refactor(ph-ai): conversation stream manager to agent executor

Author: kappa90

ee/api/conversation.py

@@ -21,11 +21,16 @@
 from posthog.exceptions_capture import capture_exception
 from posthog.models.user import User
 from posthog.rate_limit import AIBurstRateThrottle, AISustainedRateThrottle
-from posthog.temporal.ai.conversation import AssistantConversationRunnerWorkflowInputs
+from posthog.temporal.ai.chat_agent import (
+    CHAT_AGENT_STREAM_MAX_LENGTH,
+    CHAT_AGENT_WORKFLOW_TIMEOUT,
+    AssistantConversationRunnerWorkflow,
+    AssistantConversationRunnerWorkflowInputs,
+)
 from posthog.utils import get_instance_region
 
+from ee.hogai.agent.executor import AgentExecutor
 from ee.hogai.api.serializers import ConversationSerializer
-from ee.hogai.stream.conversation_stream import ConversationStreamManager
 from ee.hogai.utils.aio import async_to_sync
 from ee.hogai.utils.sse import AssistantSSESerializer
 from ee.hogai.utils.types.base import AssistantMode
@@ -181,8 +186,10 @@ async def async_stream(
             workflow_inputs: AssistantConversationRunnerWorkflowInputs,
         ) -> AsyncGenerator[bytes, None]:
             serializer = AssistantSSESerializer()
-            stream_manager = ConversationStreamManager(conversation)
-            async for chunk in stream_manager.astream(workflow_inputs):
+            stream_manager = AgentExecutor(
+                conversation, timeout=CHAT_AGENT_WORKFLOW_TIMEOUT, max_length=CHAT_AGENT_STREAM_MAX_LENGTH
+            )
+            async for chunk in stream_manager.astream(AssistantConversationRunnerWorkflow, workflow_inputs):
                 yield serializer.dumps(chunk).encode("utf-8")
 
         return StreamingHttpResponse(
@@ -200,8 +207,8 @@ def cancel(self, request: Request, *args, **kwargs):
             return Response(status=status.HTTP_204_NO_CONTENT)
 
         async def cancel_workflow():
-            conversation_manager = ConversationStreamManager(conversation)
-            await conversation_manager.cancel_conversation()
+            agent_executor = AgentExecutor(conversation)
+            await agent_executor.cancel_workflow()
 
         try:
             asgi_async_to_sync(cancel_workflow)()

ee/api/test/test_conversation.py

@@ -102,7 +102,7 @@ def test_create_conversation(self):
         conversation_id = str(uuid.uuid4())
 
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ) as mock_start_workflow_and_stream:
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -120,7 +120,7 @@ def test_create_conversation(self):
                 # Check that the method was called with workflow_inputs
                 mock_start_workflow_and_stream.assert_called_once()
                 call_args = mock_start_workflow_and_stream.call_args
-                workflow_inputs = call_args[0][0]
+                workflow_inputs = call_args[0][1]
                 self.assertEqual(workflow_inputs.user_id, self.user.id)
                 self.assertEqual(workflow_inputs.is_new_conversation, True)
                 self.assertEqual(workflow_inputs.conversation_id, conversation.id)
@@ -129,7 +129,7 @@ def test_create_conversation(self):
 
     def test_add_message_to_existing_conversation(self):
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ) as mock_start_workflow_and_stream:
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -149,7 +149,7 @@ def test_add_message_to_existing_conversation(self):
                 # Check that the method was called with workflow_inputs
                 mock_start_workflow_and_stream.assert_called_once()
                 call_args = mock_start_workflow_and_stream.call_args
-                workflow_inputs = call_args[0][0]
+                workflow_inputs = call_args[0][1]
                 self.assertEqual(workflow_inputs.user_id, self.user.id)
                 self.assertEqual(workflow_inputs.is_new_conversation, False)
                 self.assertEqual(workflow_inputs.conversation_id, conversation.id)
@@ -183,7 +183,7 @@ def test_invalid_message_format(self):
     def test_rate_limit_burst(self):
         # Create multiple requests to trigger burst rate limit
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ):
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -215,7 +215,7 @@ def test_none_content_with_existing_conversation(self):
             user=self.user, team=self.team, status=Conversation.Status.IN_PROGRESS
         )
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ) as mock_stream_conversation:
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -251,7 +251,7 @@ def test_missing_trace_id(self):
 
     def test_nonexistent_conversation(self):
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ):
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -284,7 +284,7 @@ def test_unauthenticated_request(self):
         )
         self.assertEqual(response.status_code, status.HTTP_401_UNAUTHORIZED)
 
-    @patch("ee.hogai.stream.conversation_stream.ConversationStreamManager.cancel_conversation")
+    @patch("ee.hogai.agent.executor.AgentExecutor.cancel_workflow")
     def test_cancel_conversation(self, mock_cancel):
         conversation = Conversation.objects.create(
             user=self.user,
@@ -327,7 +327,7 @@ def test_cancel_other_teams_conversation(self):
         )
         self.assertEqual(response.status_code, status.HTTP_404_NOT_FOUND)
 
-    @patch("ee.hogai.stream.conversation_stream.ConversationStreamManager.cancel_conversation")
+    @patch("ee.hogai.agent.executor.AgentExecutor.cancel_workflow")
     def test_cancel_conversation_with_async_cleanup(self, mock_cancel):
         """Test that cancel endpoint properly handles async cleanup."""
         conversation = Conversation.objects.create(
@@ -347,7 +347,7 @@ def test_cancel_conversation_with_async_cleanup(self, mock_cancel):
 
         self.assertEqual(response.status_code, status.HTTP_204_NO_CONTENT)
 
-    @patch("ee.hogai.stream.conversation_stream.ConversationStreamManager.cancel_conversation")
+    @patch("ee.hogai.agent.executor.AgentExecutor.cancel_workflow")
     def test_cancel_conversation_async_cleanup_failure(self, mock_cancel):
         """Test cancel endpoint behavior when async cleanup fails."""
         conversation = Conversation.objects.create(
@@ -415,7 +415,7 @@ def test_stream_from_in_progress_conversation(self):
             user=self.user, team=self.team, status=Conversation.Status.IN_PROGRESS
         )
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ) as mock_stream_conversation:
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -435,7 +435,7 @@ def test_resume_generation_from_idle_conversation(self):
         """Test resuming generation from an idle conversation with no new content."""
         conversation = Conversation.objects.create(user=self.user, team=self.team, status=Conversation.Status.IDLE)
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ) as mock_start_workflow_and_stream:
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -453,7 +453,7 @@ def test_resume_generation_from_idle_conversation(self):
                 # Check that the method was called with workflow_inputs
                 mock_start_workflow_and_stream.assert_called_once()
                 call_args = mock_start_workflow_and_stream.call_args
-                workflow_inputs = call_args[0][0]
+                workflow_inputs = call_args[0][1]
                 self.assertEqual(workflow_inputs.user_id, self.user.id)
                 self.assertEqual(workflow_inputs.is_new_conversation, False)
                 self.assertEqual(workflow_inputs.conversation_id, conversation.id)
@@ -595,7 +595,7 @@ def test_billing_context_validation_valid_data(self):
         conversation = Conversation.objects.create(user=self.user, team=self.team)
 
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ) as mock_start_workflow_and_stream:
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -611,15 +611,15 @@ def test_billing_context_validation_valid_data(self):
                 )
                 self.assertEqual(response.status_code, status.HTTP_200_OK)
                 call_args = mock_start_workflow_and_stream.call_args
-                workflow_inputs = call_args[0][0]
+                workflow_inputs = call_args[0][1]
                 self.assertEqual(workflow_inputs.billing_context, self.billing_context)
 
     def test_billing_context_validation_invalid_data(self):
         """Test that invalid billing context data is rejected."""
         conversation = Conversation.objects.create(user=self.user, team=self.team)
 
         with patch(
-            "ee.hogai.stream.conversation_stream.ConversationStreamManager.astream",
+            "ee.hogai.agent.executor.AgentExecutor.astream",
             return_value=_async_generator(),
         ) as mock_start_workflow_and_stream:
             with patch("ee.api.conversation.StreamingHttpResponse", side_effect=self._create_mock_streaming_response):
@@ -635,5 +635,5 @@ def test_billing_context_validation_invalid_data(self):
                 )
                 self.assertEqual(response.status_code, status.HTTP_200_OK)
                 call_args = mock_start_workflow_and_stream.call_args
-                workflow_inputs = call_args[0][0]
+                workflow_inputs = call_args[0][1]
                 self.assertEqual(workflow_inputs.billing_context, None)

ee/hogai/stream/conversation_stream.py renamed to ee/hogai/agent/executor.py

@@ -11,13 +11,12 @@
 
 from posthog.schema import AssistantEventType, FailureMessage
 
-from posthog.temporal.ai.conversation import (
-    AssistantConversationRunnerWorkflow,
-    AssistantConversationRunnerWorkflowInputs,
-)
+from posthog.temporal.ai.base import AgentBaseWorkflow
 from posthog.temporal.common.client import async_connect
 
-from ee.hogai.stream.redis_stream import (
+from ee.hogai.agent.redis_stream import (
+    CONVERSATION_STREAM_MAX_LENGTH,
+    CONVERSATION_STREAM_TIMEOUT,
     ConversationEvent,
     ConversationRedisStream,
     GenerationStatusEvent,
@@ -33,38 +32,44 @@
 logger = structlog.get_logger(__name__)
 
 
-class ConversationStreamManager:
-    """Manages conversation streaming from Redis streams."""
+class AgentExecutor:
+    """Manages executing an agent workflow and streaming the output."""
 
-    def __init__(self, conversation: Conversation) -> None:
+    def __init__(
+        self,
+        conversation: Conversation,
+        timeout: int = CONVERSATION_STREAM_TIMEOUT,
+        max_length: int = CONVERSATION_STREAM_MAX_LENGTH,
+    ) -> None:
         self._conversation = conversation
-        self._redis_stream = ConversationRedisStream(get_conversation_stream_key(conversation.id))
+        self._redis_stream = ConversationRedisStream(
+            get_conversation_stream_key(conversation.id), timeout=timeout, max_length=max_length
+        )
         self._workflow_id = f"conversation-{conversation.id}"
 
-    async def astream(
-        self, workflow_inputs: AssistantConversationRunnerWorkflowInputs
-    ) -> AsyncGenerator[AssistantOutput, Any]:
-        """Stream conversation updates from Redis stream.
+    async def astream(self, workflow: type[AgentBaseWorkflow], inputs: Any) -> AsyncGenerator[AssistantOutput, Any]:
+        """Stream agent workflow updates from Redis stream.
 
         Args:
-            workflow_inputs: Temporal workflow inputs
+            workflow: Agent temporal workflow class
+            inputs: Agent temporal workflow inputs
 
         Returns:
             AssistantOutput generator
         """
         # If this is a reconnection attempt, we resume streaming
         if self._conversation.status != Conversation.Status.IDLE:
-            if workflow_inputs.message is not None:
+            if inputs.message is not None:
                 raise ValueError("Cannot resume streaming with a new message")
             async for chunk in self.stream_conversation():
                 yield chunk
         else:
             # Otherwise, process the new message (new generation) or resume generation (no new message)
-            async for chunk in self.start_workflow(workflow_inputs):
+            async for chunk in self.start_workflow(workflow, inputs):
                 yield chunk
 
     async def start_workflow(
-        self, workflow_inputs: AssistantConversationRunnerWorkflowInputs
+        self, workflow: type[AgentBaseWorkflow], inputs: Any
     ) -> AsyncGenerator[AssistantOutput, Any]:
         try:
             # Delete the stream to ensure we start fresh
@@ -74,8 +79,8 @@ async def start_workflow(
             client = await async_connect()
 
             handle = await client.start_workflow(
-                AssistantConversationRunnerWorkflow.run,
-                workflow_inputs,
+                workflow.run,
+                inputs,
                 id=self._workflow_id,
                 task_queue=settings.MAX_AI_TASK_QUEUE,
                 id_conflict_policy=WorkflowIDConflictPolicy.USE_EXISTING,
@@ -173,7 +178,7 @@ def _failure_message(self) -> AssistantOutput:
         )
         return (AssistantEventType.MESSAGE, failure_message)
 
-    async def cancel_conversation(self) -> None:
+    async def cancel_workflow(self) -> None:
         """Cancel the current conversation and clean up resources.
 
         Raises:

ee/hogai/stream/redis_stream.py renamed to ee/hogai/agent/redis_stream.py

@@ -160,11 +160,18 @@ class StreamError(Exception):
 class ConversationRedisStream:
     """Manages conversation streaming from Redis streams."""
 
-    def __init__(self, stream_key: str):
+    def __init__(
+        self,
+        stream_key: str,
+        timeout: int = CONVERSATION_STREAM_TIMEOUT,
+        max_length: int = CONVERSATION_STREAM_MAX_LENGTH,
+    ):
         self._stream_key = stream_key
         self._redis_client = get_async_client(settings.REDIS_URL)
         self._deletion_lock = asyncio.Lock()
         self._serializer = ConversationStreamSerializer()
+        self._timeout = timeout
+        self._max_length = max_length
 
     async def wait_for_stream(self) -> bool:
         """Wait for stream to be created using linear backoff.
@@ -220,7 +227,7 @@ async def read_stream(
         start_time = asyncio.get_event_loop().time()
 
         while True:
-            if asyncio.get_event_loop().time() - start_time > CONVERSATION_STREAM_TIMEOUT:
+            if asyncio.get_event_loop().time() - start_time > self._timeout:
                 raise StreamError("Stream timeout - conversation took too long to complete")
 
             try:
@@ -283,15 +290,15 @@ async def write_to_stream(
             callback: Callback to trigger after each message is written to the stream
         """
         try:
-            await self._redis_client.expire(self._stream_key, CONVERSATION_STREAM_TIMEOUT)
+            await self._redis_client.expire(self._stream_key, self._timeout)
 
             async for chunk in generator:
                 message = self._serializer.dumps(chunk)
                 if message is not None:
                     await self._redis_client.xadd(
                         self._stream_key,
                         message,
-                        maxlen=CONVERSATION_STREAM_MAX_LENGTH,
+                        maxlen=self._max_length,
                         approximate=True,
                     )
                 if callback:
@@ -303,7 +310,7 @@ async def write_to_stream(
             await self._redis_client.xadd(
                 self._stream_key,
                 completion_message,
-                maxlen=CONVERSATION_STREAM_MAX_LENGTH,
+                maxlen=self._max_length,
                 approximate=True,
             )
 
@@ -314,7 +321,7 @@ async def write_to_stream(
             await self._redis_client.xadd(
                 self._stream_key,
                 message,
-                maxlen=CONVERSATION_STREAM_MAX_LENGTH,
+                maxlen=self._max_length,
                 approximate=True,
             )
             raise StreamError("Failed to write to stream")