Extract attachment text off the event loop; reject text-less files

- extract_text (PyMuPDF for PDFs) now runs via asyncio.to_thread in both
  /upload and stream-query, so a large PDF doesn't stall other streams.
- /upload returns 422 when extraction yields no text: a scanned PDF
  without a text layer, or an empty text file.
- Mock server: *scan*.pdf uploads return that 422.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

blueprints/conversation/__init__.py

@@ -228,7 +228,8 @@ async def _load_attached_file(file_key: str) -> tuple[str, str]:
     data, metadata = await s3_get_file(file_key)
     name = unquote(metadata.get("filename", "")) or file_key.rsplit("/", 1)[-1]
     ext = file_extension(file_key, None) or ""
-    return name, extract_text(data, ext)
+    # PyMuPDF is CPU-bound; keep it off the event loop so other streams don't stall.
+    return name, await asyncio.to_thread(extract_text, data, ext)
 
 
 @conversation_blueprint.post("/upload")
@@ -278,9 +279,16 @@ async def upload_attachment():
         )
 
     try:
-        text = extract_text(file_bytes, ext)
+        text = await asyncio.to_thread(extract_text, file_bytes, ext)
     except FileExtractionError as e:
         return jsonify({"error": str(e)}), 400
+    if not text.strip():
+        message = (
+            "This PDF has no text layer (scanned?), so Simba can't read it yet."
+            if ext == "pdf"
+            else "This file is empty, so there's nothing for Simba to read."
+        )
+        return jsonify({"error": message}), 422
 
     key = f"conversations/{conversation.id}/files/{uuid.uuid4()}.{ext}"
     await s3_upload_file(file_bytes, key, FILE_TYPES[ext], {"filename": quote(name)})

tests/unit/test_file_upload.py

@@ -203,6 +203,36 @@ class TestUploadRoute:
         resp = await _upload(app, headers, b"nope", "a.pdf", "application/pdf")
         assert resp.status_code == 400
 
+    async def test_pdf_without_text_layer_is_422(self, app, headers, s3):
+        doc = fitz.open()
+        doc.new_page()  # blank page: no text, like a scan without OCR
+        resp = await _upload(app, headers, doc.tobytes(), "scan.pdf", "application/pdf")
+        assert resp.status_code == 422
+        assert "no text layer" in (await resp.get_json())["error"]
+        s3.upload_file.assert_not_awaited()
+
+    async def test_empty_text_file_is_422(self, app, headers, s3):
+        resp = await _upload(app, headers, b"  \n", "empty.txt", "text/plain")
+        assert resp.status_code == 422
+        s3.upload_file.assert_not_awaited()
+
+    async def test_extraction_runs_off_the_event_loop(self, app, headers, s3):
+        calls = []
+
+        async def fake_to_thread(fn, *args):
+            calls.append(fn)
+            return fn(*args)
+
+        with patch.object(conv.asyncio, "to_thread", fake_to_thread):
+            resp = await _upload(
+                app, headers, _pdf_bytes("hi"), "a.pdf", "application/pdf"
+            )
+            get = AsyncMock(return_value=(b"x", {}))
+            with patch.object(conv, "s3_get_file", get):
+                await conv._load_attached_file(f"conversations/{CONV_ID}/files/u.txt")
+        assert resp.status_code == 200
+        assert calls == [conv.extract_text, conv.extract_text]
+
     async def test_other_users_conversation_is_404(self, app, headers, s3):
         resp = await _upload(
             app, headers, b"hi", "a.txt", "text/plain", conversation_id=OTHER_ID