Update app.py

1. Always run header and footer band OCR
Header band: always OCR from top of page down to he (and if he <= 0, use he = DEFAULT_ZONE_FRAC).
Footer band: always OCR from fs to bottom (and if fs >= 1.0, use fs = 1.0 - DEFAULT_ZONE_FRAC).
So we always run both bands and no longer skip them when the API returns no or full-page regions.
2. Larger default bands
DEFAULT_ZONE_FRAC increased from 0.08 to 0.12 (top 12% and bottom 12%) so header/footer are more likely to be covered. You can override with the GLMOCR_DEFAULT_ZONE_FRAC env var (e.g. 0.15 for 15%).
3. Config and formatter writes are best-effort
The config and result_formatter.py patches are wrapped in try/except. If the filesystem is read-only (e.g. on HF), the app still starts; for MaaS, header/footer now come from the client-side band OCR anyway

app.py

@@ -14,59 +14,63 @@ GLMOCR_BASE = os.path.dirname(glmocr.__file__)
 CONFIG_PATH = os.path.join(GLMOCR_BASE, "config.yaml")
 FORMATTER_PATH = os.path.join(GLMOCR_BASE, "postprocess", "result_formatter.py")
 
-DEFAULT_ZONE_FRAC = float(os.environ.get("GLMOCR_DEFAULT_ZONE_FRAC", "0.08"))
+# When MaaS is enabled, the cloud API ignores local layout config and does not return
+# header/footer regions. We always run client-side OCR on top/bottom bands as fallback.
+DEFAULT_ZONE_FRAC = float(os.environ.get("GLMOCR_DEFAULT_ZONE_FRAC", "0.12"))
 
 # ---------------------------------------------------------------------------
-# 1. Config: include headers/footers (remove from abandon, add to text mapping)
+# 1. Config: set MaaS and optionally include headers/footers for non-MaaS
 # ---------------------------------------------------------------------------
-with open(CONFIG_PATH, "r") as f:
-    config = yaml.safe_load(f)
-
-config["pipeline"]["maas"]["enabled"] = True
-config["pipeline"]["maas"]["api_key"] = "4570c28bdea5493c9efae9dae68edc66.sGbA9DLlcX1GlvqV"
-
-to_include_as_text = {"header", "footer"}
-layout_section = config.get("pipeline", {}).get("layout", {})
-if "label_task_mapping" in layout_section:
-    mapping = layout_section["label_task_mapping"]
-    # Remove header/footer from abandon so they are not discarded
-    abandon = mapping.get("abandon")
-    if isinstance(abandon, list):
-        mapping["abandon"] = [x for x in abandon if x not in to_include_as_text]
-    # Add header/footer to text so they are OCR'd and output as text
-    text_labels = mapping.get("text")
-    if isinstance(text_labels, list):
-        for label in to_include_as_text:
-            if label not in text_labels:
-                text_labels.append(label)
-
-formatter_section = config.get("pipeline", {}).get("result_formatter", {})
-if "abandon" in formatter_section:
-    abandon = formatter_section["abandon"]
-    if isinstance(abandon, list):
-        formatter_section["abandon"] = [x for x in abandon if x not in to_include_as_text]
-# Include header/footer in visualization/output text category
-if "label_visualization_mapping" in formatter_section:
-    text_vis = formatter_section["label_visualization_mapping"].get("text")
-    if isinstance(text_vis, list):
-        for label in to_include_as_text:
-            if label not in text_vis:
-                text_vis.append(label)
-
-with open(CONFIG_PATH, "w") as f:
-    yaml.dump(config, f, default_flow_style=False, sort_keys=False)
+try:
+    with open(CONFIG_PATH, "r") as f:
+        config = yaml.safe_load(f)
+    config["pipeline"]["maas"]["enabled"] = True
+    config["pipeline"]["maas"]["api_key"] = os.environ.get("GLMOCR_API_KEY", "4570c28bdea5493c9efae9dae68edc66.sGbA9DLlcX1GlvqV")
+
+    to_include_as_text = {"header", "footer"}
+    layout_section = config.get("pipeline", {}).get("layout", {})
+    if "label_task_mapping" in layout_section:
+        mapping = layout_section["label_task_mapping"]
+        abandon = mapping.get("abandon")
+        if isinstance(abandon, list):
+            mapping["abandon"] = [x for x in abandon if x not in to_include_as_text]
+        text_labels = mapping.get("text")
+        if isinstance(text_labels, list):
+            for label in to_include_as_text:
+                if label not in text_labels:
+                    text_labels.append(label)
+
+    formatter_section = config.get("pipeline", {}).get("result_formatter", {})
+    if "abandon" in formatter_section:
+        abandon = formatter_section["abandon"]
+        if isinstance(abandon, list):
+            formatter_section["abandon"] = [x for x in abandon if x not in to_include_as_text]
+    if "label_visualization_mapping" in formatter_section:
+        text_vis = formatter_section["label_visualization_mapping"].get("text")
+        if isinstance(text_vis, list):
+            for label in to_include_as_text:
+                if label not in text_vis:
+                    text_vis.append(label)
+
+    with open(CONFIG_PATH, "w") as f:
+        yaml.dump(config, f, default_flow_style=False, sort_keys=False)
+except Exception:
+    pass  # e.g. read-only package on Hugging Face; MaaS ignores layout anyway
 
 # ---------------------------------------------------------------------------
-# 2. result_formatter: stop stripping header/footer in postprocess
+# 2. result_formatter: stop stripping header/footer (best-effort; may be read-only)
 # ---------------------------------------------------------------------------
-with open(FORMATTER_PATH, "r") as f:
-    source = f.read()
-for label in ('"header"', "'header'", '"footer"', "'footer'", '"doc_header"', "'doc_header'", '"doc_footer"', "'doc_footer'"):
-    source = re.sub(r",\s*" + re.escape(label), "", source)
-    source = re.sub(re.escape(label) + r"\s*,", "", source)
-    source = re.sub(re.escape(label), "", source)
-with open(FORMATTER_PATH, "w") as f:
-    f.write(source)
+try:
+    with open(FORMATTER_PATH, "r") as f:
+        source = f.read()
+    for label in ('"header"', "'header'", '"footer"', "'footer'", '"doc_header"', "'doc_header'", '"doc_footer"', "'doc_footer'"):
+        source = re.sub(r",\s*" + re.escape(label), "", source)
+        source = re.sub(re.escape(label) + r"\s*,", "", source)
+        source = re.sub(re.escape(label), "", source)
+    with open(FORMATTER_PATH, "w") as f:
+        f.write(source)
+except Exception:
+    pass
 
 
 def get_header_footer_zones(regions, norm_height=1000):
@@ -134,7 +138,9 @@ def get_page_md_and_regions(page_result):
     regions = []
     if hasattr(page_result, "json_result"):
         jr = page_result.json_result
-        if isinstance(jr, list) and len(jr) > 0:
+        if isinstance(jr, dict) and "regions" in jr:
+            regions = jr.get("regions") or []
+        elif isinstance(jr, list) and len(jr) > 0:
             r = jr[0] if isinstance(jr[0], list) else jr
             if isinstance(r, list):
                 regions = r
@@ -175,16 +181,18 @@ def run_ocr(uploaded_file):
             page_md, regions = get_page_md_and_regions(page_result)
             header_end_frac, footer_start_frac = get_header_footer_zones(regions, 1000)
 
+            # MaaS does not return header/footer regions; always use at least default bands
             he = header_end_frac if header_end_frac is not None else DEFAULT_ZONE_FRAC
             fs = footer_start_frac if footer_start_frac is not None else (1.0 - DEFAULT_ZONE_FRAC)
-            if he <= 0 and header_end_frac is not None:
-                he = 0
-            if fs >= 1.0 and footer_start_frac is not None:
-                fs = 1.0
+            if he <= 0:
+                he = DEFAULT_ZONE_FRAC  # ensure we always OCR a top band
+            if fs >= 1.0:
+                fs = 1.0 - DEFAULT_ZONE_FRAC  # ensure we always OCR a bottom band
 
             parts = []
 
-            if he > 0 and page_num < len(page_images):
+            # Always run header band (top 8–12% of page)
+            if page_num < len(page_images):
                 img_path = page_images[page_num]
                 hdr = ""
                 if is_pdf:
@@ -197,7 +205,8 @@ def run_ocr(uploaded_file):
             if page_md:
                 parts.append(page_md)
 
-            if fs < 1.0 and page_num < len(page_images):
+            # Always run footer band (bottom 8–12% of page)
+            if page_num < len(page_images):
                 img_path = page_images[page_num]
                 ftr = ""
                 if is_pdf: