fix(pdf): flag scanned PDFs in pdf-to-text and serve text as UTF-8 (#603)

When a PDF has no text layer (scanned or image-only), pdf-to-text now
returns a 422 that points at the OCR tool instead of a silent empty file,
and text downloads carry charset=utf-8 so UTF-8 Arabic renders correctly
when the .txt is viewed inline.

Fixes #589

packages/ai/python/doc_text.py

@@ -1,4 +1,4 @@
-"""Extract plain text. Args: {"path": in, "out": out-txt-path}. Prints {"chars": N}."""
+"""Extract plain text. Args: {"path": in, "out": out-txt-path}. Prints {"chars": N, "hasText": bool}."""
 import json
 import sys
 
@@ -19,9 +19,14 @@ def main():
         parts = [page.get_text() for page in doc]
         doc.close()
         text = "\n".join(parts)
+        # hasText distinguishes a PDF with a real text layer from a scanned or
+        # image-only PDF, where every page returns "" and only the join newlines
+        # remain. len(text) alone can't tell them apart, so the caller uses this
+        # to offer OCR instead of handing back an empty file (#589).
+        has_text = any(part.strip() for part in parts)
         with open(out, "w", encoding="utf-8") as fh:
             fh.write(text)
-        print(json.dumps({"chars": len(text)}))
+        print(json.dumps({"chars": len(text), "hasText": has_text}))
     except Exception as exc:  # noqa: BLE001
         print(json.dumps({"error": str(exc)}))
         sys.exit(1)