services/upload.py`
-- [ ] Add `UploadError` exception type
-- [ ] Add result dataclass (e.g., `UploadJobResult`) with:
+- [ ] Create `UploadError` exception
+- [ ] Create `UploadJobResult` dataclass with:
   - [ ] `document_id`
   - [ ] `job_id`
   - [ ] `stored_path`
   - [ ] `original_filename`
-- [ ] Implement safe filename handling:
-  - [ ] basename normalization
-  - [ ] collision-safe naming (e.g., UUID prefix)
-  - [ ] reject path traversal patterns
-- [ ] Implement extension validation (`.jpg/.jpeg/.png/.tif/.tiff/.pdf`)
-- [ ] Implement empty-bytes validation
-- [ ] Ensure `UPLOAD_DIR` exists (`mkdir(parents=True, exist_ok=True)`)
-- [ ] Persist file bytes to disk
-- [ ] Insert `Document` and `Job(status=queued)` in one transaction
-- [ ] Return structured `UploadJobResult`
-- [ ] Add logging for upload success/failure (no sensitive payload logging)
+- [ ] Add filename safety handling:
+  - [ ] normalize to basename
+  - [ ] avoid path traversal
+  - [ ] collision-safe stored name (e.g., UUID prefix/suffix)
+- [ ] Validate upload payload:
+  - [ ] non-empty bytes required
+  - [ ] extension in supported set (`.jpg/.jpeg/.png/.tif/.tiff/.pdf`)
+- [ ] Ensure upload directory exists (`mkdir(parents=True, exist_ok=True)`)
+- [ ] Write file bytes to `UPLOAD_DIR`
+- [ ] Persist DB records in one transaction:
+  - [ ] `Document(filename, file_path)`
+  - [ ] `Job(document_id=..., status=queued)`
+- [ ] Return `UploadJobResult`
+- [ ] Add logging for success/failure boundaries
 
 ---
 
-## Phase B — Worker core (`worker.py`)
+## Phase B — Implement worker core (`src/transcription/worker.py`)
 
-- [ ] Create `src/transcription/worker.py`
-- [ ] Implement `process_next_queued_job(...) -> bool`
-  - [ ] fetch oldest queued job
-  - [ ] return `False` if none found
-  - [ ] set job to `processing` + update timestamp
-  - [ ] call `transcribe_document_image(image_path=...)`
-  - [ ] success path: persist transcript text, mark `transcribed`
-  - [ ] failure path: persist error detail, mark `failed`
-  - [ ] return `True` when a job is processed
-- [ ] Implement transcript upsert behavior (avoid duplicate unique `job_id` insert issues)
-- [ ] Add structured logs for lifecycle transitions and outcomes
+- [ ] Add `process_next_queued_job(...) -> bool`
+- [ ] Fetch oldest queued job
+- [ ] Return `False` when no queued jobs exist
+- [ ] Transition picked job to `processing` and update timestamp
+- [ ] Resolve associated `Document.file_path`
+- [ ] Call `transcribe_document_image(image_path=...)`
+- [ ] On success:
+  - [ ] insert/update transcript text
+  - [ ] clear error detail
+  - [ ] mark job `transcribed`
+  - [ ] update timestamp
+- [ ] On failure:
+  - [ ] insert/update transcript with `text=None`, `error_detail=...`
+  - [ ] mark job `failed`
+  - [ ] update timestamp
+- [ ] Commit terminal state and return `True`
+- [ ] Add logs around job pickup, transition, and terminal outcome
 
 ---
 
-## Phase C — Worker loop (`worker.py`)
+## Phase C — Implement worker loop (`src/transcription/worker.py`)
 
-- [ ] Implement `run_worker_loop(...)`
-  - [ ] accepts stop signal/event
-  - [ ] configurable `poll_interval_seconds`
-  - [ ] repeatedly calls `process_next_queued_job`
-  - [ ] sleeps when no work
-  - [ ] exits cleanly on stop request
-- [ ] Keep implementation single-process/simple for MVP assumptions
+- [ ] Add `run_worker_loop(...)`
+- [ ] Accept configurable stop event/signal
+- [ ] Accept configurable poll interval
+- [ ] Repeatedly call `process_next_queued_job`
+- [ ] Sleep only when queue is empty
+- [ ] Exit cleanly when stop event is set
 
 ---
 
-## Phase D — Service exports
+## Phase D — Exports
 
-- [ ] Update `src/transcription/services/__init__.py` exports
-  - [ ] include upload service symbols
-  - [ ] keep transcription exports intact
+- [ ] Update `src/transcription/services/__init__.py` to expose upload APIs
+- [ ] Keep existing transcription exports intact
 
 ---
 
-## Phase E — Tests (scaffold -> fill)
+## Phase E — Tests via MCP scaffold -> fill flow
 
-### E1 Scaffold
-- [ ] Create `tests/services/test_upload.py` with class/method skeletons + one-line docstrings
-- [ ] Create `tests/services/test_worker.py` with class/method skeletons + one-line docstrings
-- [ ] Validate scaffold collection:
-  - [ ] `uv run pytest --collect-only -q`
+## E1 Scaffold (structure only)
 
-### E2 Fill
+Use scaffold prompt workflow first for:
+- `src/transcription/services/upload.py`
+- `src/transcription/worker.py`
 
-#### Upload tests
-- [ ] `test_create_upload_job_writes_file_and_creates_records` (`integration`)
-- [ ] `test_create_upload_job_rejects_empty_bytes` (`unit`)
-- [ ] `test_create_upload_job_rejects_unsupported_extension` (`unit`)
-- [ ] `test_create_upload_job_uses_unique_stored_filename` (`integration`)
-- [ ] `test_create_upload_job_sets_job_status_queued` (`integration`)
+Expected scaffold targets:
+- `tests/services/test_upload.py`
+- `tests/services/test_worker.py`
 
-#### Worker tests
-- [ ] `test_process_next_queued_job_returns_false_when_queue_empty` (`integration`)
-- [ ] `test_process_next_queued_job_transitions_processing_to_transcribed` (`integration`)
-- [ ] `test_process_next_queued_job_persists_transcript_text_on_success` (`integration`)
-- [ ] `test_process_next_queued_job_sets_failed_and_error_detail_on_failure` (`integration`)
-- [ ] `test_process_next_queued_job_updates_existing_transcript_if_present` (`integration`)
-- [ ] `test_run_worker_loop_stops_when_stop_event_is_set` (`unit`)
+Scaffold rules:
+- [ ] Class hierarchy + method names + one-line docstrings only
+- [ ] No assertions or implementation details in scaffold phase
+- [ ] Keep method names concise and behavior-focused
 
-### Marker usage
-- [ ] Mark pure logic tests as `@pytest.mark.unit`
-- [ ] Mark DB/service orchestration tests as `@pytest.mark.integration`
-- [ ] Reuse existing `external` only for future live endpoint checks (not required in Step 4)
+Validation:
+- [ ] `uv run pytest --collect-only -q`
+
+## E2 Fill scaffold (implementation)
+
+Use fill prompt workflow for:
+- `tests/services/test_upload.py`
+- `tests/services/test_worker.py`
+- stack: `sqlalchemy-sync` (or `mixed` if combining pure + DB behaviors)
+- marker lane preference: `unit` and `integration` as appropriate
+- strategy: minimal deterministic implementation
+
+Fill rules (invariants):
+- [ ] Preserve scaffold class names, method names, and one-line docstrings
+- [ ] Do not rename/re-nest scaffolded tests unless explicitly approved
+- [ ] One behavior target per test
+- [ ] Minimal mocking; mock only network/nondeterministic boundaries
+
+Suggested test coverage:
+
+### `tests/services/test_upload.py`
+- [ ] creates file + document + queued job (`integration`)
+- [ ] rejects empty bytes (`unit`)
+- [ ] rejects unsupported extension (`unit`)
+- [ ] writes collision-safe unique filename (`integration`)
+- [ ] persisted job status is `queued` (`integration`)
+
+### `tests/services/test_worker.py`
+- [ ] returns `False` when queue empty (`integration`)
+- [ ] transitions `queued -> processing -> transcribed` on success (`integration`)
+- [ ] stores transcript text on success (`integration`)
+- [ ] transitions to `failed` and stores `error_detail` on failure (`integration`)
+- [ ] updates existing transcript instead of duplicate create (`integration`)
+- [ ] worker loop exits when stop event set (`unit`)
 
 ---
 
-## Phase F — Verification
+## Marker Strategy
+
+- `unit`: pure logic tests (filename handling, loop stop behavior, validation logic)
+- `integration`: DB + service orchestration tests (SQLite/session/contracts)
+- `external`: opt-in live provider tests only (not part of default Step 4 lane)
+
+No new marker needed; reuse existing marker registration.
+
+---
+
+## Validation Sequence (strict order)
 
 - [ ] `uv run pytest --collect-only -q`
+- [ ] `uv run pytest -m unit -q` *(if unit tests touched)*
 - [ ] `uv run pytest tests/services/test_upload.py -q`
 - [ ] `uv run pytest tests/services/test_worker.py -q`
-- [ ] `uv run pytest -m "unit or integration" -q`
 - [ ] `uv run pytest -q`
 
 ---
 
-## Implementation Guardrails
+## Reporting Requirements (after implementation)
 
-- [ ] Do not couple worker directly to provider SDK internals
-- [ ] Do not swallow exceptions silently
-- [ ] Always persist terminal failure details
-- [ ] Keep status transitions explicit and timestamped
-- [ ] Keep default test suite deterministic and fast (no live network)
+Implementation report must include:
+
+1. Files created/updated
+2. Fixture and marker decisions
+3. MCP references used and why
+4. Validation command results
+5. Remaining risks/open questions (only blockers)
+
+---
+
+## Guardrails
+
+- Keep Step 4 independent from UI concerns.
+- Do not call provider SDK directly from worker.
+- Do not silently swallow exceptions.
+- Always persist terminal job outcome.
+- Keep default suite deterministic and fast.
+- Preserve scaffold invariants during fill phase.
 
 ---
 
 ## Definition of Done (Step 4)
 
-- [ ] Upload service persists file + creates `Document` and queued `Job`
-- [ ] Worker processes queued jobs end-to-end via Step 3 service
-- [ ] Success writes transcript text and final `transcribed` status
-- [ ] Failure writes `error_detail` and final `failed` status
-- [ ] Queue-empty worker call returns cleanly without error
-- [ ] Full test suite passes with `uv run pytest -q`
+Step 4 is complete when:
+
+- [ ] Upload service writes file and creates `Document` + queued `Job`
+- [ ] Worker processes queued jobs end-to-end using Step 3 transcription service
+- [ ] Success path persists transcript text and sets `transcribed`
+- [ ] Failure path persists error detail and sets `failed`
+- [ ] Queue-empty path returns cleanly
+- [ ] New tests pass and full suite is green (`uv run pytest -q`)
+- [ ] Output report includes MCP reference usage + validation evidence
\ No newline at end of file
diff --git a/tests/artifacts/transcriptions/Book_Two_-_page_02.jpg.txt b/tests/artifacts/transcriptions/Book_Two_-_page_02.jpg.txt
new file mode 100644
index 0000000..817d991
--- /dev/null
+++ b/tests/artifacts/transcriptions/Book_Two_-_page_02.jpg.txt
@@ -0,0 +1,42 @@
+source: Book Two - page 02.jpg
+provider: openrouter
+model: google/gemini-2.5-flash
+---
+BY WAY OF INTRODUCTION:-
+
+These few paragraphs of introduction may help you read BOOK 2 which covers a
+wider range than did BOOK 1 (Pioneer Days).
+
+BOOK 1 had 54 pages; 14 chapters. BOOK 2 has 70 pages; 18 chapters. BOOK 1 con-
+sisted largely of first generation family history. BOOK 2 throws more light on
+the second generation. Sidney promises a BOOK 3 and that may begin to do justice
+to the third generation. We suggest that Sidney get the help of Louis Shinn
+who has a chapter in this book (Chapter 16 - The Last 25 Years on the Doumeeq
+Plains. Louis has the gift of seeing, recalling and telling. One sentence in
+his chapter gives a great tribute to the Doumeeqers--so far as he knows no one
+on the Doumeeq Plains went on relief during the depression. That in a nutshell
+shows the sturdy character of the residents of the Doumeeq Plains.
+
+We promised in BOOK 1 that in BOOK 2 we would give the story of the trip of
+John E. Cochran and wife to Tennessee, Cuba and the Panama Canal. You will see
+by the Table of Contents that the first four chapters have been given to those
+trips. Those chapters are worth reading and re-reading. Mr. Cochran has eyes
+to see and a pen to tell. We think the people in Tennessee will read with
+great pleasure the comments he makes on conditions today.
+
+Some who get this book will consider the group picture the best thing in the
+book. It took a lot of preliminary photographing to reduce some pictures, enlarge
+others and bring out the tin types. We wish that instead of 44 faces we could
+have given 88. Do not blame Ethel Cochran-Shinn for the selection. She furnished
+enough pictures but we had to take only part of them. We think there are great
+possibilities in reproducing old pictures. We wish we had a Pickard group. Some
+Pickard descendant may wish to make a collection.
+
+We are much impressed with the future possibilities of getting a complete genealog-
+ogy of the Pickard family. Mr. Cochran has a fine chapter on the Pickards but
+to date we have not had the pleasure of finding all of the family dates. We had
+intended to give more family data in this book but it takes time to get the
+correct dates. Often times it requires trips to cemeteries to get dates on the
+tombstones. Winter is no time to collect dates on tombstones.
+
+-2-
diff --git a/tests/artifacts/transcriptions/Omie_Writes_Home.pdf.txt b/tests/artifacts/transcriptions/Omie_Writes_Home.pdf.txt
new file mode 100644
index 0000000..f4290d3
--- /dev/null
+++ b/tests/artifacts/transcriptions/Omie_Writes_Home.pdf.txt
@@ -0,0 +1,111 @@
+source: Omie Writes Home.pdf
+provider: openrouter
+model: google/gemini-2.5-flash
+---
+JOHN E. COCHRAN
+FAMILY ASSOCIATION
+Family Only
+Home | Sibling's Stories | 1st Cousins | Ancestor's Stories | Reunion History | Next Reunion | JECMEF
+OMIE WRITES HOME
+Ed. Note: The following letter was written by Omie Cochran in Nome, Alaska and sent to her
+sister, Ethel Shinn, in Canfield, Idaho in 1923. It has been stored away these 63 years in the
+original envelope with its 2 cent stamp. The letter has a number of references to the Shinn
+children. Peter was Louis; Polly was Edith; and the 'little black rascal' referred to Maurice.
+Miss Saville was the nurse at the Nome Hospital that was mentioned in the article by Inez in
+the family newsletter two years ago.
+Nome Alaska August 26, 1923
+My Dear Ethel et al.
+I don't know when I did write or when you did
+but I am going to write now however and never
+the less. But I wish I could talk (I can yet but I
+mean to tell you all) instead and see ole Unc Pete
+and Polly sit up and listen and that little black
+rascal of yours would fairly sparkle with
+listening. Can't I see him listening now to all the
+yarns we told last summer?
+You see, we-Miss Saville and I, took a trip north
+on the Buford and it was very interesting. We
+went north thru the Bering Strait into the Arctic and as far as the Ice Pack. There the captain
+of our craft and some other mighty hunters went out first in kayaks and later in row boats and
+shot seven walrus. When they also took a movie man and camera, so you will likely see all
+this in the movies before I get to tell you. They came back on board and the ship went up
+along the icebergs and the walrus were 'heisted' on board by the use of the crane which loads
+tons of freight and the beasts were so huge that they made the pulleys just creak. They were
+over 12 feet long, as big around as three cows, had no feet but toenails on their flippers or
+flappers, no head but their body just suddenly ended with a hole for a mouth and big bristles
+
+all around it similar to a currycomb in coarseness; no ears but huge tusks of ivory. They are
+the most repulsive looking animals imaginable and tho I have always read about them I never
+expect such disagreeable looking creatures. They had a rough brown hairy skin and some of
+them looked warty. They must have weighed two ton at least. Ere we got them back to Nome
+to the natives they were getting extremely odiferous—in fact, you could scarcely stay on the
+ship with any degree of comfort unless you had per chance lost your sense of smell.
+
+Then we went north to a few minutes beyond the 70th degree of latitude and thot for awhile
+we would go to Wrangell Island where some men from Steffonsons ship were supposed to be
+stranded but we didn't get there and instead stopped at a small native village at Cape Serdz in
+Siberia. These Eskimo were very primitive. One white squaw man lived there and had for 23
+years. He was a Swede--who else could. Their houses were circular and built up with dirt 2 or
+3 feet and then skins were stretched over it and weighted down with rocks. Inside, the room
+was partitioned off at the sides with skins for sleeping quarters. In the main part they had the
+fire on the ground and the fish drying on lines and the skins hanging around and the dogs and
+babies and children. They wore skin clothes entirely. The women's were made like bloomers
+and were heavily padded for warmth. They wore high mukluks and really looked very
+comfortable. The babies were in fur skins with the fur inside and they looked like little Teddy
+bears with faces. I guess they had never seen white women, not so many at one time anyway.
+We went ashore in 2 life boats and a launch pulling them. On the launch was a six piece band
+playing and the rear of the last life boat was the movie man. 'Twas very thrilling.
+
+The other place we stopped was at Whalen, a trading post in Siberia. There these 'towerists'
+went wild. They rushed helter-skelter, hither and thither, here and there, trying to find
+something to buy. Prices raised right before your eyes. One would but something for $1.00
+and the next might have to pay $2.00, $4.00 or $10.00. That made no difference. They had to
+have it. One man I was sort of taking care of, tho he had his son along for the purpose,
+bought 2 ivory tusks, 1 pup, 2 moccasins, 3 or 4 billikens, 6 or 8 ivory and silver rings, one
+fishing line, hooks, floats, etc. and two bird slings. The slings have rocks at the end and the
+little natives throw them at the flocks of geese and ducks which fly close over the village and
+the slings entangle their wings and legs, sometimes more than one, and they can't fly. They
+come down and the natives capture them. There was more junk brot aboard than baggage, I
+do believe. And they say that at the first stop it was worse than here. The red flag was flying
+over Whalen and the Russian soldiers were there—a few, one or two or three, I forget the
+number.
+
+We got home yesterday morning at 5 a.m. but missed the first lighter in so had to stay out
+until 2:30. The girls had prepared a big meal for us and invited up the Hartfords and then let
+us talk. Miss Saville talked quite a bit. Any how if you folks don't like this I don't care, it is
+all I had to write about and I know Buster'ud listen anyway and I'd soak ole Peter's head if he
+didn't and Polly would in my lap and I don't know much about the youngest one of yours so
+likely he would be squawling. But we did surely enjoy our trip and were gone just long enuf.
+
+I expect there were 150 passengers on board and almost or more of the crew and helpers. We
+had a stateroom down next to the kitchen and 'twas pretty fierce for odor at times.
+
+I have had jobs nearly all summer but not very much in them. Next week, September 4,
+school opens. I wish they would wait for a week but you know these school men. Wouldn't
+make any special difference I suppose for I would just fritter away the time but still one likes
+to postpone the inevitable.
+
+I just now stopped and re-read your letter and it sure was a bird. I don't especially blame Ray
+for reading over your shoulder. It would seem, then that you have bright children. Maybe
+they do know something about Geography. But it is ridiculous to speak of Louis finishing the
+eighth grade. Why you and I were grown children when we finished and he is only a baby. I
+am rather afraid he doesn't know much. I quite remember your little timid Maurice and how
+he shifted his affection from his Aunt Om and yelled and howled and screamed steadily for a
+week while his mother went to S.S. Ask him is he recalls this little interview. Do you suppose
+he does?
+
+Ever hear from Zen? or Inez? They don't seem to be very writing inclined tho Zen has done
+well this year. Even sent me a telegram a few weeks ago. Well, if anything else ever happens,
+I'll write again. Don't suppose it ever will, tho.
+
+Lots of love to all,
+
+Ome
+
+Reprinted from Cochran Chronicles, Volume 9, Number 1, November 1986
+
+© JECFA 1986
+
+Up
+
+jecochranclan.org ~ Contact webmaster
diff --git a/tests/artifacts/transcriptions/Rod_Moser_Letter_-_p1.jpg.txt b/tests/artifacts/transcriptions/Rod_Moser_Letter_-_p1.jpg.txt
new file mode 100644
index 0000000..102f81a
--- /dev/null
+++ b/tests/artifacts/transcriptions/Rod_Moser_Letter_-_p1.jpg.txt
@@ -0,0 +1,31 @@
+source: Rod Moser Letter - p1.jpg
+provider: openrouter
+model: google/gemini-2.5-flash
+---
+JOHN ISBILL R. T. MOSER
+
+ISBILL & MOSER
+DEALERS IN
+GENERAL MERCHANDISE
+
+Vonore, Tenn. Jany 27 – 1913
+Dear Much Aunt Louie
+How are you—
+few nights ago received a
+letter from your folks, So
+I decided to write you
+a few lines myself &
+I am contemplating a
+trip out west next summer
+& would like for [illegible] to go
+where I am.
+
+Am getting
+up in years & unmarried
+so you see the object of
+my trip is to get a wife
+& if there is any old maid
+or widdow out there, I
+want you to kiss them
+at my [illegible] for me at They
+as soon as I get there
diff --git a/tests/fixtures/images/invalid/empty.bin b/tests/fixtures/images/invalid/empty.bin
new file mode 100644
index 0000000..e69de29
diff --git a/tests/fixtures/images/invalid/wrong_extension.txt b/tests/fixtures/images/invalid/wrong_extension.txt
new file mode 100644
index 0000000..2cebe82
--- /dev/null
+++ b/tests/fixtures/images/invalid/wrong_extension.txt
@@ -0,0 +1 @@
+not an image fixture
\ No newline at end of file
diff --git a/tests/fixtures/images/real/Book Two - page 02.jpg b/tests/fixtures/images/real/Book Two - page 02.jpg
new file mode 100644
index 0000000..7f2267a
Binary files /dev/null and b/tests/fixtures/images/real/Book Two - page 02.jpg differ
diff --git a/tests/fixtures/images/real/Omie Writes Home.pdf b/tests/fixtures/images/real/Omie Writes Home.pdf
new file mode 100644
index 0000000..6f30949
Binary files /dev/null and b/tests/fixtures/images/real/Omie Writes Home.pdf differ
diff --git a/tests/fixtures/images/real/Rod Moser Letter - p1.jpg b/tests/fixtures/images/real/Rod Moser Letter - p1.jpg
new file mode 100644
index 0000000..a1e9432
Binary files /dev/null and b/tests/fixtures/images/real/Rod Moser Letter - p1.jpg differ
diff --git a/tests/fixtures/images/valid/single_page_pdf.pdf b/tests/fixtures/images/valid/single_page_pdf.pdf
new file mode 100644
index 0000000..d61520a
--- /dev/null
+++ b/tests/fixtures/images/valid/single_page_pdf.pdf
@@ -0,0 +1,2 @@
+%PDF-1.4
+%fixture
diff --git a/tests/fixtures/images/valid/small_jpg.jpg b/tests/fixtures/images/valid/small_jpg.jpg
new file mode 100644
index 0000000..5ed4353
--- /dev/null
+++ b/tests/fixtures/images/valid/small_jpg.jpg
@@ -0,0 +1 @@
+ÿØÿàfixture
\ No newline at end of file
diff --git a/tests/fixtures/images/valid/small_png.png b/tests/fixtures/images/valid/small_png.png
new file mode 100644
index 0000000..aa62e8b
--- /dev/null
+++ b/tests/fixtures/images/valid/small_png.png
@@ -0,0 +1,3 @@
+‰PNG
+
+fixture
\ No newline at end of file
diff --git a/tests/fixtures/images/valid/small_tiff.tiff b/tests/fixtures/images/valid/small_tiff.tiff
new file mode 100644
index 0000000..bf1d7e9
Binary files /dev/null and b/tests/fixtures/images/valid/small_tiff.tiff differ
diff --git a/tests/fixtures/prompts/invalid/empty_prompt.md b/tests/fixtures/prompts/invalid/empty_prompt.md
new file mode 100644
index 0000000..e69de29
diff --git a/tests/fixtures/prompts/valid/transcribe_document_min.md b/tests/fixtures/prompts/valid/transcribe_document_min.md
new file mode 100644
index 0000000..f2ea2db
--- /dev/null
+++ b/tests/fixtures/prompts/valid/transcribe_document_min.md
@@ -0,0 +1,6 @@
+# Minimal Verbatim Prompt
+
+Transcribe the document verbatim.
+Do not summarize or paraphrase.
+Mark uncertain text using square brackets and a question mark.
+Mark unreadable text as [illegible].
\ No newline at end of file
diff --git a/tests/services/test_transcription_external.py b/tests/services/test_transcription_external.py
new file mode 100644
index 0000000..7cb3aed
--- /dev/null
+++ b/tests/services/test_transcription_external.py
@@ -0,0 +1,70 @@
+"""Opt-in external tests for real document image transcription."""
+
+import os
+from pathlib import Path
+
+import pytest
+
+from transcription.services.transcription import transcribe_document_image
+
+
+HAS_OPENROUTER_KEY = bool(os.getenv("OPENROUTER_API_KEY"))
+
+REAL_IMAGES_DIR = Path(__file__).resolve().parents[1] / "fixtures" / "images" / "real"
+ARTIFACTS_DIR = Path(__file__).resolve().parents[1] / "artifacts" / "transcriptions"
+SUPPORTED_EXTENSIONS = {".jpg", ".jpeg", ".png", ".tif", ".tiff", ".pdf"}
+
+pytestmark = [
+    pytest.mark.external,
+    pytest.mark.skipif(
+        not HAS_OPENROUTER_KEY,
+        reason="Set OPENROUTER_API_KEY to run external real-image tests.",
+    ),
+]
+
+
+def _real_image_paths() -> list[Path]:
+    if not REAL_IMAGES_DIR.exists():
+        return []
+    return sorted(
+        [
+            p
+            for p in REAL_IMAGES_DIR.iterdir()
+            if p.is_file() and p.suffix.lower() in SUPPORTED_EXTENSIONS
+        ]
+    )
+
+
+def _artifact_filename(image_path: Path) -> str:
+    safe_stem = image_path.stem.replace(" ", "_")
+    safe_suffix = image_path.suffix.lower().replace(".", "")
+    return f"{safe_stem}.{safe_suffix}.txt"
+
+
+class TestRealImageExternalTranscription:
+    """Validate transcription against real local fixtures via live provider."""
+
+    def test_real_image_fixture_set_exists(self):
+        """At least one supported real-image fixture exists for external tests."""
+        assert REAL_IMAGES_DIR.exists()
+        assert _real_image_paths()
+
+    @pytest.mark.parametrize("image_path", _real_image_paths(), ids=lambda p: p.name)
+    def test_transcribes_real_image_fixture(self, image_path: Path):
+        """Real fixture image produces a non-empty transcription result."""
+        result = transcribe_document_image(image_path)
+        assert result.provider == "openrouter"
+        assert isinstance(result.model, str) and result.model.strip()
+        assert isinstance(result.text, str) and result.text.strip()
+
+        ARTIFACTS_DIR.mkdir(parents=True, exist_ok=True)
+        artifact_path = ARTIFACTS_DIR / _artifact_filename(image_path)
+        artifact_text = (
+            f"source: {image_path.name}\n"
+            f"provider: {result.provider}\n"
+            f"model: {result.model}\n"
+            "---\n"
+            f"{result.text}\n"
+        )
+        artifact_path.write_text(artifact_text, encoding="utf-8")
+        assert artifact_path.exists()
\ No newline at end of file