Create app.py

app.py

@@ -0,0 +1,61 @@
+# [سيرفر النموذج] app.py
+import os, json, asyncio
+from fastapi import FastAPI, Request, HTTPException
+from fastapi.responses import StreamingResponse
+from fastapi.middleware.cors import CORSMiddleware
+from huggingface_hub import hf_hub_download
+from llama_cpp import Llama
+from prompts import build_system_prompt
+from search_engine import search_web
+
+app = FastAPI()
+app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])
+
+MODEL_REPO = "bartowski/Qwen_Qwen3.6-35B-A3B-GGUF"
+MODEL_FILE = "Qwen_Qwen3.6-35B-A3B-IQ3_M.gguf"
+llm = None
+
+def load_model():
+    global llm
+    if llm is None:
+        print("⬇️ جاري تحميل النموذج...")
+        model_path = hf_hub_download(repo_id=MODEL_REPO, filename=MODEL_FILE)
+        llm = Llama(
+            model_path=model_path,
+            n_ctx=2048,
+            n_threads=4,
+            n_gpu_layers=0,
+            use_mmap=True,
+            verbose=False
+        )
+        print("✅ تم تحميل النموذج بنجاح.")
+
+@app.on_event("startup")
+def startup():
+    load_model()
+
+async def generate_stream(messages: list, mode: str):
+    system_prompt = build_system_prompt(mode)
+    prompt = f"<|system|>\n{system_prompt}\n<|user|>\n{messages[-1]['content']}\n<|assistant|>\n"
+    
+    if mode == "search":
+        query = messages[-1]['content']
+        search_res = search_web(query)
+        prompt = f"<|system|>\n{system_prompt}\n<|user|>\n{query}\n[SEARCH RESULTS]\n{search_res}\n<|assistant|>\n"
+
+    for token in llm(prompt, max_tokens=1500, stop=["<|user|>", "<|end|>"], stream=True, temperature=0.7):
+        yield json.dumps({"token": token["choices"][0]["text"]}) + "\n"
+        await asyncio.sleep(0.01)
+
+@app.post("/v1/chat/completions")
+async def chat_completions(request: Request):
+    data = await request.json()
+    messages = data.get("messages", [])
+    mode = data.get("mode", "chat")
+    if not messages:
+        raise HTTPException(400, "No messages provided")
+    return StreamingResponse(generate_stream(messages, mode), media_type="application/json")
+
+@app.get("/health")
+def health():
+    return {"status": "ok", "model": MODEL_FILE}