Markdown message (half) support

backend/src/inference.py

@@ -33,8 +33,8 @@ def get_llama_cpp_model(app, model_file):
         return last_model_instance
     
     if last_model_instance is not None:
-        #last_model_instance = None
-        del last_model_instance
+        last_model_instance = None
+        #del last_model_instance
 
     last_model_file = model_file
     last_model_instance = Llama(    
@@ -176,7 +176,7 @@ async def replicate_base_completion(app, model, input):
 
 def run_replicate_inference_loop(replicate_client, model, input, content_stream):
     print("START INFERENCE")
-    completion = replicate_client.run(model, input=input)
+    completion = replicate_client.run(model, input=input, stream=True)
     for event in completion:
         if event:
             content_stream.feed(str(event))