update app

app.py

@@ -78,6 +78,7 @@ MODEL_MAP = {
     "Qwen2.5-VL-7B-Instruct": (processor_m7b, model_m7b),
     "Qwen2.5-VL-3B-Instruct": (processor_x3b, model_x3b),
 }
+
 MODEL_CHOICES = list(MODEL_MAP.keys())
 
 image_examples = [
@@ -211,22 +212,19 @@ EXAMPLE_CARDS_HTML = build_example_cards_html()
 
 
 def load_example_data(idx_str):
-    idx_raw = str(idx_str).strip()
-    if idx_raw == "":
+    if idx_str is None:
+        return json.dumps({"status": "error", "message": "Invalid example index"})
+    idx_str = str(idx_str).strip()
+    if idx_str == "":
         return json.dumps({"status": "error", "message": "Invalid example index"})
-
     try:
-        if "." in idx_raw:
-            idx = int(float(idx_raw))
-        else:
-            idx = int(idx_raw)
+        idx = int(float(idx_str))
     except Exception:
         return json.dumps({"status": "error", "message": "Invalid example index"})
-
     if idx < 0 or idx >= len(all_examples):
         return json.dumps({"status": "error", "message": "Example index out of range"})
-
     ex = all_examples[idx]
+
     payload = {
         "status": "ok",
         "query": ex["query"],
@@ -334,16 +332,17 @@ def generate_image(model_name: str, text: str, image: Image.Image,
 
 
 @spaces.GPU(duration=calc_timeout_generic)
-def generate_pdf(model_name: str, text: str, pages,
+def generate_pdf(model_name: str, text: str, state,
                  max_new_tokens: int = 2048, temperature: float = 0.6,
                  top_p: float = 0.9, top_k: int = 50,
                  repetition_penalty: float = 1.2, gpu_timeout: int = 120):
-    if not pages:
+    if not state or not state["pages"]:
         raise gr.Error("Please upload a PDF file first.")
     processor, model = select_model(model_name)
+    page_images = state["pages"]
     full_response = ""
-    for i, image in enumerate(pages):
-        page_header = f"--- Page {i+1}/{len(pages)} ---\n"
+    for i, image in enumerate(page_images):
+        page_header = f"--- Page {i+1}/{len(page_images)} ---\n"
         yield full_response + page_header
         messages = [{"role": "user", "content": [{"type": "image"}, {"type": "text", "text": text}]}]
         prompt_full = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
@@ -449,7 +448,7 @@ def generate_gif(model_name: str, text: str, gif_path: str,
         torch.cuda.empty_cache()
 
 
-def run_router(tab_kind, model_name, text, image_b64, pdf_b64, gif_b64,
+def run_router(tab_kind, model_name, text, image_b64, pdf_b64, gif_b64, pdf_state_json,
                max_new_tokens_v, temperature_v, top_p_v, top_k_v, repetition_penalty_v, gpu_timeout_v):
     if tab_kind == "pdf":
         temp_pdf_path = b64_to_temp_file(pdf_b64)
@@ -457,10 +456,11 @@ def run_router(tab_kind, model_name, text, image_b64, pdf_b64, gif_b64,
             raise gr.Error("Could not decode uploaded PDF.")
         try:
             pages = convert_pdf_to_images(temp_pdf_path)
+            state = {"pages": pages, "total_pages": len(pages), "current_page_index": 0}
             yield from generate_pdf(
                 model_name=model_name,
                 text=text,
-                pages=pages,
+                state=state,
                 max_new_tokens=max_new_tokens_v,
                 temperature=temperature_v,
                 top_p=top_p_v,
@@ -827,6 +827,7 @@ function init() {
     const runBtnEl = document.getElementById('custom-run-btn');
     const outputArea = document.getElementById('custom-output-textarea');
     const mediaStatus = document.getElementById('sb-media-status');
+    const exampleResultContainer = document.getElementById('example-result-data');
     const pdfPrevBtn = document.getElementById('pdf-prev-btn');
     const pdfNextBtn = document.getElementById('pdf-next-btn');
     const pdfPageInfo = document.getElementById('pdf-page-info');
@@ -842,7 +843,6 @@ function init() {
     let toastTimer = null;
     let examplePoller = null;
     let lastSeenExamplePayload = null;
-    let suppressNextExampleEmpty = false;
 
     function showToast(message, type) {
         let toast = document.getElementById('app-toast');
@@ -901,17 +901,28 @@ function init() {
 
     function setGradioValue(containerId, value) {
         const container = document.getElementById(containerId);
-        if (!container) return;
-        container.querySelectorAll('input, textarea').forEach(el => {
-            if (el.type === 'file' || el.type === 'range' || el.type === 'checkbox') return;
-            const proto = el.tagName === 'TEXTAREA' ? HTMLTextAreaElement.prototype : HTMLInputElement.prototype;
-            const ns = Object.getOwnPropertyDescriptor(proto, 'value');
-            if (ns && ns.set) {
-                ns.set.call(el, value);
-                el.dispatchEvent(new Event('input', {bubbles:true, composed:true}));
-                el.dispatchEvent(new Event('change', {bubbles:true, composed:true}));
-            }
-        });
+        if (!container) return false;
+        const el = container.querySelector('textarea, input');
+        if (!el) return false;
+        const proto = el.tagName === 'TEXTAREA' ? HTMLTextAreaElement.prototype : HTMLInputElement.prototype;
+        const ns = Object.getOwnPropertyDescriptor(proto, 'value');
+        if (ns && ns.set) {
+            ns.set.call(el, value);
+            el.dispatchEvent(new Event('input', {bubbles:true, composed:true}));
+            el.dispatchEvent(new Event('change', {bubbles:true, composed:true}));
+            return true;
+        }
+        return false;
+    }
+
+    function setGradioValueWithRetry(containerId, value, tries = 25, delay = 120) {
+        let attempt = 0;
+        function tick() {
+            attempt += 1;
+            const ok = setGradioValue(containerId, value);
+            if (!ok && attempt < tries) setTimeout(tick, delay);
+        }
+        tick();
     }
 
     function syncMediaToGradio() {
@@ -995,6 +1006,7 @@ function init() {
         mediaState = null;
         renderPreview();
     }
+    window.__clearPreview = clearPreview;
 
     function processFile(file) {
         if (!file) return;
@@ -1106,6 +1118,9 @@ function init() {
         if (mediaState && mediaState.mode !== mode) clearPreview();
     }
 
+    window.__activateModeTab = activateModeTab;
+    window.__activateModelTab = activateModelTab;
+
     if (uploadClick) uploadClick.addEventListener('click', () => fileInput.click());
     if (btnUpload) btnUpload.addEventListener('click', () => fileInput.click());
     if (btnClear) btnClear.addEventListener('click', clearPreview);
@@ -1330,6 +1345,32 @@ function init() {
         }, 150);
     }
 
+    function triggerExampleLoad(idx) {
+        setGradioValueWithRetry('example-result-data', '');
+        setGradioValueWithRetry('example-idx-input', String(idx));
+
+        let triggerAttempts = 0;
+        function tryClick() {
+            triggerAttempts += 1;
+            const idxNow = getValueFromContainer('example-idx-input');
+            const btnWrap = document.getElementById('example-load-btn');
+            const btn = btnWrap ? (btnWrap.querySelector('button') || btnWrap) : null;
+
+            if (idxNow === String(idx) && btn) {
+                btn.click();
+                startExamplePolling();
+                return;
+            }
+            if (triggerAttempts < 25) {
+                setTimeout(tryClick, 120);
+            } else {
+                document.querySelectorAll('.example-card.loading').forEach(c => c.classList.remove('loading'));
+                showToast('Failed to initialize example loader', 'error');
+            }
+        }
+        setTimeout(tryClick, 180);
+    }
+
     document.querySelectorAll('.example-card[data-idx]').forEach(card => {
         card.addEventListener('click', () => {
             const idx = card.getAttribute('data-idx');
@@ -1337,23 +1378,10 @@ function init() {
                 showToast('Invalid example index', 'error');
                 return;
             }
-
             document.querySelectorAll('.example-card.loading').forEach(c => c.classList.remove('loading'));
             card.classList.add('loading');
             showToast('Loading example...', 'info');
-
-            suppressNextExampleEmpty = true;
-            setGradioValue('example-result-data', '');
-            setGradioValue('example-idx-input', String(idx).trim());
-
-            setTimeout(() => {
-                const btn = document.getElementById('example-load-btn');
-                if (btn) {
-                    const b = btn.querySelector('button');
-                    if (b) b.click(); else btn.click();
-                }
-                startExamplePolling();
-            }, 220);
+            triggerExampleLoad(idx);
         });
     });
 
@@ -1361,14 +1389,6 @@ function init() {
     if (observerTarget) {
         const obs = new MutationObserver(() => {
             const current = getValueFromContainer('example-result-data');
-
-            if (suppressNextExampleEmpty && !current) {
-                suppressNextExampleEmpty = false;
-                return;
-            }
-
-            if (!current) return;
-
             if (current && current !== lastSeenExamplePayload) {
                 lastSeenExamplePayload = current;
                 if (examplePoller) {
@@ -1454,6 +1474,7 @@ with gr.Blocks() as demo:
     hidden_image_b64 = gr.Textbox(value="", elem_id="hidden-image-b64", elem_classes="hidden-input", container=False)
     hidden_pdf_b64 = gr.Textbox(value="", elem_id="hidden-pdf-b64", elem_classes="hidden-input", container=False)
     hidden_gif_b64 = gr.Textbox(value="", elem_id="hidden-gif-b64", elem_classes="hidden-input", container=False)
+    hidden_pdf_state = gr.Textbox(value="", elem_id="hidden-pdf-state", elem_classes="hidden-input", container=False)
     prompt = gr.Textbox(value="", elem_id="prompt-gradio-input", elem_classes="hidden-input", container=False)
     hidden_model_name = gr.Textbox(value="Qwen3-VL-4B-Instruct", elem_id="hidden-model-name", elem_classes="hidden-input", container=False)
 
@@ -1466,7 +1487,7 @@ with gr.Blocks() as demo:
 
     result = gr.Textbox(value="", elem_id="gradio-result", elem_classes="hidden-input", container=False)
 
-    example_idx = gr.Textbox(value="", elem_id="example-idx-input", elem_classes="hidden-input", container=False)
+    example_idx = gr.Textbox(value="-1", elem_id="example-idx-input", elem_classes="hidden-input", container=False)
     example_result = gr.Textbox(value="", elem_id="example-result-data", elem_classes="hidden-input", container=False)
     example_load_btn = gr.Button("Load Example", elem_id="example-load-btn")
 
@@ -1631,6 +1652,7 @@ with gr.Blocks() as demo:
             hidden_image_b64,
             hidden_pdf_b64,
             hidden_gif_b64,
+            hidden_pdf_state,
             max_new_tokens,
             temperature,
             top_p,
@@ -1639,7 +1661,7 @@ with gr.Blocks() as demo:
             gpu_duration_state,
         ],
         outputs=[result],
-        js=r"""(mode, model, p, img, pdf, gif, mnt, t, tp, tk, rp, gd) => {
+        js=r"""(mode, model, p, img, pdf, gif, pdfs, mnt, t, tp, tk, rp, gd) => {
             const modelEl = document.querySelector('.model-tab.active');
             const modeEl = document.querySelector('.mode-tab.active');
             const modelVal = modelEl ? modelEl.getAttribute('data-model') : model;
@@ -1665,7 +1687,7 @@ with gr.Blocks() as demo:
                 if (inner) gifVal = inner.value;
             }
 
-            return [modeVal, modelVal, promptVal, imgVal, pdfVal, gifVal, mnt, t, tp, tk, rp, gd];
+            return [modeVal, modelVal, promptVal, imgVal, pdfVal, gifVal, pdfs, mnt, t, tp, tk, rp, gd];
         }""",
     )