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

[kappa90]

Problem

ConversationStreamManager is now AgentExecutor, a class to run an agent workflow and stream its content, taking as input the workflow type and workflow inputs. This will allow us to run deep research using a different workflow than the base chat agent one.

Changes

• Renamed ConversationStreamManager to AgentExecutor and moved it from ee/hogai/stream/ to ee/hogai/agent/
• Renamed AssistantConversationRunnerWorkflow to ChatAgentWorkflow for better clarity
• Created a new base class AgentBaseWorkflow to support different types of agent workflows
• Moved Redis stream implementation to the agent package
• Made timeout and max length parameters configurable in the executor

How did you test this code?

• Migrated all existing tests to the new structure

Changelog: Yes

[kappa90]

• feat(ph-ai): agent artifacts schema #41834
• feat(ph-ai): agent artifacts model #41830
• refactor(ph-ai): clean up scaffolding #41713
• refactor(ph-ai): move chat agent specifics to dedicated classes #41697
• refactor(ph-ai): conversation stream manager to agent executor #41613 👈 (View in Graphite)
• feat(ph-ai): agent modes #41284
• master

This stack of pull requests is managed by Graphite. Learn more about stacking.

[greptile-apps]

_{10 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile
React with 👍 or 👎 to share your feedback on this new summary format}

[sortafreel]

The missing part is the consistency test (because you changed the activity name), easy fix.

[skoob13]

Looks good. Let's not merge it into the agent modes branch to keep this isolated.

[Copilot]

Pull Request Overview

This PR refactors the conversation streaming infrastructure to support multiple types of agent workflows. The main change renames ConversationStreamManager to AgentExecutor and introduces a base workflow class to enable different workflow types beyond the basic chat agent.

Key changes:

• Introduced AgentBaseWorkflow as a base class for different agent workflow types
• Refactored ConversationStreamManager to AgentExecutor with configurable timeout and max_length parameters
• Moved Redis stream implementation from ee/hogai/stream/ to ee/hogai/agent/

Reviewed Changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
posthog/temporal/ai/base.py	Adds new AgentBaseWorkflow base class for agent workflows
posthog/temporal/ai/chat_agent.py	Updates imports and constants for chat agent workflow; maintains existing workflow class name
posthog/temporal/ai/init.py	Updates imports to reference chat_agent module
posthog/temporal/tests/ai/test_chat_agent_activity.py	Updates test imports and patches (contains bugs - patching wrong class)
ee/hogai/agent/redis_stream.py	Makes timeout and max_length configurable parameters in ConversationRedisStream
ee/hogai/agent/executor.py	Renames class to AgentExecutor, adds workflow type parameter to support multiple workflows
ee/hogai/agent/test/test_redis_stream.py	Updates import paths for redis_stream module
ee/hogai/agent/test/test_agent_executor.py	Updates test class name and patches for AgentExecutor
ee/api/conversation.py	Updates imports and usage to work with AgentExecutor and new workflow structure
ee/api/test/test_conversation.py	Updates test patches and call argument indices for new method signature

Comments suppressed due to low confidence (8)

posthog/temporal/tests/ai/test_chat_agent_activity.py:106

• The test is patching ChatAgentManager but the actual implementation uses Assistant.create. This patch should be "posthog.temporal.ai.chat_agent.Assistant.create" instead of "posthog.temporal.ai.chat_agent.ChatAgentManager" to properly mock the Assistant creation in the activity.
  posthog/temporal/tests/ai/test_chat_agent_activity.py:152
• The test is patching ChatAgentManager but the actual implementation uses Assistant.create. This patch should be "posthog.temporal.ai.chat_agent.Assistant.create" instead of "posthog.temporal.ai.chat_agent.ChatAgentManager" to properly mock the Assistant creation in the activity.
  ee/hogai/agent/test/test_agent_executor.py:42
• The docstring should be updated to reflect the new class name. It should say "Test AgentExecutor initialization" instead of "Test ConversationStreamManager initialization".
  ee/hogai/agent/executor.py:182
• The method name cancel_workflow is generic but the docstring still says "Cancel the current conversation". While the method name was updated for generality, the docstring should be updated to say "Cancel the current workflow" to match the new naming convention and reflect that this executor can run different types of workflows, not just conversations.
  ee/hogai/agent/executor.py:64
• The method call self.stream_conversation() is using the old conversation-specific naming. For consistency with the refactoring to support multiple workflow types, this should be renamed to something more generic like self.stream_workflow(). However, since stream_conversation is not being changed in this PR, this is noted for future refactoring.
  posthog/temporal/tests/ai/test_chat_agent_activity.py:187
• The test is patching ChatAgentManager but the actual implementation uses Assistant.create. This patch should be "posthog.temporal.ai.chat_agent.Assistant.create" instead of "posthog.temporal.ai.chat_agent.ChatAgentManager" to properly mock the Assistant creation in the activity.
  posthog/temporal/tests/ai/test_chat_agent_activity.py:217
• The test is patching ChatAgentManager but the actual implementation uses Assistant.create. This patch should be "posthog.temporal.ai.chat_agent.Assistant.create" instead of "posthog.temporal.ai.chat_agent.ChatAgentManager" to properly mock the Assistant creation in the activity.
  ee/hogai/agent/executor.py:58
• The docstring should be updated to reflect the new generic nature of the method. It should say "Stream agent workflow updates" instead of just repeating the first line, and the description should clarify that this method can handle different types of agent workflows.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The run method should be marked as abstract using the @abstractmethod decorator from the abc module. Additionally, AgentBaseWorkflow should inherit from ABC to properly define it as an abstract base class. This makes the interface contract clearer and prevents instantiation of the base class.

Suggested change

	from posthog.temporal.common.base import PostHogWorkflow
	class AgentBaseWorkflow(PostHogWorkflow):
	"""Base temporal workflow for processing agents asynchronously."""
	from abc import ABC, abstractmethod
	from posthog.temporal.common.base import PostHogWorkflow
	class AgentBaseWorkflow(PostHogWorkflow, ABC):
	"""Base temporal workflow for processing agents asynchronously."""
	@abstractmethod

[Copilot]

The check for deep research mode was replaced with a NotImplementedError. While this prevents the use of an incomplete feature, the variable is_deep_research is still being extracted from serializer.validated_data on line 140 but is no longer used. Consider removing the unused variable assignment.