0.2

app.py

@@ -17,9 +17,9 @@ I am capable of ( slowly ) responding to any query you may have, as I am a LLM a
 """
 
 INITIAL_STATE = json.dumps({
-    "path": "bartowski/mamba-2.8b-hf-GGUF",
-    "filename": "mamba-2.8b-hf-Q8_0.gguf",
-    "context": 4096,
+    "path": "Qwen/Qwen1.5-0.5B-Chat-GGUF",
+    "filename": "qwen1_5-0_5b-chat-q8_0.gguf",
+    "context": 32786,
     "messages": [
         {
             "role": "assistant",
@@ -94,11 +94,7 @@ def markdownify_chats(chats):
     """
     print(chats)
     nl = "\n"
-    return f"""# R3BC - Reload / Reboot / Reset Bot Controller
-
-{format_greeting(current_settings['path'], current_settings['filename'], current_settings['context'])}
-
----\n\n\n\n""" + "\n".join([f"### {'R3BC' if msg['role'] == 'assistant' else 'Human'}\n{msg['content'].strip()}\n\n{'---' + nl*2 if msg['role'] == 'assistant' else ''}" for msg in chats['messages']])
+    return f"# R3BC - Simple. Clean. Chat.\n\n" + "\n".join([f"### {'R3BC' if msg['role'] == 'assistant' else 'Human'}\n{msg['content'].strip()}\n\n{'---' + nl*2 if msg['role'] == 'assistant' else ''}" for msg in chats['messages']])
 
 def llm_chat(inpt: str, state_raw: str):
     """
@@ -107,18 +103,18 @@ def llm_chat(inpt: str, state_raw: str):
     print("llm_chat called", inpt, state_raw)
     global llm, current_settings
     if not llm:
-        return "Language model not loaded."
+        return state_raw, ""
     
     # Assuming inpt is a list of messages to process
     if inpt is None or not inpt:
-        return state_raw
+        return state_raw, ""
     try:
         state = json.loads(state_raw)
     except Exception as e:
-        return INITIAL_STATE
+        return INITIAL_STATE, ""
     
     if not "messages" in state:
-        return INITIAL_STATE
+        return INITIAL_STATE, ""
     
     # If the last message is the same as the last message in the state, pretend they sent "...".
     state['messages'].append({
@@ -146,7 +142,11 @@ def llm_chat(inpt: str, state_raw: str):
         },
     ]
     run_messages.extend(state['messages'])
-    
+    if not state['messages'][-1]['role'] == "assistant":
+        run_messages.append({
+            "role": "user",
+            "content": "Continue as though I have not spoken."
+        })
     # Generate a response using the language model
     response = llm.create_chat_completion(messages=run_messages, max_tokens=current_settings['context'], top_k=16, top_p=0.85, temperature=0.369, presence_penalty=1.12, stop=["</s>", "<|im_end|>", "\n\n", "< | im_start | >", "< | im_end | >", "<user>"])['choices'][0]['message']['content']
     print(f"Request: {inpt}\nResponse: {response}")
@@ -180,7 +180,8 @@ def main():
             mdn = gr.Markdown(markdownify_chats(json.loads(INITIAL_STATE)))
         with gr.Row():
             inp = gr.Textbox(placeholder="Enter your message ( Shift+Enter to Send )", lines=2, max_lines=32, label=None, show_label=False, show_copy_button=True)
-            inp.submit(llm_chat, inputs=[inp, inv['jsn']], outputs=[inv['jsn'], inp, mdn])
+            inp.submit(llm_chat, inputs=[inp, inv['jsn']], outputs=[inv['jsn'], inp])
+            inv['jsn'].change(lambda ijn: markdownify_chats(json.loads(ijn)), inputs=[inv['jsn']], outputs=[mdn])
     blk.launch(debug=True, show_api=False)
 
 if __name__ == "__main__":