debug "answer()"

app.py

@@ -132,43 +132,46 @@ def store_doc(doc_text: str, user_id="demo"):
     """UI callback: take the textbox content and shove it into the KB."""
     n = add_docs(user_id, [doc_text])
     if n == 0:
-        return "⚠️ Nothing stored (empty input)."
-    return f"📚 Stored ✅ — KB now has {len(kb[user_id]['texts'])} passage(s)."
+        return "Nothing stored (empty input)."
+    return f"Stored — KB now has {len(kb[user_id]['texts'])} passage(s)."
 
 def answer(question: str, user_id="demo"):
     """UI callback: retrieve, build prompt with Qwen tags, generate answer."""
-    if not question.strip():
-        return "⚠️ Please ask a question."
-    if not kb[user_id]["texts"]:
-        return "⚠️ No reference passage yet. Add one first."
-
-    # 1️⃣  Retrieve top-k similar passages
-    q_vec  = embed(question)
-    store  = kb[user_id]
-    sims   = torch.matmul(store["vecs"], q_vec)          # [N]
-    k      = min(4, sims.numel())
-    idxs   = torch.topk(sims, k=k).indices.tolist()
-    context = "\n".join(store["texts"][i] for i in idxs)
-
-    # 2️⃣  Build a Qwen-chat prompt (helper defined earlier)
-    prompt = build_qwen_prompt(context, question)
-
-    # 3️⃣  Generate and strip everything before the assistant tag
-    load_chat()
-    inputs = tokenizer(prompt, return_tensors="pt").to(chat_model.device)
-    output = chat_model.generate(**inputs, max_new_tokens=512)
-    full   = tokenizer.decode(output[0], skip_special_tokens=True)
-    reply  = full.split("<|im_start|>assistant")[-1].strip()
-    return reply
+    try:
+        if not question.strip():
+            return "Please ask a question."
+        if not kb[user_id]["texts"]:
+            return "No reference passage yet. Add one first."
+
+        # 1️⃣  Retrieve top-k similar passages
+        q_vec  = embed(question)
+        store  = kb[user_id]
+        sims   = torch.matmul(store["vecs"], q_vec)          # [N]
+        k      = min(4, sims.numel())
+        idxs   = torch.topk(sims, k=k).indices.tolist()
+        context = "\n".join(store["texts"][i] for i in idxs)
+
+        # 2️⃣  Build a Qwen-chat prompt (helper defined earlier)
+        prompt = build_qwen_prompt(context, question)
+
+        # 3️⃣  Generate and strip everything before the assistant tag
+        load_chat()
+        inputs = tokenizer(prompt, return_tensors="pt").to(chat_model.device)
+        output = chat_model.generate(**inputs, max_new_tokens=512)
+        full   = tokenizer.decode(output[0], skip_special_tokens=True)
+        reply  = full.split("<|im_start|>assistant")[-1].strip()
+        return reply
+    except Exception as e:
+        return f"Error in app.py: {e}"
 
 # ---- UI layout (feel free to tweak cosmetics) -----------------------------
 with gr.Blocks() as demo:
-    gr.Markdown("### 📥 Tiny-RAG playground &nbsp;–&nbsp; 1) paste a passage → store&nbsp;&nbsp; 2) ask a question")
+    gr.Markdown("### Tiny-RAG playground &nbsp;–&nbsp; 1) paste a passage → store&nbsp;&nbsp; 2) ask a question")
 
     # ---- passage ingestion ----
     with gr.Row():
         passage_box = gr.Textbox(lines=6, label="Reference passage")
-        store_btn   = gr.Button("➕ Store passage")
+        store_btn   = gr.Button("Store passage")
     status_box = gr.Markdown()
     store_btn.click(fn=store_doc,
                     inputs=passage_box,
@@ -176,7 +179,7 @@ with gr.Blocks() as demo:
 
     # ---- Q & A ----
     question_box = gr.Textbox(lines=2, label="Ask a question")
-    answer_btn   = gr.Button("🤖 Answer")
+    answer_btn   = gr.Button("Answer")
     answer_box   = gr.Textbox(lines=6, label="Assistant reply")
 
     answer_btn.click(fn=answer,