Auto-updater, limites CPU 60% et facturation corrigés

modules/ai_openrouter.py

@@ -1,31 +1,15 @@
 import os
 import time
+import json
 from dotenv import load_dotenv
-import openai
 from openai import OpenAI
 from modules.logger import log_event
 
 load_dotenv()
 
 api_key_openrouter = os.getenv("OPENROUTER_API_KEY") or "sk-or-v1-cle-non-definie"
-
-client = OpenAI(
-    base_url="https://openrouter.ai/api/v1",
-    api_key=api_key_openrouter
-)
-
-CATALOGUE_MODELES = {
-    "gratuit": {
-        "openrouter_chat": "google/gemma-4-31b-it:free",
-        "openrouter_code": "cohere/north-mini-code:free",
-        "openrouter_droit": "z-ai/glm-5.2:free"
-    },
-    "premium": {
-        "openrouter_chat": "openai/gpt-4o-mini",
-        "openrouter_code": "anthropic/claude-3.5-sonnet",
-        "openrouter_droit": "mistralai/mistral-large-2411"
-    }
-}
+client = OpenAI(base_url="https://openrouter.ai/api/v1", api_key=api_key_openrouter)
+CATALOGUE_FILE = "/DATA/AppData/MULTI-IA-AETHAS38/catalogue_dynamique.json"
 
 def verifier_cle_api():
     cle = os.getenv("OPENROUTER_API_KEY")
@@ -33,23 +17,42 @@ def verifier_cle_api():
         raise Exception("Clé OPENROUTER_API_KEY absente dans le fichier .env.")
     return cle
 
-def interroger_openrouter(prompt: str, type_modele: str = "openrouter_chat", premium: bool = False, max_retries: int = 2):
+def obtenir_modele_dynamique(premium: bool):
+    """Charge le catalogue dynamiquement selon l'état Premium ou Gratuit."""
+    try:
+        if os.path.exists(CATALOGUE_FILE):
+            with open(CATALOGUE_FILE, "r", encoding="utf-8") as f:
+                data = json.load(f)
+                if premium:
+                    paid_models = data.get("paid_models", [])
+                    claude = [m for m in paid_models if "claude-3.5-sonnet" in m.lower()]
+                    return claude[0] if claude else (paid_models[0] if paid_models else "anthropic/claude-3.5-sonnet")
+                else:
+                    free_models = data.get("free_models", [])
+                    priorites = [m for m in free_models if "gemma" in m.lower() or "nemotron" in m.lower()]
+                    return priorites[0] if priorites else (free_models[0] if free_models else "google/gemma-2-9b-it:free")
+    except Exception:
+        pass
+    # Fallbacks absolus si fichier introuvable
+    return "anthropic/claude-3.5-sonnet" if premium else "google/gemma-2-9b-it:free"
+
+def interroger_openrouter(prompt: str, type_modele: str = "openrouter_chat", premium: bool = False, max_retries: int = 2, chat_id: str = None):
     verifier_cle_api()
-    gamme = "premium" if premium else "gratuit"
-    nom_modele_api = CATALOGUE_MODELES[gamme].get(type_modele, CATALOGUE_MODELES[gamme]["openrouter_chat"])
+    nom_modele_api = obtenir_modele_dynamique(premium)
     nom_affichage = f"OpenRouter ({nom_modele_api.split('/')[-1]})"
     couleur = "#d97757" if premium else "#4a90e2"
+    
+    # Injection du Tracker Financier OpenRouter
+    extra_body_data = {"session_id": chat_id} if chat_id else {}
 
     for tentative in range(max_retries):
         try:
             response = client.chat.completions.create(
                 model=nom_modele_api,
                 messages=[{"role": "user", "content": prompt}],
-                extra_headers={
-                    "HTTP-Referer": "https://aethas38.duckdns.org",
-                    "X-Title": "Aethas38 Multi-IA"
-                },
-                timeout=20.0
+                extra_headers={"HTTP-Referer": "https://aethas38.duckdns.org", "X-Title": "Aethas38 Multi-IA"},
+                extra_body=extra_body_data,
+                timeout=25.0
             )
             texte = response.choices[0].message.content
             usage = response.usage
@@ -60,19 +63,23 @@ def interroger_openrouter(prompt: str, type_modele: str = "openrouter_chat", pre
         except Exception as e:
             log_event("WARNING", "OPENROUTER", f"Tentative {tentative + 1} échouée pour {nom_modele_api} : {e}")
             if tentative == max_retries - 1:
+                # Force la MAJ du catalogue si le modèle crashe
+                from modules.model_updater import trigger_update_now
+                trigger_update_now()
                 raise e
             time.sleep(1)
 
     return None, None, None, None, 0, 0
 
-def interroger_openrouter_multi(prompt: str, type_modele: str = "openrouter_code", premium: bool = False) -> list:
+def interroger_openrouter_multi(prompt: str, type_modele: str = "openrouter_code", premium: bool = False, chat_id: str = None) -> list:
     verifier_cle_api()
+    extra_body_data = {"session_id": chat_id} if chat_id else {}
     
+    modeles = [obtenir_modele_dynamique(premium)]
     if premium:
-        modeles = ["anthropic/claude-3.5-sonnet", "openai/gpt-4o-mini"]
+        modeles.append("openai/gpt-4o-mini")
     else:
-        # Modèles gratuits réactifs sans blocage de quota
-        modeles = ["cohere/north-mini-code:free", "google/gemma-4-31b-it:free"]
+        modeles.append("nvidia/nemotron-3-nano-30b-a3b:free")
 
     reponses = []
     for modele in modeles:
@@ -80,10 +87,8 @@ def interroger_openrouter_multi(prompt: str, type_modele: str = "openrouter_code
             res = client.chat.completions.create(
                 model=modele,
                 messages=[{"role": "user", "content": prompt}],
-                extra_headers={
-                    "HTTP-Referer": "https://aethas38.duckdns.org",
-                    "X-Title": "Aethas38 Multi-IA"
-                },
+                extra_headers={"HTTP-Referer": "https://aethas38.duckdns.org", "X-Title": "Aethas38 Multi-IA"},
+                extra_body=extra_body_data,
                 timeout=15.0
             )
             contenu = res.choices[0].message.content