chore: refactor control loop and delete legacy code (custom_agent and on_message_cb)

Author: philipph-askui

src/askui/agent_base.py

@@ -13,7 +13,6 @@
 from askui.container import telemetry
 from askui.locators.locators import Locator
 from askui.models.shared.agent_message_param import MessageParam
-from askui.models.shared.agent_on_message_cb import OnMessageCb
 from askui.models.shared.conversation import Conversation, Speakers
 from askui.models.shared.settings import (
     ActSettings,
@@ -109,13 +108,12 @@ def __init__(
         self.locate_settings = LocateSettings()
         self.caching_settings = CachingSettings()
 
-    @telemetry.record_call(exclude={"goal", "on_message", "act_settings", "tools"})
+    @telemetry.record_call(exclude={"goal", "act_settings", "tools"})
     @validate_call(config=ConfigDict(arbitrary_types_allowed=True))
     def act(
         self,
         goal: Annotated[str | list[MessageParam], Field(min_length=1)],
         act_settings: ActSettings | None = None,
-        on_message: OnMessageCb | None = None,
         tools: list[Tool] | ToolCollection | None = None,
         caching_settings: CachingSettings | None = None,
     ) -> None:
@@ -134,9 +132,6 @@ def act(
             act_model (ActModel | None, optional): Model to use for this act
                 execution.
                 Overrides the agent's default model if provided.
-            on_message (OnMessageCb | None, optional): Callback for new messages. If
-                it returns `None`, stops and does not add the message. Cannot be used
-                with caching_settings strategy "record" or "both".
             tools (list[Tool] | ToolCollection | None, optional): The tools for the
                 agent. Defaults to default tools depending on the selected model.
             caching_settings (CachingSettings | None, optional): The caching settings
@@ -153,8 +148,6 @@ def act(
             MaxTokensExceededError: If the model reaches the maximum token limit
                 defined in the agent settings.
             ModelRefusalError: If the model refuses to process the request.
-            ValueError: If on_message callback is provided with caching strategy
-                "record" or "both".
 
         Example:
             Basic usage without caching:
@@ -248,9 +241,8 @@ def act(
         self._conversation.cache_manager = cache_manager
 
         # Use conversation-based architecture for execution
-        self._conversation.start(
+        self._conversation.execute_conversation(
             messages=messages,
-            on_message=on_message,
             tools=_tools,
             settings=_act_settings,
         )

src/askui/custom_agent.py

@@ -6,7 +6,6 @@
 
 from askui.locators.locators import Locator
 from askui.models.shared.agent_message_param import MessageParam
-from askui.models.shared.agent_on_message_cb import OnMessageCb
 from askui.models.shared.settings import ActSettings, GetSettings, LocateSettings
 from askui.models.shared.tools import ToolCollection
 from askui.models.types.geometry import Point, PointList
@@ -140,7 +139,6 @@ def act(
                 self,
                 messages: list[MessageParam],
                 model: str,
-                on_message: OnMessageCb | None = None,
                 tools: list[Tool] | None = None,
                 settings: AgentSettings | None = None,
             ) -> None:
@@ -155,7 +153,6 @@ def act(
         self,
         messages: list[MessageParam],
         act_settings: ActSettings,
-        on_message: OnMessageCb | None = None,
         tools: ToolCollection | None = None,
     ) -> None:
         """
@@ -174,14 +171,6 @@ def act(
         Args:
             messages (list[MessageParam]): The message history to start that
                 determines the actions and following messages.
-            on_message (OnMessageCb | None, optional): Callback for new messages
-                from either an assistant/agent or a user (including
-                automatic/programmatic tool use, e.g., taking a screenshot).
-                If it returns `None`, the acting is canceled and `act()` returns
-                immediately. If it returns a `MessageParam`, this `MessageParma` is
-                added to the message history and the acting continues based on the
-                message. The message may be modified by the callback to allow for
-                directing the assistant/agent or tool use.
             tools (ToolCollection | None, optional): The tools for the agent.
                 Defaults to `None`.
             act_settings (ActSettings): The settings for this act operation,