Simulator quiet run and progress bar

sdk/ai/azure-ai-generative/azure/ai/generative/synthetic/simulator/_conversation/conversation.py

@@ -90,8 +90,6 @@ async def simulate_conversation(
     else:
         conversation_id = None
     first_prompt = first_response["samples"][0]
-    logger.info(f"First turn: {first_prompt}")
-
     # Add all generated turns into array to pass for each bot while generating
     # new responses. We add generated response and the person generating it.
     # in the case of the first turn, it is supposed to be the user search query
@@ -115,7 +113,6 @@ async def simulate_conversation(
             current_character_idx = current_turn % len(bots)
             current_bot = bots[current_character_idx]
             # invoke Bot to generate response given the input request
-            logger.info("-- Sending to %s", current_bot.role.value)
             # pass only the last generated turn without passing the bot name.
             response, request, time_taken, full_response = await current_bot.generate_response(
                 session=session,
@@ -137,7 +134,6 @@ async def simulate_conversation(
                     request=request,
                 )
             )
-            logger.info("Last turn: %s", conversation_history[-1])
             if mlflow_logger is not None:
                 logger_tasks.append(  # schedule logging but don't get blocked by it
                     asyncio.create_task(mlflow_logger.log_successful_response(time_taken))

sdk/ai/azure-ai-generative/azure/ai/generative/synthetic/simulator/_conversation/conversation_bot.py

@@ -57,21 +57,13 @@ def __init__(
         self.model = model
 
         self.logger = logging.getLogger(repr(self))
-
         self.conversation_starter = None  # can either be a dictionary or jinja template
         if role == ConversationRole.USER:
             if "conversation_starter" in self.persona_template_args:
                 conversation_starter_content = self.persona_template_args["conversation_starter"]
                 if isinstance(conversation_starter_content, dict):
-                    msg = f"{conversation_starter_content} instead of generating a turn using a LLM"
-                    self.logger.info(
-                        "This simulated bot will use the provided conversation starter (passed in as dictionary): %s",
-                        msg,
-                    )
                     self.conversation_starter = conversation_starter_content
                 else:
-                    msg = f"{repr(conversation_starter_content)[:400]} instead of generating a turn using a LLM"
-                    self.logger.info("This simulated bot will use the provided conversation starter %s", msg)
                     self.conversation_starter = jinja2.Template(
                         conversation_starter_content, undefined=jinja2.StrictUndefined
                     )
@@ -107,12 +99,8 @@ async def generate_response(
         if turn_number == 0 and self.conversation_starter is not None:
             # if conversation_starter is a dictionary, pass it into samples as is
             if isinstance(self.conversation_starter, dict):
-                self.logger.info("Returning conversation starter: %s", self.conversation_starter)
                 samples = [self.conversation_starter]
             else:
-                self.logger.info(
-                    "Returning conversation starter: %s", repr(self.persona_template_args["conversation_starter"])[:400]
-                )
                 samples = [self.conversation_starter.render(**self.persona_template_args)]  # type: ignore[attr-defined]
             time_taken = 0
 

sdk/ai/azure-ai-generative/azure/ai/generative/synthetic/simulator/_model_tools/models.py

@@ -47,8 +47,9 @@ def __init__(self, n_retry, retry_timeout, logger, retry_options=None):
         # Set up async HTTP client with retry
 
         trace_config = TraceConfig()  # set up request logging
-        trace_config.on_request_start.append(self.on_request_start)
-        trace_config.on_request_end.append(self.on_request_end)
+        trace_config.on_request_end.append(self.delete_auth_header)
+        # trace_config.on_request_start.append(self.on_request_start)
+        # trace_config.on_request_end.append(self.on_request_end) 
         if retry_options is None:
             retry_options = RandomRetry(  # set up retry configuration
                 statuses=[104, 408, 409, 424, 429, 500, 502,
@@ -67,6 +68,13 @@ async def on_request_start(self, session, trace_config_ctx, params):
             current_attempt, params.method, params.url
         ))
 
+    async def delete_auth_header(self, session, trace_config_ctx, params):
+        request_headers = dict(params.response.request_info.headers)
+        if "Authorization" in request_headers:
+            del request_headers["Authorization"]
+        if "api-key" in request_headers:
+            del request_headers["api-key"]
+
     async def on_request_end(self, session, trace_config_ctx, params):
         current_attempt = trace_config_ctx.trace_request_ctx["current_attempt"]
         request_headers = dict(params.response.request_info.headers)

sdk/ai/azure-ai-generative/azure/ai/generative/synthetic/simulator/simulator/_callback_conversation_bot.py

@@ -46,6 +46,17 @@ async def generate_response(
                     "id": None,
                     "template_parameters": {}
                 }
+        if not result:
+            result = {
+                "messages": [{
+                    "content": "Callback did not return a response.",
+                    "role": "assistant"
+                }],
+                "finish_reason": ["stop"],
+                "id": None,
+                "template_parameters": {}
+            } 
+
         self.logger.info("Using user provided callback returning response.")
 
         time_taken = 0

sdk/ai/azure-ai-generative/azure/ai/generative/synthetic/simulator/simulator/_utils.py

@@ -1,7 +1,7 @@
 # ---------------------------------------------------------
 # Copyright (c) Microsoft Corporation. All rights reserved.
 # ---------------------------------------------------------
-# pylint: disable=C0303
+# pylint: skip-file
 """
 This module contains a utility class for managing a list of JSON lines.
 """
@@ -52,14 +52,17 @@ def to_eval_qa_json_lines(self):
                     assistant_message = message['content']
                 if 'context' in message:
                     context = message.get("context", None)
-            if user_message and assistant_message:
-                if context:
-                    json_lines += json.dumps({
-                        'question': user_message, 
-                        'answer': assistant_message, 
-                        'context': context}) + "\n"
-                else:
-                    json_lines += json.dumps({
-                        'question': user_message, 
-                        'answer': assistant_message}) + "\n"
+                if user_message and assistant_message:
+                    if context:
+                        json_lines += json.dumps({
+                            'question': user_message, 
+                            'answer': assistant_message, 
+                            'context': context}) + "\n"
+                        user_message = assistant_message = context = None
+                    else:
+                        json_lines += json.dumps({
+                            'question': user_message, 
+                            'answer': assistant_message}) + "\n"
+                        user_message = assistant_message = None
+
         return json_lines