Add telegram bot webhook handler

api/endpoints.py

@@ -41,12 +41,6 @@ BACKUP_HF_TOKEN = os.getenv("BACKUP_HF_TOKEN")
 if not BACKUP_HF_TOKEN:
     logger.warning("BACKUP_HF_TOKEN is not set. Fallback to secondary model will not work if primary token fails.")
 
-
-BACKUP_HF_TOKEN_2 = os.getenv("BACKUP_HF_TOKEN_2")
-if not BACKUP_HF_TOKEN_2:
-    logger.warning("BACKUP_HF_TOKEN_2 is not set. Second backup token not available.")
-
-
 ROUTER_API_URL = os.getenv("ROUTER_API_URL", "https://router.huggingface.co")
 API_ENDPOINT = os.getenv("API_ENDPOINT", "https://router.huggingface.co/v1")
 FALLBACK_API_ENDPOINT = os.getenv("FALLBACK_API_ENDPOINT", "https://api-inference.huggingface.co/v1")

utils/generation.py

@@ -37,7 +37,6 @@ LATEX_DELIMS = [
 # إعداد العميل لـ Hugging Face API
 HF_TOKEN = os.getenv("HF_TOKEN")
 BACKUP_HF_TOKEN = os.getenv("BACKUP_HF_TOKEN")
-BACKUP_HF_TOKEN_2 = os.getenv("BACKUP_HF_TOKEN_2")
 ROUTER_API_URL = os.getenv("ROUTER_API_URL", "https://router.huggingface.co")
 API_ENDPOINT = os.getenv("API_ENDPOINT", "https://router.huggingface.co/v1")
 FALLBACK_API_ENDPOINT = os.getenv("FALLBACK_API_ENDPOINT", "https://api-inference.huggingface.co/v1")
@@ -61,7 +60,7 @@ PROVIDER_ENDPOINTS = {
 }
 
 def check_model_availability(model_name: str, api_key: str) -> tuple[bool, str, str]:
-    """التحقق من توفر النموذج — مع استثناء لنماذج الصور ودعم عدة مفاتيح احتياطية."""
+    """التحقق من توفر النموذج — مع استثناء لنماذج الصور."""
     
     # ✅ القائمة الشاملة لنماذج الصور (تحليل أو توليد)
     IMAGE_MODELS = [
@@ -78,54 +77,28 @@ def check_model_availability(model_name: str, api_key: str) -> tuple[bool, str,
         clean_model_name = model_name.split(":")[0]  # عشان نشيل أي provider مثل :novita
         return True, api_key, f"{IMAGE_INFERENCE_API}/{clean_model_name}"
 
-    # قائمة بكل المفاتيح المتاحة (بالترتيب)
-    available_tokens = []
-    
-    # أضف المفتاح الأساسي إذا كان موجوداً
-    if HF_TOKEN:
-        available_tokens.append(HF_TOKEN)
-    
-    # أضف المفتاح الاحتياطي الأول إذا كان موجوداً
-    if BACKUP_HF_TOKEN:
-        available_tokens.append(BACKUP_HF_TOKEN)
-    
-    # أضف المفتاح الاحتياطي الثاني إذا كان موجوداً (الجديد)
-    if BACKUP_HF_TOKEN_2:
-        available_tokens.append(BACKUP_HF_TOKEN_2)
-    
-    # إذا لم يكن هناك أي مفتاح
-    if not available_tokens:
-        logger.error("❌ No HF tokens are set in environment variables!")
+    # ✅ لو مش صورة — نستخدم الطريقة العادية (للدردشة)
+    try:
+        response = requests.get(
+            f"{ROUTER_API_URL}/v1/models/{model_name}",
+            headers={"Authorization": f"Bearer {api_key}"},
+            timeout=30
+        )
+        logger.debug(f"📡 Checking model {model_name}: {response.status_code} - {response.text}")
+        if response.status_code == 200:
+            logger.info(f"✅ Model {model_name} is available at {API_ENDPOINT}")
+            return True, api_key, API_ENDPOINT
+        elif response.status_code == 429 and BACKUP_HF_TOKEN and api_key != BACKUP_HF_TOKEN:
+            logger.warning(f"⚠️ Rate limit reached for token {api_key}. Switching to backup token.")
+            return check_model_availability(model_name, BACKUP_HF_TOKEN)
+        logger.error(f"❌ Model {model_name} not available: {response.status_code} - {response.text}")
+        return False, api_key, API_ENDPOINT
+    except Exception as e:
+        logger.error(f"🔥 Failed to check model availability for {model_name}: {e}")
+        if BACKUP_HF_TOKEN and api_key != BACKUP_HF_TOKEN:
+            logger.warning(f"🔁 Retrying with backup token for {model_name}")
+            return check_model_availability(model_name, BACKUP_HF_TOKEN)
         return False, api_key, API_ENDPOINT
-
-    # جرب كل المفاتيح بالترتيب
-    for i, token in enumerate(available_tokens):
-        try:
-            logger.debug(f"📡 Checking model {model_name} with token #{i+1} of {len(available_tokens)}")
-            response = requests.get(
-                f"{ROUTER_API_URL}/v1/models/{model_name}",
-                headers={"Authorization": f"Bearer {token}"},
-                timeout=30
-            )
-            logger.debug(f"📡 Response status for token #{i+1}: {response.status_code}")
-            
-            if response.status_code == 200:
-                logger.info(f"✅ Model {model_name} is available with token #{i+1}")
-                return True, token, API_ENDPOINT
-            elif response.status_code == 429:
-                logger.warning(f"⚠️ Rate limit reached for token #{i+1}, trying next token...")
-                continue  # جرب المفتاح التالي
-            else:
-                logger.warning(f"⚠️ Token #{i+1} failed with status {response.status_code}, trying next...")
-                continue
-                
-        except Exception as e:
-            logger.warning(f"🔥 Token #{i+1} error: {e}, trying next...")
-            continue
-    
-    # إذا فشلت جميع المفاتيح
-    logger.error(f"❌ All {len(available_tokens)} tokens failed for model {model_name}")
-    return False, api_key, API_ENDPOINT
 
 def select_model(query: str, input_type: str = "text", preferred_model: Optional[str] = None) -> tuple[str, str]:
     if preferred_model and preferred_model in MODEL_ALIASES:
@@ -506,56 +479,28 @@ def request_generation(
 
     except Exception as e:
         logger.error(f"[Gateway] Streaming failed for model {model_name}: {e}")
-        
-        # ============================================================
-        # محاولة استخدام المفاتيح الاحتياطية بالترتيب
-        # ============================================================
-        
-        # قائمة بكل المفاتيح المتاحة (بعد المفتاح الحالي)
-        available_backup_tokens = []
-        if BACKUP_HF_TOKEN:
-            available_backup_tokens.append(BACKUP_HF_TOKEN)
-        if BACKUP_HF_TOKEN_2:
-            available_backup_tokens.append(BACKUP_HF_TOKEN_2)
-        
-        # جرب كل المفاتيح الاحتياطية بالترتيب
-        backup_used = False
-        for backup_token in available_backup_tokens:
-            if selected_api_key != backup_token:
-                logger.warning(f"🔄 Retrying with backup token for model {model_name}")
-                try:
-                    for chunk in request_generation(
-                        api_key=backup_token,
-                        api_base=selected_endpoint,
-                        message=message,
-                        system_prompt=system_prompt,
-                        model_name=model_name,
-                        chat_history=chat_history,
-                        temperature=temperature,
-                        max_new_tokens=max_new_tokens,
-                        reasoning_effort=reasoning_effort,
-                        tools=tools,
-                        tool_choice=tool_choice,
-                        deep_search=deep_search,
-                        input_type=input_type,
-                        audio_data=audio_data,
-                        image_data=image_data,
-                        output_format=output_format,
-                    ):
-                        yield chunk
-                    backup_used = True
-                    return
-                except Exception as backup_error:
-                    logger.warning(f"⚠️ Backup token also failed: {backup_error}, trying next...")
-                    continue  # جرب المفتاح الاحتياطي التالي
-        
-        if backup_used:
+        if selected_api_key != BACKUP_HF_TOKEN and BACKUP_HF_TOKEN:
+            logger.warning(f"Retrying with backup token for {model_name}")
+            for chunk in request_generation(
+                api_key=BACKUP_HF_TOKEN,
+                api_base=selected_endpoint,
+                message=message,
+                system_prompt=system_prompt,
+                model_name=model_name,
+                chat_history=chat_history,
+                temperature=temperature,
+                max_new_tokens=max_new_tokens,
+                reasoning_effort=reasoning_effort,
+                tools=tools,
+                tool_choice=tool_choice,
+                deep_search=deep_search,
+                input_type=input_type,
+                audio_data=audio_data,
+                image_data=image_data,
+                output_format=output_format,
+            ):
+                yield chunk
             return
-            
-        # ============================================================
-        # إذا فشلت جميع المفاتيح، جرب النماذج الاحتياطية
-        # ============================================================
-        
         if model_name == MODEL_NAME:
             fallback_model = SECONDARY_MODEL_NAME
             fallback_endpoint = FALLBACK_API_ENDPOINT
@@ -725,7 +670,6 @@ def request_generation(
         else:
             yield f"Error: Failed to load model {model_name}: {e}"
             return
-            
 def format_final(analysis_text: str, visible_text: str) -> str:
     reasoning_safe = html.escape((analysis_text or "").strip())
     response = (visible_text or "").strip()

utils/telegram_bot.py

@@ -15,8 +15,7 @@ from fastapi import APIRouter, Request, HTTPException, Header
 from utils.generation import request_generation
 from api.endpoints import enhance_system_prompt
 from utils.constants import API_ENDPOINT, MODEL_NAME
-from utils.generation import HF_TOKEN, BACKUP_HF_TOKEN, BACKUP_HF_TOKEN_2
-
+from utils.generation import HF_TOKEN, BACKUP_HF_TOKEN
 
 logger = logging.getLogger(__name__)
 router = APIRouter()
@@ -428,61 +427,76 @@ def format_ai_response(raw_response: str) -> str:
 # ============================================================
 
 async def call_ai_with_fallback(user_message: str, enhanced_prompt: str) -> str:
-    """Calls the AI, tries all available tokens (primary, backup1, backup2) in sequence."""
-    
-    # قائمة بكل المفاتيح المتاحة (بالترتيب)
-    available_tokens = []
-    
-    if HF_TOKEN:
-        available_tokens.append(HF_TOKEN)
-    if BACKUP_HF_TOKEN:
-        available_tokens.append(BACKUP_HF_TOKEN)
-    if BACKUP_HF_TOKEN_2:
-        available_tokens.append(BACKUP_HF_TOKEN_2)
-    
-    if not available_tokens:
-        logger.error("No HF tokens are set!")
-        return get_error_message("en")
-    
-    last_error = None
+    """Calls the AI, automatically falls back to BACKUP_HF_TOKEN if primary fails."""
+    # First try with primary token
+    primary_token = HF_TOKEN
+    if not primary_token:
+        logger.error("HF_TOKEN is not set!")
+        return get_error_message("en")  # Fallback to error message
     
-    for i, token in enumerate(available_tokens):
-        try:
-            logger.info(f"🔄 Trying token #{i+1} of {len(available_tokens)}...")
-            
-            response_chunks = []
-            stream = request_generation(
-                api_key=token,
-                api_base=API_ENDPOINT,
-                message=user_message,
-                system_prompt=enhanced_prompt,
-                model_name=MODEL_NAME,
-                temperature=0.7,
-                max_new_tokens=1500,
-                deep_search=True,
-                input_type="text",
-                output_format="text"
-            )
-            
-            for chunk in stream:
-                if isinstance(chunk, str) and chunk not in ["analysis", "assistantfinal"]:
-                    response_chunks.append(chunk)
+    try:
+        response_chunks = []
+        stream = request_generation(
+            api_key=primary_token,
+            api_base=API_ENDPOINT,
+            message=user_message,
+            system_prompt=enhanced_prompt,
+            model_name=MODEL_NAME,
+            temperature=0.7,
+            max_new_tokens=1500,
+            deep_search=True,
+            input_type="text",
+            output_format="text"
+        )
+        
+        for chunk in stream:
+            if isinstance(chunk, str) and chunk not in ["analysis", "assistantfinal"]:
+                response_chunks.append(chunk)
+        
+        bot_reply = "".join(response_chunks).strip()
+        if bot_reply and len(bot_reply) >= 5:
+            return bot_reply
             
-            bot_reply = "".join(response_chunks).strip()
-            if bot_reply and len(bot_reply) >= 5:
-                logger.info(f"✅ Token #{i+1} succeeded!")
-                return bot_reply
-            else:
-                raise Exception("Empty or insufficient response")
+        # If we get here, primary failed or returned empty
+        raise Exception("Primary token returned empty or insufficient response")
+        
+    except Exception as primary_error:
+        logger.warning(f"Primary token failed: {primary_error}. Trying backup token...")
+        
+        # If backup token exists, try it
+        if BACKUP_HF_TOKEN:
+            try:
+                response_chunks = []
+                stream = request_generation(
+                    api_key=BACKUP_HF_TOKEN,
+                    api_base=API_ENDPOINT,
+                    message=user_message,
+                    system_prompt=enhanced_prompt,
+                    model_name=MODEL_NAME,
+                    temperature=0.7,
+                    max_new_tokens=1500,
+                    deep_search=True,
+                    input_type="text",
+                    output_format="text"
+                )
                 
-        except Exception as e:
-            logger.warning(f"⚠️ Token #{i+1} failed: {e}")
-            last_error = e
-            continue  # جرب المفتاح التالي
-    
-    # إذا فشلت جميع المفاتيح
-    logger.error(f"❌ All {len(available_tokens)} tokens failed. Last error: {last_error}")
-    raise Exception("All tokens failed.")
+                for chunk in stream:
+                    if isinstance(chunk, str) and chunk not in ["analysis", "assistantfinal"]:
+                        response_chunks.append(chunk)
+                
+                bot_reply = "".join(response_chunks).strip()
+                if bot_reply and len(bot_reply) >= 5:
+                    logger.info("Backup token succeeded.")
+                    return bot_reply
+                else:
+                    raise Exception("Backup token returned empty response")
+                    
+            except Exception as backup_error:
+                logger.error(f"Backup token also failed: {backup_error}")
+                raise Exception("Both primary and backup tokens failed.")
+        else:
+            logger.error("No backup token configured.")
+            raise primary_error
 
             
 # ============================================================