Update app.py

app.py

@@ -160,7 +160,7 @@ def make_cache_key(image: Image.Image, prompt: str) -> str:
 
 # --- 7. Inference with mixed precision & error handling ---
 def run_inference(image: Image.Image, prompt: str = "") -> str:
-    prompt_text = prompt.strip() or "Read information from the document."
+    prompt_text = prompt.strip() or ""
     img = normalize_image(image)
     key = make_cache_key(img, prompt_text)
 
@@ -186,7 +186,7 @@ def run_inference(image: Image.Image, prompt: str = "") -> str:
             with torch.cuda.amp.autocast(enabled=(device.type=='cuda')):
                 gen = model.generate(
                     **inputs,
-                    max_new_tokens=512,
+                    max_new_tokens=2048,
                     do_sample=False,
                     eos_token_id=processor.tokenizer.eos_token_id
                 )
@@ -280,17 +280,17 @@ STRICT INSTRUCTIONS – read carefully and follow EXACTLY:
 2. DO NOT add, remove, rename, or reorder any XML tags.
 3. DO NOT include explanations, markdown, notes, comments, or extra spacing outside the XML block.
 4. For every tag, fill in the exact value read from the image.
-   • NEVER copy or repeat the label/placeholder text.  
+   • NEVER copy or repeat the label/placeholder text.
    • NEVER guess or invent values.
 5. If a value is missing or unreadable, leave the tag EMPTY (e.g. <tag></tag>).
 6. DO NOT include Vietnamese text or translations inside tag values.
 7. The output MUST start with the root tag and end with its correct closing tag; all tags must be well-formed.
 8. Dates must be in YYYY-MM-DD format.
-9. Boolean tags must be exactly true or false (lower-case, no quotes).  
+9. Boolean tags must be exactly true or false (lower-case, no quotes).
    ✔ √ Yes Passed ⇒ true | ✘ X No Fail ⇒ false
-10. **Inside each value**  
-    • Replace every internal line-break with “, ” (comma + space).  
-    • Trim leading/trailing whitespace.  
+10. **Inside each value**
+    • Replace every internal line-break with “, ” (comma + space).
+    • Trim leading/trailing whitespace.
     • Escape XML special characters: & → &amp;, < → &lt;, > → &gt;.
 11. **Phone / contact fields** – digits, “+”, “–”, spaces only; if multiple numbers, separate with “, ”.
 12. **Signature fields** – fill ONLY if the signature appears as legible text; if it is handwritten, leave the tag empty.
@@ -327,17 +327,17 @@ STRICT INSTRUCTIONS – read carefully and follow EXACTLY:
 2. DO NOT add, remove, rename, or reorder any XML tags.
 3. DO NOT include explanations, markdown, notes, comments, or extra spacing outside the XML block.
 4. For every tag, fill in the exact value read from the image.
-   • NEVER copy or repeat the label/placeholder text.  
+   • NEVER copy or repeat the label/placeholder text.
    • NEVER guess or invent values.
 5. If a value is missing or unreadable, leave the tag EMPTY (e.g. <tag></tag>).
 6. DO NOT include Vietnamese text or translations inside tag values.
 7. The output MUST start with the root tag and end with its correct closing tag; all tags must be well-formed.
 8. Dates must be in YYYY-MM-DD format.
-9. Boolean tags must be exactly true or false (lower-case, no quotes).  
+9. Boolean tags must be exactly true or false (lower-case, no quotes).
    ✔ √ Yes Passed ⇒ true | ✘ X No Fail ⇒ false
-10. **Inside each value**  
-    • Replace every internal line-break with “, ” (comma + space).  
-    • Trim leading/trailing whitespace.  
+10. **Inside each value**
+    • Replace every internal line-break with “, ” (comma + space).
+    • Trim leading/trailing whitespace.
     • Escape XML special characters: & → &amp;, < → &lt;, > → &gt;.
 11. **Phone / contact fields** – digits, “+”, “–”, spaces only; if multiple numbers, separate with “, ”.
 12. **Signature fields** – fill ONLY if the signature appears as legible text; if it is handwritten, leave the tag empty.
@@ -366,17 +366,17 @@ STRICT INSTRUCTIONS – read carefully and follow EXACTLY:
 2. DO NOT add, remove, rename, or reorder any XML tags.
 3. DO NOT include explanations, markdown, notes, comments, or extra spacing outside the XML block.
 4. For every tag, fill in the exact value read from the image.
-   • NEVER copy or repeat the label/placeholder text.  
+   • NEVER copy or repeat the label/placeholder text.
    • NEVER guess or invent values.
 5. If a value is missing or unreadable, leave the tag EMPTY (e.g. <tag></tag>).
 6. DO NOT include Vietnamese text or translations inside tag values.
 7. The output MUST start with the root tag and end with its correct closing tag; all tags must be well-formed.
 8. Dates must be in YYYY-MM-DD format.
-9. Boolean tags must be exactly true or false (lower-case, no quotes).  
+9. Boolean tags must be exactly true or false (lower-case, no quotes).
    ✔ √ Yes Passed ⇒ true | ✘ X No Fail ⇒ false
-10. **Inside each value**  
-    • Replace every internal line-break with “, ” (comma + space).  
-    • Trim leading/trailing whitespace.  
+10. **Inside each value**
+    • Replace every internal line-break with “, ” (comma + space).
+    • Trim leading/trailing whitespace.
     • Escape XML special characters: & → &amp;, < → &lt;, > → &gt;.
 11. **Phone / contact fields** – digits, “+”, “–”, spaces only; if multiple numbers, separate with “, ”.
 12. **Signature fields** – fill ONLY if the signature appears as legible text; if it is handwritten, leave the tag empty.
@@ -401,17 +401,17 @@ STRICT INSTRUCTIONS – read carefully and follow EXACTLY:
 2. DO NOT add, remove, rename, or reorder any XML tags.
 3. DO NOT include explanations, markdown, notes, comments, or extra spacing outside the XML block.
 4. For every tag, fill in the exact value read from the image.
-   • NEVER copy or repeat the label/placeholder text.  
+   • NEVER copy or repeat the label/placeholder text.
    • NEVER guess or invent values.
 5. If a value is missing or unreadable, leave the tag EMPTY (e.g. <tag></tag>).
 6. DO NOT include Vietnamese text or translations inside tag values.
 7. The output MUST start with the root tag and end with its correct closing tag; all tags must be well-formed.
 8. Dates must be in YYYY-MM-DD format.
-9. Boolean tags must be exactly true or false (lower-case, no quotes).  
+9. Boolean tags must be exactly true or false (lower-case, no quotes).
    ✔ √ Yes Passed ⇒ true | ✘ X No Fail ⇒ false
-10. **Inside each value**  
-    • Replace every internal line-break with “, ” (comma + space).  
-    • Trim leading/trailing whitespace.  
+10. **Inside each value**
+    • Replace every internal line-break with “, ” (comma + space).
+    • Trim leading/trailing whitespace.
     • Escape XML special characters: & → &amp;, < → &lt;, > → &gt;.
 11. **Phone / contact fields** – digits, “+”, “–”, spaces only; if multiple numbers, separate with “, ”.
 12. **Signature fields** – fill ONLY if the signature appears as legible text; if it is handwritten, leave the tag empty.
@@ -437,17 +437,17 @@ STRICT INSTRUCTIONS – read carefully and follow EXACTLY:
 2. DO NOT add, remove, rename, or reorder any XML tags.
 3. DO NOT include explanations, markdown, notes, comments, or extra spacing outside the XML block.
 4. For every tag, fill in the exact value read from the image.
-   • NEVER copy or repeat the label/placeholder text.  
+   • NEVER copy or repeat the label/placeholder text.
    • NEVER guess or invent values.
 5. If a value is missing or unreadable, leave the tag EMPTY (e.g. <tag></tag>).
 6. DO NOT include Vietnamese text or translations inside tag values.
 7. The output MUST start with the root tag and end with its correct closing tag; all tags must be well-formed.
 8. Dates must be in YYYY-MM-DD format.
-9. Boolean tags must be exactly true or false (lower-case, no quotes).  
+9. Boolean tags must be exactly true or false (lower-case, no quotes).
    ✔ √ Yes Passed ⇒ true | ✘ X No Fail ⇒ false
-10. **Inside each value**  
-    • Replace every internal line-break with “, ” (comma + space).  
-    • Trim leading/trailing whitespace.  
+10. **Inside each value**
+    • Replace every internal line-break with “, ” (comma + space).
+    • Trim leading/trailing whitespace.
     • Escape XML special characters: & → &amp;, < → &lt;, > → &gt;.
 11. **Phone / contact fields** – digits, “+”, “–”, spaces only; if multiple numbers, separate with “, ”.
 12. **Signature fields** – fill ONLY if the signature appears as legible text; if it is handwritten, leave the tag empty.
@@ -475,17 +475,17 @@ STRICT INSTRUCTIONS – read carefully and follow EXACTLY:
 2. DO NOT add, remove, rename, or reorder any XML tags.
 3. DO NOT include explanations, markdown, notes, comments, or extra spacing outside the XML block.
 4. For every tag, fill in the exact value read from the image.
-   • NEVER copy or repeat the label/placeholder text.  
+   • NEVER copy or repeat the label/placeholder text.
    • NEVER guess or invent values.
 5. If a value is missing or unreadable, leave the tag EMPTY (e.g. <tag></tag>).
 6. DO NOT include Vietnamese text or translations inside tag values.
 7. The output MUST start with the root tag and end with its correct closing tag; all tags must be well-formed.
 8. Dates must be in YYYY-MM-DD format.
-9. Boolean tags must be exactly true or false (lower-case, no quotes).  
+9. Boolean tags must be exactly true or false (lower-case, no quotes).
    ✔ √ Yes Passed ⇒ true | ✘ X No Fail ⇒ false
-10. **Inside each value**  
-    • Replace every internal line-break with “, ” (comma + space).  
-    • Trim leading/trailing whitespace.  
+10. **Inside each value**
+    • Replace every internal line-break with “, ” (comma + space).
+    • Trim leading/trailing whitespace.
     • Escape XML special characters: & → &amp;, < → &lt;, > → &gt;.
 11. **Phone / contact fields** – digits, “+”, “–”, spaces only; if multiple numbers, separate with “, ”.
 12. **Signature fields** – fill ONLY if the signature appears as legible text; if it is handwritten, leave the tag empty.
@@ -535,54 +535,6 @@ def export_json(image_name, result_text):
     except Exception as e:
         return "", f"[Export JSON Failed]: {e}"
 
-# --- 10. Gradio UI ---
-# with gr.Blocks(title="Camel-Doc-OCR") as demo:
-#     gr.Markdown("Camel-Doc-OCR (Qwen2.5-VL, 4-bit)")
-
-#     status_txt = gr.Textbox(label="Status & Memory", interactive=False)
-#     cache_txt = gr.Textbox(label="Cache Stats", interactive=False)
-#     clear_btn = gr.Button("Clear Cache")
-#     clear_btn.click(fn=lambda: (cache_clear(), f"Cache: {len(_mru_cache)}/{CACHE_MAX_SIZE}"), outputs=[cache_txt])
-
-#     file_input = gr.File(label="Tải ảnh hoặc PDF", file_types=[".jpg", ".jpeg", ".png", ".pdf"])
-#     prompt_input = gr.Textbox(label="Prompt thuần", lines=2)
-#     config_input = gr.Textbox(label="JSON Prompt", lines=12)
-
-#     gr.Markdown("Chọn mẫu prompt:")
-#     with gr.Row():
-#         # for key in prompt_templates:
-#         #     btn = gr.Button(f"Mẫu {key}")
-#         #     btn.click(fn=insert_template, inputs=[gr.State(key)], outputs=config_input)
-#         for key in prompt_templates:
-#             gr.Button(f"Mẫu {key}").click(
-#                 fn=lambda k=key: insert_template(k),
-#                 outputs=config_input
-#             )
-
-
-#     run_btn = gr.Button("Chạy OCR")
-#     export_btn = gr.Button("Xuất JSON", visible=False)
-
-#     hidden_name = gr.Textbox(visible=False)
-#     result_output = gr.Textbox(label="Kết quả trích xuất", lines=20)
-#     json_file = gr.File(label="File JSON", visible=False, file_types=[".json"])
-#     json_text = gr.Code(label="JSON Output", language="json", lines=20)
-
-#     # Run inference
-#     run_btn.click(
-#         fn=handle_file,
-#         inputs=[file_input, prompt_input, config_input],
-#         outputs=[hidden_name, result_output]
-#     )
-
-#     # Update memory status
-#     run_btn.click(fn=lambda: get_memory_info(), outputs=[status_txt])
-#     run_btn.click(fn=lambda: f"Cache: {len(_mru_cache)}/{CACHE_MAX_SIZE}", outputs=[cache_txt])
-#     run_btn.click(fn=lambda: gr.update(visible=True), outputs=[export_btn])
-
-#     # Export
-#     export_btn.click(fn=export_json, inputs=[hidden_name, result_output], outputs=[json_file, json_text])
-#     export_btn.click(fn=lambda: gr.update(visible=True), outputs=[json_file])
 # --- 10. Gradio UI ---
 css = """
 .gradio-textbox textarea {
@@ -629,43 +581,43 @@ css = """
 
 with gr.Blocks(title="Camel-Doc-OCR", css=css) as demo:
     gr.Markdown("## 🧾 Camel-Doc-OCR (Qwen2.5-VL, 4-bit)")
-    
+
     # --- Status Bar (Full width) ---
     with gr.Row():
         status_txt = gr.Textbox(label="Status & Memory", interactive=False, scale=2)
         cache_txt = gr.Textbox(label="Cache Stats", interactive=False, scale=1)
         clear_btn = gr.Button("Clear Cache", scale=1)
         clear_btn.click(fn=lambda: (cache_clear(), f"Cache: {len(_mru_cache)}/{CACHE_MAX_SIZE}"), outputs=[cache_txt])
-    
+
     # --- Main Layout: 2 Columns ---
     with gr.Row():
         # === LEFT COLUMN: Input ===
         with gr.Column(scale=1):
             gr.Markdown("### 📥 INPUT")
-            
+
             # File Input
             file_input = gr.File(
-                label="📤 Tải ảnh hoặc PDF", 
+                label="📤 Tải ảnh hoặc PDF",
                 file_types=[".jpg", ".jpeg", ".png", ".pdf"],
                 height=100
             )
-            
+
             # Prompt Input
             prompt_input = gr.Textbox(
-                label="Prompt thuần", 
+                label="Prompt thuần",
                 lines=2,
                 placeholder="Nhập prompt tùy chỉnh...",
                 max_lines=3
             )
-            
+
             # JSON Config
             config_input = gr.Textbox(
-                label="JSON Prompt", 
+                label="JSON Prompt",
                 lines=6,
                 placeholder="Cấu hình JSON sẽ xuất hiện ở đây...",
                 max_lines=8
             )
-            
+
             # Prompt Templates
             gr.Markdown("### 📑 Mẫu:")
             with gr.Row():
@@ -675,64 +627,64 @@ with gr.Blocks(title="Camel-Doc-OCR", css=css) as demo:
                         inputs=[],
                         outputs=config_input
                     )
-            
+
             # Run Button
             run_btn = gr.Button("🚀 Chạy OCR", variant="primary")
-            
+
         # === RIGHT COLUMN: Output ===
         with gr.Column(scale=1):
             gr.Markdown("### 📤 OUTPUT")
-            
+
             # Result Output
             result_output = gr.Textbox(
-                label="Kết quả trích xuất", 
+                label="Kết quả trích xuất",
                 lines=10,
                 placeholder="Kết quả sẽ hiển thị ở đây sau khi chạy OCR...",
                 max_lines=12
             )
-            
+
             # Export Section
             with gr.Row():
                 export_btn = gr.Button("📦 Xuất JSON", visible=False, variant="secondary", size="sm")
-                
+
             # JSON Output
             json_text = gr.Code(
-                label="JSON Output", 
-                language="json", 
+                label="JSON Output",
+                language="json",
                 lines=6,
                 visible=False
             )
-            
+
             # Download File
             json_file = gr.File(
-                label="File JSON để tải", 
-                visible=False, 
+                label="File JSON để tải",
+                visible=False,
                 file_types=[".json"]
             )
-    
+
     # --- Hidden Fields ---
     hidden_name = gr.Textbox(visible=False)
-    
+
     # --- Event Handlers ---
-    
+
     # Run Inference
     run_btn.click(
         fn=handle_file,
         inputs=[file_input, prompt_input, config_input],
         outputs=[hidden_name, result_output]
     )
-    
+
     run_btn.click(fn=get_memory_info, outputs=[status_txt])
     run_btn.click(fn=lambda: f"Cache: {len(_mru_cache)}/{CACHE_MAX_SIZE}", outputs=[cache_txt])
     run_btn.click(fn=lambda: gr.update(visible=True), outputs=[export_btn])
-    
+
     # Export JSON
     export_btn.click(
         fn=export_json,
         inputs=[hidden_name, result_output],
         outputs=[json_file, json_text]
     )
-    
+
     export_btn.click(fn=lambda: gr.update(visible=True), outputs=[json_file])
     export_btn.click(fn=lambda: gr.update(visible=True), outputs=[json_text])