Update app.py

app.py

@@ -1458,8 +1458,14 @@ class Hive:
         if hasattr(self, 'client') and self.client: # Remote Inference
             stop_sequences = ["</s>", "Assistant:"] + [self.tok.decode(st) for st in self.stop_tokens]
             try:
-                for token in self.client.text_generation(prompt, max_new_tokens=int(max_new_tokens), temperature=float(temperature), do_sample=True, stop_sequences=stop_sequences, stream=True): # type: ignore
-                    yield token
+                messages = [{"role": "user", "content": prompt}]
+                for chunk in self.client.chat_completion(
+                    messages=messages, max_tokens=int(max_new_tokens), temperature=float(temperature), 
+                    do_sample=True, stop=stop_sequences, stream=True
+                ):
+                    content = chunk.choices[0].delta.content
+                    if content:
+                        yield content
             except Exception as e:
                 print(f"[ModelBridge] Remote inference stream failed: {e}")
                 yield "[Error: Could not get response from remote model]"
@@ -1485,7 +1491,7 @@ class Hive:
             yield new_text
 
     def chat(self, message:str, effective_role:str, caller_id: Optional[str],
-             k:int=None, max_new_tokens:int=256, temperature:float=None, prompt_override: Optional[str] = None) -> str: # type: ignore
+             k:int=None, max_new_tokens:int=1024, temperature:float=None, prompt_override: Optional[str] = None) -> str: # type: ignore
         temp = temperature if temperature is not None else (self.decoding_temperature if not self.lite_mode else 0.7)
         
         # This logic was previously in _prepare_chat_input
@@ -1865,7 +1871,7 @@ def launch_ui(bootstrap_instance: "Bootstrap"):
 
                 eff_role = role if mode == "admin" else "user"
                 final_message, intent = hive_instance._prepare_chat_input(user_text, "en", False, None) # type: ignore
-                max_tokens = 512 if intent == "essay_review" else 256
+                max_tokens = 1024 if intent == "essay_review" else 1024 # Increased for longer responses
                 full_prompt = hive_instance.compiler.compile(final_message, [], intent=intent) # type: ignore
 
                 full_reply = ""