Update app.py

app.py

@@ -1,36 +1,54 @@
 import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch, os
-from fastapi import FastAPI, Request
-import uvicorn
 
+# ✅ 1. تحديد الموديل مع إعدادات آمنة
 MODEL_ID = "Qwen/Qwen2.5-1.5B-Instruct"
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
-model = AutoModelForCausalLM.from_pretrained(MODEL_ID, device_map="auto", torch_dtype=torch.float16)
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_ID,
+    device_map="auto",
+    torch_dtype=torch.float16
+)
+model.eval()
 
+# ✅ 2. قالب التلقين
 PROMPT_TEMPLATE = """You are a GAIA final-answer extractor.
-Extract only what follows "Final Answer:" from the text, or infer if missing.
+Your job is to extract only what comes after 'Final Answer:' in the text below.
+If no 'Final Answer:' is present, infer the most likely short final answer (one line only).
+
 Text:
 {text}
+
+Return only the final answer (no extra words).
 """
 
-def extract_final_answer(raw):
+# ✅ 3. دالة استخراج الإجابة النهائية
+def extract_final_answer(raw: str) -> str:
+    if not raw or not raw.strip():
+        return "No input provided."
     prompt = PROMPT_TEMPLATE.format(text=raw)
     inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
-    out = model.generate(**inputs, max_new_tokens=64, temperature=0.2)
-    return tokenizer.decode(out[0], skip_special_tokens=True).splitlines()[-1]
-
-app = FastAPI()
-
-@app.post("/api/predict")
-async def predict(request: Request):
-    data = await request.json()
-    text = data.get("data", [""])[0]
-    return {"data": [extract_final_answer(text)]}
+    with torch.no_grad():
+        out = model.generate(**inputs, max_new_tokens=64, temperature=0.2)
+    decoded = tokenizer.decode(out[0], skip_special_tokens=True)
+    
+    # نحاول التقاط آخر سطر كناتج نهائي
+    lines = [l.strip() for l in decoded.splitlines() if l.strip()]
+    if not lines:
+        return decoded.strip()
+    return lines[-1]
 
-iface = gr.Interface(fn=extract_final_answer, inputs="text", outputs="text", title="Final Answer Agent")
+# ✅ 4. إعداد واجهة Gradio متوافقة مع Client.predict() في الوكيل الأساسي
+iface = gr.Interface(
+    fn=extract_final_answer,
+    inputs=gr.Textbox(label="Input Text", lines=6, placeholder="Paste the reasoning text from main agent..."),
+    outputs=gr.Textbox(label="Extracted Final Answer", lines=1),
+    title="Final Answer Extractor",
+    description="Extracts the GAIA Final Answer from reasoning text."
+)
 
+# ✅ 5. تشغيل Gradio فقط (بدون FastAPI + Uvicorn)
+#   الوكيل الأساسي يستخدم gradio_client للتواصل، لذلك لا نحتاج FastAPI هنا.
 if __name__ == "__main__":
-    import threading
-    threading.Thread(target=lambda: iface.launch(server_name="0.0.0.0", server_port=7863)).start()
-    uvicorn.run(app, host="0.0.0.0", port=7863)
+    iface.launch(server_name="0.0.0.0", server_port=int(os.getenv("PORT", 7863)))