little things

OS/01/assistant/assistant.py

@@ -93,21 +93,19 @@ def queue_listener():
             to_user.put(chunk)
             
             # Speak full sentences out loud
-            accumulated_text += chunk["content"]
+            if chunk["type"] == "assistant":
-            sentences = split_into_sentences(accumulated_text)
+                accumulated_text += chunk["content"]
-            if is_full_sentence(sentences[-1]):
+                sentences = split_into_sentences(accumulated_text)
-                for sentence in sentences:
+                if is_full_sentence(sentences[-1]):
-                    for audio_chunk in tts(sentence):
+                    for sentence in sentences:
-                        to_user.put(audio_chunk)
+                        for audio_chunk in tts(sentence):
-                accumulated_text = ""
+                            to_user.put(audio_chunk)
-            else:
+                    accumulated_text = ""
-                for sentence in sentences[:-1]:
+                else:
-                    for audio_chunk in tts(sentence):
+                    for sentence in sentences[:-1]:
-                        to_user.put(audio_chunk)
+                        for audio_chunk in tts(sentence):
-                accumulated_text = sentences[-1]
+                            to_user.put(audio_chunk)
-
+                    accumulated_text = sentences[-1]
-            if chunk["type"] == "message" and "content" in sentence:
-                sentence += chunk.get("content")
             
             # If we have a new message, save our progress and go back to the top
             if not to_assistant.empty():

OS/01/assistant/create_interpreter.py

@@ -87,6 +87,9 @@ Remember: You can run Python code. Be very concise. Ensure that you actually run
         # This is the name that will appear to the LLM.
         name = "python"
 
+        def __init__(self):
+            self.halt = False
+
         def run(self, code):
             """Generator that yields a dictionary in LMC Format."""
 
@@ -98,17 +101,18 @@ Remember: You can run Python code. Be very concise. Ensure that you actually run
             response = requests.post(f"http://localhost:{computer_port}/run", json=data, stream=True)
             # Stream the response
             for chunk in response.iter_content(chunk_size=100000000):
+                if self.halt:
+                    self.halt = False
+                    break
                 if chunk:  # filter out keep-alive new lines
                     yield json.loads(chunk.decode())
 
         def stop(self):
-            """Stops the code."""
+            self.halt = True
-            # Not needed here, because e2b.run_code isn't stateful.
-            pass
 
         def terminate(self):
             """Terminates the entire process."""
-            # Not needed here, because e2b.run_code isn't stateful.
+            # dramatic!!
             pass
 
     interpreter.computer.languages = [Python]