main.py

@@ -34,11 +34,13 @@ ENV_FILE = os.path.join(os.path.dirname(__file__), ".env")
 AI_PROVIDER = os.getenv("AI_PROVIDER", "google").lower()
 OPENAI_API_KEY = os.getenv("OPENAI_API_KEY", "")
 GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY", "")
+NVIDIA_API_KEY = os.getenv("NVIDIA_API_KEY", "")
 OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://127.0.0.1:11434/v1")
 
 GOOGLE_MODEL = os.getenv("GOOGLE_MODEL", "gemini-2.0-flash")
 OPENAI_MODEL = os.getenv("OPENAI_MODEL", "gpt-4o")
 OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "llama3")
+NVIDIA_MODEL = os.getenv("NVIDIA_MODEL", "moonshotai/kimi-k2.5")
 
 # --- DATENBANK INITIALISIERUNG (ERWEITERT) ---
 def init_db():
@@ -364,6 +366,7 @@ async def get_settings():
         "provider": AI_PROVIDER,
         "google_model": GOOGLE_MODEL,
         "openai_model": OPENAI_MODEL,
+        "mvidia_model": NVIDIA_MODEL
         "ollama_model": OLLAMA_MODEL,
         "ollama_base_url": OLLAMA_BASE_URL # URL ans Frontend schicken
     }
@@ -371,7 +374,7 @@ async def get_settings():
 @app.post("/api/settings")
 async def update_settings(request: Request):
     # WICHTIG: OLLAMA_BASE_URL als global deklarieren
-    global AI_PROVIDER, GOOGLE_MODEL, OPENAI_MODEL, OLLAMA_MODEL, OLLAMA_BASE_URL
+    global AI_PROVIDER, GOOGLE_MODEL, OPENAI_MODEL, NVIDIA_MODEL, OLLAMA_MODEL, OLLAMA_BASE_URL
     
     data = await request.json()
     provider = data.get("provider")
@@ -388,6 +391,9 @@ async def update_settings(request: Request):
     elif provider == "openai" and model:
         OPENAI_MODEL = model
         set_key(ENV_FILE, "OPENAI_MODEL", model)
+    elif provider == "nvidia" and model:
+        NVIDIA_MODEL = model
+        set_key(ENV_FILE, "NVIDIA_MODEL", model)
     elif provider == "ollama" and model:
         OLLAMA_MODEL = model
         set_key(ENV_FILE, "OLLAMA_MODEL", model)
@@ -421,7 +427,26 @@ async def get_models(provider: str, url: str = None):
             # Nur GPT-Modelle filtern, um die Liste sauber zu halten
             models = [m.id for m in response.data if "gpt" in m.id or "o1" in m.id]
             models.sort()
+        
+        elif provider == "nvidia":
+            if not NVIDIA_API_KEY or "hier" in NVIDIA_API_KEY: return {"models": []}
             
+            # NVIDIA nutzt das OpenAI SDK! 
+            import openai
+            client = openai.AsyncOpenAI(
+                api_key=NVIDIA_API_KEY,
+                base_url="https://integrate.api.nvidia.com/v1"
+            )
+            
+            try:
+                response = await client.models.list()
+                # Bei NVIDIA heißen die Modelle oft nach dem Schema "nvidia/llama-3.1-405b"
+                models = [m.id for m in response.data]
+                models.sort()
+            except Exception as e:
+                print(f"Fehler bei NVIDIA Abfrage: {e}")
+                return {"models": []}
+
         elif provider == "google":
             if not GOOGLE_API_KEY:
                 return {"models": ["API-Key fehlt"]}