Fixes batch 2

deployment/docker-compose-azure.yaml

@@ -14,7 +14,7 @@ services:
       context: ../application
       args:
         INSTALL_DOCLING: ${INSTALL_DOCLING:-false}
-        INSTALL_TESSERACT: ${INSTALL_TESSERACT:-true}
+        INSTALL_TESSERACT: ${INSTALL_TESSERACT:-false}
     env_file:
       - ../.env
     environment:
@@ -39,7 +39,7 @@ services:
       context: ../application
       args:
         INSTALL_DOCLING: ${INSTALL_DOCLING:-false}
-        INSTALL_TESSERACT: ${INSTALL_TESSERACT:-true}
+        INSTALL_TESSERACT: ${INSTALL_TESSERACT:-false}
     # `parsing` queue carries read_document/parse_document; required for its await to resolve.
     command: celery -A application.app.celery worker -l INFO -Q docsgpt,parsing
     env_file:

deployment/docker-compose.yaml

@@ -21,8 +21,10 @@ services:
         # Bake the optional docling engine (layout-model OCR backend, structured
         # output) into the image: set INSTALL_DOCLING=true in ../.env or the shell.
         INSTALL_DOCLING: ${INSTALL_DOCLING:-false}
-        # The tesseract binary behind OCR_ENABLED=true; on by default (~35 MB).
-        INSTALL_TESSERACT: ${INSTALL_TESSERACT:-true}
+        # Bake the tesseract binary behind OCR_ENABLED=true (~35 MB): set
+        # INSTALL_TESSERACT=true in ../.env or the shell (setup.sh does this
+        # when OCR is enabled). Off by default, like docling.
+        INSTALL_TESSERACT: ${INSTALL_TESSERACT:-false}
     env_file:
       - ../.env
     environment:
@@ -49,7 +51,7 @@ services:
       context: ../application
       args:
         INSTALL_DOCLING: ${INSTALL_DOCLING:-false}
-        INSTALL_TESSERACT: ${INSTALL_TESSERACT:-true}
+        INSTALL_TESSERACT: ${INSTALL_TESSERACT:-false}
     # Consumes the default queue AND the dedicated `parsing` queue (read_document /
     # parse_document). Without `parsing` here the read_document await never resolves.
     # For heavy/OCR parsing run a separate worker with `-Q parsing`.