allowing image inputs too

app.py

@@ -92,11 +92,18 @@ def render_page(pdf_path: str, page_num: int, dpi: int) -> Image.Image:
         doc.close()
 
 
-def parse(pdf_file, page_num, dpi, text_in_pic, table_format):
-    if pdf_file is None:
-        raise gr.Error("Please upload a PDF first.")
+def load_input(file_path: str, page_num: int, dpi: int) -> Image.Image:
+    """Return an RGB image from either a PDF page or an image file."""
+    if file_path.lower().endswith(".pdf"):
+        return render_page(file_path, page_num, dpi)
+    return Image.open(file_path).convert("RGB")
 
-    image = render_page(pdf_file, int(page_num), int(dpi))
+
+def parse(input_file, page_num, dpi, text_in_pic, table_format):
+    if input_file is None:
+        raise gr.Error("Please upload a PDF or image first.")
+
+    image = load_input(input_file, int(page_num), int(dpi))
 
     fourth = "<predict_text_in_pic>" if text_in_pic else "<predict_no_text_in_pic>"
     task_prompt = f"</s><s><predict_bbox><predict_classes><output_markdown>{fourth}"
@@ -130,17 +137,23 @@ def parse(pdf_file, page_num, dpi, text_in_pic, table_format):
 with gr.Blocks(title="Nemotron Parse — Repair Manuals") as demo:
     gr.Markdown(
         "# 🔧 Nemotron Parse v1.2 — Repair Manual Explorer\n"
-        "Upload a PDF, choose a page, and parse it with "
+        "Upload a PDF (choose a page) or an image, and parse it with "
         "[NVIDIA Nemotron Parse v1.2](https://huggingface.co/nvidia/NVIDIA-Nemotron-Parse-v1.2) "
         "on ZeroGPU. Returns structured markdown, a JSON of elements, and an "
         "annotated page image."
     )
     with gr.Row():
         with gr.Column(scale=1):
-            pdf_in = gr.File(label="PDF", file_types=[".pdf"], type="filepath")
-            page_in = gr.Number(label="Page", value=1, precision=0, minimum=1)
+            pdf_in = gr.File(
+                label="PDF or image",
+                file_types=[".pdf", ".png", ".jpg", ".jpeg", ".webp"],
+                type="filepath",
+            )
+            page_in = gr.Number(
+                label="Page (PDF only)", value=1, precision=0, minimum=1
+            )
             dpi_in = gr.Slider(
-                label="Render DPI", minimum=72, maximum=300, value=150, step=10
+                label="Render DPI (PDF only)", minimum=72, maximum=300, value=150, step=10
             )
             text_in_pic_in = gr.Checkbox(
                 label="Extract text inside pictures/diagrams", value=False