Update app.py

app.py

@@ -1,6 +1,6 @@
 """
 Backend API cho HT_MATH_WEB - Chạy trên Hugging Face Spaces (Docker Version)
-Phiên bản: 8.3 (Debug Mode: Route Logging & Dynamic 404 Handler)
+Phiên bản: 8.4 (Refactor with APIRouter & Deep Debugging)
 Tác giả: Hoàng Tấn Thiên
 """
 
@@ -16,7 +16,7 @@ import uuid
 import math
 from typing import List, Optional
 
-from fastapi import FastAPI, File, UploadFile, HTTPException, Form, Request
+from fastapi import FastAPI, File, UploadFile, HTTPException, Form, Request, APIRouter
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse, FileResponse
 from fastapi.staticfiles import StaticFiles
@@ -33,7 +33,7 @@ except ImportError:
 except OSError:
     print("CRITICAL WARNING: pandoc binary not found in system path.")
 
-# --- SUPABASE ---
+# --- SUPABASE SETUP ---
 try:
     from supabase import create_client, Client
     SUPABASE_AVAILABLE = True
@@ -47,10 +47,10 @@ GEMINI_API_KEYS = os.getenv("GEMINI_API_KEYS", "").split(",")
 GEMINI_MODELS = os.getenv("GEMINI_MODELS", "gemini-2.5-flash,gemini-1.5-pro").split(",")
 SUPABASE_URL = os.getenv("SUPABASE_URL", "")
 SUPABASE_KEY = os.getenv("SUPABASE_KEY", "")
-MAX_THREADS = int(os.getenv("MAX_THREADS", "5")) 
+MAX_THREADS = int(os.getenv("MAX_THREADS", "5"))
 ADMIN_SECRET_KEY = os.getenv("ADMIN_SECRET_KEY", "admin123")
 
-# Setup Supabase
+# Setup Supabase Client
 supabase = None
 if SUPABASE_AVAILABLE and SUPABASE_URL and SUPABASE_KEY:
     try:
@@ -58,8 +58,10 @@ if SUPABASE_AVAILABLE and SUPABASE_URL and SUPABASE_KEY:
     except Exception as e:
         print(f"Warning: Không thể kết nối Supabase: {e}")
 
-app = FastAPI(title="HT_MATH_WEB API", version="8.3")
+# ===== KHỞI TẠO APP =====
+app = FastAPI(title="HT_MATH_WEB API", version="8.4")
 
+# CORS Middleware
 app.add_middleware(
     CORSMiddleware,
     allow_origins=["*"],
@@ -68,30 +70,33 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# --- SETUP STATIC FILES ---
+# Static Files
 os.makedirs("uploads", exist_ok=True)
 app.mount("/uploads", StaticFiles(directory="uploads"), name="uploads")
 
-# --- DEBUG: LOG ROUTES ON STARTUP ---
+# ===== ROUTER SETUP (QUAN TRỌNG) =====
+# Gom nhóm tất cả API vào router để đảm bảo prefix luôn đúng
+api_router = APIRouter(prefix="/api")
+
+# ===== LOGGING STARTUP =====
 @app.on_event("startup")
 async def startup_event():
-    print("\n" + "="*50)
-    print(">> SERVER STARTUP: Checking Registered Routes...")
-    routes = [route.path for route in app.routes]
-    for route in routes:
-        print(f"   - {route}")
-    print("="*50 + "\n")
-
+    print("\n" + "="*60)
+    print(">> SERVER V8.4 STARTING UP...")
+    print(">> REGISTERED ROUTES:")
+    for route in app.routes:
+        print(f"   path: {route.path} | name: {route.name} | methods: {route.methods}")
+    print("="*60 + "\n")
+
+# ===== EXCEPTION HANDLERS =====
 @app.exception_handler(404)
 async def not_found_handler(request: Request, exc):
-    # Dynamic route listing for better debugging
-    routes = [route.path for route in request.app.routes]
     return JSONResponse(
         status_code=404,
         content={
             "detail": f"Route not found: {request.url.path}",
-            "available_routes": routes,
-            "suggestion": "Check URL path and HTTP Method (POST/GET)"
+            "method": request.method,
+            "available_routes": [r.path for r in request.app.routes]
         }
     )
 
@@ -113,7 +118,7 @@ class ApiKeyManager:
 key_manager = ApiKeyManager(GEMINI_API_KEYS)
 
 ip_rate_limits = {}
-RATE_LIMIT_DURATION = 7  
+RATE_LIMIT_DURATION = 7
 
 def check_rate_limit(request: Request):
     forwarded = request.headers.get("X-Forwarded-For")
@@ -125,142 +130,46 @@ def check_rate_limit(request: Request):
             print(f"[RateLimit] IP {client_ip} requesting too fast.")
     ip_rate_limits[client_ip] = now
 
-# ===== CHIẾN LƯỢC PROMPT KÉP =====
-
+# ===== HELPER FUNCTIONS =====
 STRONG_PROMPT_LATEX = r"""Role: Chuyên viên nhập liệu Toán học chính xác cao.
 Task: Số hóa lại tài liệu trong ảnh thành Markdown/LaTeX để lưu trữ.
+⚠️ YÊU CẦU: Trích xuất KHÔNG BỎ SÓT. Giữ nguyên định dạng gốc."""
 
-⚠️ YÊU CẦU TUYỆT ĐỐI (CRITICAL):
-1. **TÍNH TOÀN VẸN**: Trích xuất KHÔNG BỎ SÓT bất kỳ từ ngữ, con số, hay câu hỏi nào. Nếu ảnh có 10 câu, phải xuất đủ 10 câu.
-2. **KHÔNG TÓM TẮT**: Tuyệt đối không được tóm tắt hay làm gọn nội dung. Viết lại y hệt bản gốc.
-3. **ĐỊNH DẠNG**: 
-   - Công thức toán để trong $...$ hoặc $$...$$
-   - Giữ nguyên số thứ tự câu (Câu 1, Bài 2...).
-   - In đậm tiêu đề.
+SAFE_PROMPT_LATEX = r"""Role: Trợ lý hỗ trợ người khiếm thị.
+Task: Mô tả lại nội dung văn bản và toán học trong ảnh."""
 
-Đây là tác vụ nhập liệu kỹ thuật số, hãy làm việc như một máy quét."""
+def clean_latex_formulas(text: str) -> str:
+    return re.sub(r'\$\s+(.*?)\s+\$', lambda m: f'${m.group(1).strip()}$', text)
 
-SAFE_PROMPT_LATEX = r"""Role: Trợ lý hỗ trợ người khiếm thị tiếp cận tài liệu.
-Task: Mô tả lại nội dung văn bản và toán học trong ảnh một cách chi tiết để người dùng có thể đọc được.
+def hash_password(password: str) -> str:
+    return hashlib.sha256(password.encode()).hexdigest()
 
-Yêu cầu:
-- Chuyển đổi toàn bộ nội dung sang dạng văn bản và LaTeX.
-- Cố gắng giữ nguyên ý nghĩa và các con số của bài toán.
-- Trình bày rõ ràng, dễ đọc.
-"""
+def verify_password(password: str, hashed: str) -> bool:
+    return hash_password(password) == hashed
 
 # ===== STITCHING ALGORITHM =====
 def stitch_text(text_a: str, text_b: str, min_overlap_chars: int = 20) -> str:
     if not text_a: return text_b
     if not text_b: return text_a
-    
     a_lines = text_a.splitlines()
     b_lines = text_b.splitlines()
-    
     scan_window = min(len(a_lines), len(b_lines), 30) 
-    
     best_overlap_idx = 0
-    
     for i in range(scan_window, 0, -1):
         tail_a = "\n".join(a_lines[-i:]).strip()
         head_b = "\n".join(b_lines[:i]).strip()
-        
         if len(tail_a) >= min_overlap_chars and tail_a == head_b:
             best_overlap_idx = i
             break
-            
     if best_overlap_idx > 0:
         return text_a + "\n" + "\n".join(b_lines[best_overlap_idx:])
     else:
         return text_a + "\n\n" + text_b
 
-# ===== HELPER FUNCTIONS =====
-def clean_latex_formulas(text: str) -> str:
-    return re.sub(r'\$\s+(.*?)\s+\$', lambda m: f'${m.group(1).strip()}$', text)
-
-def hash_password(password: str) -> str:
-    return hashlib.sha256(password.encode()).hexdigest()
-
-def verify_password(password: str, hashed: str) -> bool:
-    return hash_password(password) == hashed
-
-# ===== API ENDPOINTS =====
-
-@app.get("/")
-@app.get("/health")
-async def root():
-    pandoc_status = "Not Found"
-    try:
-        pandoc_status = pypandoc.get_pandoc_version()
-    except:
-        pass
-    return {
-        "status": "ok",
-        "service": "HT_MATH_WEB API v8.3 (Debug Mode)",
-        "keys_loaded": key_manager.get_key_count(),
-        "pandoc_version": pandoc_status
-    }
-
-@app.get("/api/models")
-async def get_models():
-    return {"models": GEMINI_MODELS}
-
-# --- AUTH API ---
-@app.post("/api/register")
-async def register(email: str = Form(...), password: str = Form(...)):
-    if not supabase: raise HTTPException(status_code=500, detail="DB Error")
-    res = supabase.table("users").select("email").eq("email", email).execute()
-    if res.data: raise HTTPException(status_code=400, detail="Email tồn tại")
-    user_data = {"email": email, "password": hash_password(password), "status": "pending", "created_at": time.strftime("%Y-%m-%d %H:%M:%S")}
-    supabase.table("users").insert(user_data).execute()
-    return {"success": True, "message": "Đăng ký thành công, chờ duyệt."}
-
-@app.post("/api/login")
-async def login(request: Request, email: str = Form(...), password: str = Form(...)):
-    if not supabase: raise HTTPException(status_code=500, detail="DB Error")
-    res = supabase.table("users").select("*").eq("email", email).execute()
-    if not res.data: raise HTTPException(status_code=401, detail="Sai email/pass")
-    user = res.data[0]
-    if not verify_password(password, user["password"]): raise HTTPException(status_code=401, detail="Sai email/pass")
-    if user.get("status") != "active": raise HTTPException(status_code=403, detail="Tài khoản chưa kích hoạt")
-    token = secrets.token_urlsafe(32)
-    try: supabase.table("sessions").delete().eq("email", email).execute()
-    except: pass
-    supabase.table("sessions").insert({"email": email, "token": token, "last_seen": time.strftime("%Y-%m-%d %H:%M:%S")}).execute()
-    return {"success": True, "token": token, "email": email}
-
-@app.post("/api/check-session")
-async def check_session(email: str = Form(...), token: str = Form(...)):
-    if not supabase: raise HTTPException(status_code=500, detail="DB Error")
-    res = supabase.table("sessions").select("token").eq("email", email).execute()
-    if not res.data or res.data[0]['token'] != token: raise HTTPException(status_code=401, detail="Session expired")
-    supabase.table("sessions").update({"last_seen": time.strftime("%Y-%m-%d %H:%M:%S")}).eq("email", email).execute()
-    return {"status": "valid"}
-
-@app.post("/api/logout")
-async def logout(request: Request):
-    try:
-        data = await request.json()
-        email = data.get("email")
-        if email and supabase: supabase.table("sessions").delete().eq("email", email).execute()
-    except: pass
-    return {"status": "success"}
-
-@app.post("/api/upload-image")
-async def upload_image(file: UploadFile = File(...)):
-    try:
-        file_ext = os.path.splitext(file.filename)[1] or ".png"
-        file_name = f"{uuid.uuid4().hex}{file_ext}"
-        file_path = f"uploads/{file_name}"
-        with open(file_path, "wb") as f: f.write(await file.read())
-        return {"url": file_path}
-    except Exception as e: raise HTTPException(status_code=500, detail=str(e))
-
-# --- CORE CONVERT LOGIC ---
+# ===== CORE AI LOGIC =====
 
 async def process_image_with_gemini(image: Image.Image, model_id: str, prompt_mode: str, max_retries: int = 3) -> str:
-    current_prompt = STRONG_PROMPT_LATEX if prompt_mode == "latex" else "Trích xuất văn bản chính xác."
-    
+    current_prompt = STRONG_PROMPT_LATEX if prompt_mode == "latex" else "Trích xuất văn bản."
     safety_settings = [
         {"category": "HARM_CATEGORY_HARASSMENT", "threshold": "BLOCK_NONE"},
         {"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "BLOCK_NONE"},
@@ -274,47 +183,34 @@ async def process_image_with_gemini(image: Image.Image, model_id: str, prompt_mo
             api_key = key_manager.get_next_key()
             if not api_key: raise ValueError("No API Key")
             genai.configure(api_key=api_key)
-            
             model = genai.GenerativeModel(model_id, generation_config=generation_config)
             
-            print(f"[Gemini] Sending request (Attempt {attempt+1})...")
             response = model.generate_content([current_prompt, image], safety_settings=safety_settings)
             
             if response.candidates:
                 candidate = response.candidates[0]
                 if candidate.content and candidate.content.parts:
-                    text = response.text.strip()
-                    if len(text) > 10: 
-                        return text
+                    return response.text.strip()
                 
                 finish_reason = candidate.finish_reason
                 print(f"[Gemini] Blocked. Reason: {finish_reason}")
                 
                 if finish_reason == 4 and current_prompt == STRONG_PROMPT_LATEX:
-                    print("[Gemini] Copyright Triggered! Switching to SAFE_PROMPT...")
+                    print("[Gemini] Switching to SAFE_PROMPT...")
                     current_prompt = SAFE_PROMPT_LATEX
-                    continue 
+                    continue
                 
-                if finish_reason == 4:
-                    return "\n> *[Nội dung bị ẩn do bản quyền (Đã thử mọi cách)]*\n"
-                if finish_reason == 3:
-                    return "\n> *[Nội dung bị ẩn do Safety Filter]*\n"
-
-            else:
-                print(f"[Gemini] No candidates returned. Feedback: {response.prompt_feedback}")
-
+                if finish_reason == 4: return "\n> *[Nội dung bị ẩn do bản quyền]*\n"
+                if finish_reason == 3: return "\n> *[Nội dung bị ẩn do Safety Filter]*\n"
         except Exception as e:
             print(f"[Gemini] Error attempt {attempt}: {str(e)}")
-            if "429" in str(e): 
-                time.sleep(2)
-                continue
+            if "429" in str(e): time.sleep(2); continue
             
-    return "" 
+    return ""
 
 async def process_large_image(image: Image.Image, model: str, prompt_mode: str, semaphore: asyncio.Semaphore) -> str:
     CHUNK_HEIGHT = 1536 
     OVERLAP_HEIGHT = 300 
-    
     width, height = image.size
     
     if height <= CHUNK_HEIGHT:
@@ -331,7 +227,7 @@ async def process_large_image(image: Image.Image, model: str, prompt_mode: str,
         if bottom == height: break
         y += (CHUNK_HEIGHT - OVERLAP_HEIGHT)
         
-    print(f"[Split] Image height {height}px -> {len(chunks)} chunks (Aggressive Splitting).")
+    print(f"[Split] Height {height}px -> {len(chunks)} chunks.")
     
     async def process_chunk(chunk_img, index):
         async with semaphore:
@@ -340,17 +236,90 @@ async def process_large_image(image: Image.Image, model: str, prompt_mode: str,
 
     tasks = [process_chunk(chunk, i) for i, chunk in enumerate(chunks)]
     chunk_results = await asyncio.gather(*tasks)
-    
     chunk_results.sort(key=lambda x: x[0])
     ordered_texts = [text for _, text in chunk_results]
     
     final_text = ordered_texts[0]
     for i in range(1, len(ordered_texts)):
         final_text = stitch_text(final_text, ordered_texts[i], min_overlap_chars=20)
-        
     return final_text
 
-@app.post("/api/convert")
+# ===== ROUTES DEFINITION =====
+
+@app.get("/")
+@app.get("/health")
+async def root():
+    return {
+        "status": "ok", 
+        "service": "HT_MATH_WEB API v8.4", 
+        "keys": key_manager.get_key_count()
+    }
+
+# --- DEBUG ROUTE ---
+@app.get("/debug/routes")
+async def debug_routes():
+    """Liệt kê tất cả route đang hoạt động để kiểm tra lỗi 404"""
+    return {
+        "routes": [{"path": route.path, "name": route.name, "methods": list(route.methods)} for route in app.routes]
+    }
+
+# --- API ROUTES (Được gắn vào api_router) ---
+
+@api_router.get("/models")
+async def get_models():
+    return {"models": GEMINI_MODELS}
+
+@api_router.post("/register")
+async def register(email: str = Form(...), password: str = Form(...)):
+    if not supabase: raise HTTPException(status_code=500, detail="DB Error")
+    res = supabase.table("users").select("email").eq("email", email).execute()
+    if res.data: raise HTTPException(status_code=400, detail="Email tồn tại")
+    user_data = {"email": email, "password": hash_password(password), "status": "pending", "created_at": time.strftime("%Y-%m-%d %H:%M:%S")}
+    supabase.table("users").insert(user_data).execute()
+    return {"success": True}
+
+@api_router.post("/login")
+async def login(email: str = Form(...), password: str = Form(...)):
+    if not supabase: raise HTTPException(status_code=500, detail="DB Error")
+    res = supabase.table("users").select("*").eq("email", email).execute()
+    if not res.data: raise HTTPException(status_code=401, detail="Sai email/pass")
+    user = res.data[0]
+    if not verify_password(password, user["password"]): raise HTTPException(status_code=401, detail="Sai email/pass")
+    if user.get("status") != "active": raise HTTPException(status_code=403, detail="Tài khoản chưa kích hoạt")
+    token = secrets.token_urlsafe(32)
+    try: supabase.table("sessions").delete().eq("email", email).execute()
+    except: pass
+    supabase.table("sessions").insert({"email": email, "token": token, "last_seen": time.strftime("%Y-%m-%d %H:%M:%S")}).execute()
+    return {"success": True, "token": token, "email": email}
+
+@api_router.post("/check-session")
+async def check_session(email: str = Form(...), token: str = Form(...)):
+    if not supabase: raise HTTPException(status_code=500, detail="DB Error")
+    res = supabase.table("sessions").select("token").eq("email", email).execute()
+    if not res.data or res.data[0]['token'] != token: raise HTTPException(status_code=401, detail="Session expired")
+    supabase.table("sessions").update({"last_seen": time.strftime("%Y-%m-%d %H:%M:%S")}).eq("email", email).execute()
+    return {"status": "valid"}
+
+@api_router.post("/logout")
+async def logout(request: Request):
+    try:
+        data = await request.json()
+        email = data.get("email")
+        if email and supabase: supabase.table("sessions").delete().eq("email", email).execute()
+    except: pass
+    return {"status": "success"}
+
+@api_router.post("/upload-image")
+async def upload_image(file: UploadFile = File(...)):
+    try:
+        file_ext = os.path.splitext(file.filename)[1] or ".png"
+        file_name = f"{uuid.uuid4().hex}{file_ext}"
+        file_path = f"uploads/{file_name}"
+        with open(file_path, "wb") as f: f.write(await file.read())
+        return {"url": file_path}
+    except Exception as e: raise HTTPException(status_code=500, detail=str(e))
+
+@api_router.post("/convert")
 async def convert_file(
     request: Request,
     file: UploadFile = File(...),
@@ -358,15 +327,12 @@ async def convert_file(
     mode: str = Form("latex")
 ):
     check_rate_limit(request)
-    if key_manager.get_key_count() == 0:
-        raise HTTPException(status_code=500, detail="Chưa cấu hình API Key")
+    if key_manager.get_key_count() == 0: raise HTTPException(status_code=500, detail="Chưa cấu hình API Key")
     
     try:
         file_content = await file.read()
         file_ext = os.path.splitext(file.filename)[1].lower()
-        
         global_semaphore = asyncio.Semaphore(MAX_THREADS)
-        
         results = []
         
         if file_ext == ".pdf":
@@ -374,8 +340,7 @@ async def convert_file(
             async def process_page_wrapper(page, idx):
                 pix = page.get_pixmap(dpi=300)
                 img = Image.open(io.BytesIO(pix.tobytes("png")))
-                text = await process_large_image(img, model, mode, global_semaphore)
-                return idx, text
+                return idx, await process_large_image(img, model, mode, global_semaphore)
             
             tasks = [process_page_wrapper(doc[i], i) for i in range(len(doc))]
             page_results = await asyncio.gather(*tasks)
@@ -384,8 +349,7 @@ async def convert_file(
             
         elif file_ext in [".png", ".jpg", ".jpeg", ".bmp"]:
             img = Image.open(io.BytesIO(file_content))
-            text = await process_large_image(img, model, mode, global_semaphore)
-            results.append(text)
+            results.append(await process_large_image(img, model, mode, global_semaphore))
         else:
              raise HTTPException(status_code=400, detail="Định dạng file không hỗ trợ")
 
@@ -393,34 +357,22 @@ async def convert_file(
         return {"success": True, "result": clean_latex_formulas(final_text)}
         
     except Exception as e:
-        import traceback
-        traceback.print_exc()
+        import traceback; traceback.print_exc()
         raise HTTPException(status_code=500, detail=str(e))
 
-@app.post("/api/export-docx")
+@api_router.post("/export-docx")
 async def export_docx(markdown_text: str = Form(...)):
     try:
         with tempfile.NamedTemporaryFile(suffix=".docx", delete=False) as tmp_file:
             output_filename = tmp_file.name
-
-        pypandoc.convert_text(
-            markdown_text,
-            to='docx',
-            format='markdown',
-            outputfile=output_filename,
-            extra_args=['--standalone']
-        )
-
-        return FileResponse(
-            output_filename,
-            media_type="application/vnd.openxmlformats-officedocument.wordprocessingml.document",
-            filename="Ket_qua_HT_MATH_Pandoc.docx"
-        )
+        pypandoc.convert_text(markdown_text, to='docx', format='markdown', outputfile=output_filename, extra_args=['--standalone'])
+        return FileResponse(output_filename, media_type="application/vnd.openxmlformats-officedocument.wordprocessingml.document", filename="Ket_qua.docx")
     except Exception as e:
-        import traceback
-        traceback.print_exc()
         raise HTTPException(status_code=500, detail=f"Lỗi xuất Word: {str(e)}")
 
+# ===== ĐĂNG KÝ ROUTER VÀO APP =====
+app.include_router(api_router)
+
 if __name__ == "__main__":
     import uvicorn
     uvicorn.run(app, host="0.0.0.0", port=int(os.getenv("PORT", "7860")))