Markdown message (half) support

backend/src/agents/expert.py

@@ -71,6 +71,8 @@ class ExpertAgent(Agent):
         elif model_engine == "replicate":
             if 'parameters' in db_model:
                 opts['parameters'] = db_model['parameters']
+            if 'format' in db_model:
+                opts['format'] = db_model['format']
 
             response = await inference.replicate_completion(self.app, messages, model=db_model.value['model'], opts=opts)
 

backend/src/app_mixins/chat.py

@@ -122,7 +122,7 @@ class ChatMixin:
                 sender=sender,
                 kind=kind or 'message',
                 subkind=subkind,
-                #format=format,
+                format=format,
                 content=initial_content,
                 thread_id=thread_id
             )
@@ -159,7 +159,7 @@ class ChatMixin:
             for agent_id in agent_ids:
                 if str(agent_id) != str(sender):
                     await self.notify_message_to_agent(agent_id, message)
-        
+
         def notify_users():
             nonlocal updating
             self.notify_message_to_user(chat_owner_id, message, updating)

backend/src/chat.py

@@ -8,13 +8,14 @@ import temporary as tmp
 MESSAGE_PAGE_SIZE = 256
 
 class Message:
-    def __init__(self, id, content, sender, time=None, kind='message', subkind=None, thread_id=None):
+    def __init__(self, id, content, sender, time=None, kind='message', subkind=None, format=None, thread_id=None):
         self.id = id
         self.content = content
         self.sender = sender
         self.time = time
         self.kind = kind
         self.subkind = subkind
+        self.format = format
         self.thread_id = thread_id
     
     def from_db_record(record):
@@ -32,6 +33,8 @@ class Message:
             "time": self.time,
             "sender": self.sender,
             "kind": self.kind,
+            "subkind": self.subkind,
+            "format": self.format,
             "content": self.content,
             "threadId": self.thread_id,
         }

backend/src/database.py

@@ -31,7 +31,16 @@ async def init_database (data_directory):
 
 # Checks whether an object is an instance of any database model
 def is_instance (value):
-    isinstance(value, models.Model)
+    import temporary as tmp
+    if hasattr(value, 'id'):
+        if tmp.is_tmp_id(value.id):
+            return False
+
+    if isinstance(value, models.Model):
+        return True
+
+    print("NOT A DB INSTANCE??? " + repr(value))
+    return False
 
 class User(models.Model):
     id = fields.IntField(pk=True)

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))

backend/src/server.py

@@ -12,12 +12,12 @@ from stream import Stream
 # relative to the module file???? wtf
 frontend_path = '../../frontend/dist'
 
-def start_server(app):
+async def start_server(app):
     nest_asyncio.apply()
 
     server = Quart(__name__, static_folder=frontend_path)
     server.secret_key = 'horde-secret'
-    auth_manager = QuartAuth(server, mode="brearer")
+    auth_manager = QuartAuth(server, mode="bearer")
 
     stream_listeners = {}
 
@@ -116,7 +116,7 @@ def start_server(app):
                 return
 
         # TODO: print username instead, same in the finally block
-        print("User logged in:", user_id) 
+        print("User logged in:", user_id)
         await websocket.send(json.dumps({'message': 'authentication successful'}))
 
         send_queue = asyncio.Queue()
@@ -307,4 +307,4 @@ def start_server(app):
             print("User logged out:", user_id)
             app.remove_user_connection(user_id, send_queue)
 
-    server.run(debug=True)
+    server.run(debug=True)