v0.6.1 - reinstated UI, relay > cortex pipeline working

cortex/llm/llm_router.py

@@ -57,18 +57,17 @@ async def call_llm(
         raise RuntimeError(f"Backend '{backend}' missing url/model in env")
 
     # -------------------------------
-    # Provider: VLLM (your MI50)
+    # Provider: MI50 (llama.cpp server)
     # -------------------------------
-    if provider == "vllm":
+    if provider == "mi50":
         payload = {
-            "model": model,
             "prompt": prompt,
-            "max_tokens": max_tokens,
+            "n_predict": max_tokens,
             "temperature": temperature
         }
-        r = requests.post(url, json=payload, timeout=120)
+        r = requests.post(f"{url}/completion", json=payload, timeout=120)
         data = r.json()
-        return data["choices"][0]["text"]
+        return data["content"]
 
     # -------------------------------
     # Provider: OLLAMA (your 3090)