###

main.py

@@ -17,18 +17,13 @@ if "invoice_counter" not in st.session_state:
     st.session_state.invoice_counter = 0
 
 # ==========================================
-# 1. API-CLIENT FÜR EXTERNE BACKENDS
+# 1. API-CLIENT FÜR DAS vLLM-BACKEND
 # ==========================================
 class LLMClient:
-    """Client für Ollama und vLLM (OpenAI-kompatibel) mit JSON-Schema-Unterstützung."""
     
-    def __init__(self, backend_type="vllm", model="gemma431b", base_url=None):
-        self.backend_type = backend_type.lower()
+    def __init__(self, model="gemma431b", base_url="http://10.192.0.140:8002"):
         self.model = model
         
-        if not base_url:
-            base_url = "http://10.192.0.140:8002" if self.backend_type == "vllm" else "http://localhost:11434"
-            
         base_url = base_url.strip()
         if base_url.startswith("[") and ")" in base_url:
             match = re.search(r"\((https?://.*?)\)", base_url)
@@ -42,31 +37,7 @@ class LLMClient:
             {"role": "system", "content": system_prompt},
             {"role": "user", "content": prompt}
         ]
-
-        if self.backend_type == "ollama":
-            return self._ollama_request(messages, temperature, json_schema)
-        else:
-            return self._openai_compatible_request(messages, temperature, json_schema)
-
-    def _ollama_request(self, messages, temperature, json_schema):
-        url = f"{self.base_url}/api/chat"
-        payload = {
-            "model": self.model,
-            "messages": messages,
-            "stream": False,
-            "options": {
-                "temperature": temperature
-            }
-        }
-        if json_schema:
-            payload["format"] = json_schema
-            
-        try:
-            response = requests.post(url, json=payload, timeout=120)
-            response.raise_for_status()
-            return response.json().get("message", {}).get("content", "")
-        except Exception as e:
-            return f"Ollama Error: {e}"
+        return self._openai_compatible_request(messages, temperature, json_schema)
 
     def _openai_compatible_request(self, messages, temperature, json_schema):
         url = f"{self.base_url}/v1/chat/completions"
@@ -361,8 +332,7 @@ def get_cached_llm():
     if "llm_instance" not in st.session_state:
         llm_url = os.environ.get('VLLM_URL', 'http://10.192.0.140:8002')
         llm_model = os.environ.get('LLM_MODEL', 'gemma431b')
-        llm_backend = os.environ.get('LLM_BACKEND', 'vllm')
-        st.session_state.llm_instance = LLMClient(backend_type=llm_backend, model=llm_model, base_url=llm_url)
+        st.session_state.llm_instance = LLMClient(model=llm_model, base_url=llm_url)
     return st.session_state.llm_instance