fix input text action with no text in GET steps API (#421)

skyvern/exceptions.py

@@ -266,12 +266,7 @@ class UnknownElementTreeFormat(SkyvernException):
 
 class StepTerminationError(SkyvernException):
     def __init__(self, step_id: str, reason: str) -> None:
-        super().__init__(f"Step {step_id} cannot be executed and task is failed. Reason: {reason}")
+        super().__init__(f"Step {step_id} cannot be executed and task is terminated. Reason: {reason}")
-
-
-class StepUnableToExecuteError(SkyvernException):
-    def __init__(self, step_id: str, reason: str) -> None:
-        super().__init__(f"Step {step_id} cannot be executed and task execution is stopped. Reason: {reason}")
 
 
 class UnsupportedActionType(SkyvernException):

skyvern/forge/agent.py

@@ -18,7 +18,6 @@ from skyvern.exceptions import (
     InvalidWorkflowTaskURLState,
     MissingBrowserStatePage,
     StepTerminationError,
-    StepUnableToExecuteError,
     TaskNotFound,
 )
 from skyvern.forge import app
@@ -37,7 +36,6 @@ from skyvern.forge.sdk.workflow.models.workflow import Workflow, WorkflowRun
 from skyvern.webeye.actions.actions import (
     Action,
     ActionType,
-    ActionTypeUnion,
     CompleteAction,
     UserDefinedError,
     WebAction,
@@ -53,7 +51,7 @@ LOG = structlog.get_logger()
 
 
 class ActionLinkedNode:
-    def __init__(self, action: ActionTypeUnion) -> None:
+    def __init__(self, action: Action) -> None:
         self.action = action
         self.next: ActionLinkedNode | None = None
 
@@ -330,16 +328,9 @@ class ForgeAgent:
 
             return step, detailed_output, next_step
         # TODO (kerem): Let's add other exceptions that we know about here as custom exceptions as well
-        except StepUnableToExecuteError:
-            LOG.error(
-                "Step cannot be executed. Task execution stopped",
-                task_id=task.task_id,
-                step_id=step.step_id,
-            )
-            raise
         except StepTerminationError as e:
             LOG.error(
-                "Step cannot be executed. Task failed.",
+                "Step cannot be executed. Task terminated",
                 task_id=task.task_id,
                 step_id=step.step_id,
             )
@@ -841,7 +832,7 @@ class ForgeAgent:
         # Get action results from the last app.SETTINGS.PROMPT_ACTION_HISTORY_WINDOW steps
         steps = await app.DATABASE.get_task_steps(task_id=task.task_id, organization_id=task.organization_id)
         window_steps = steps[-1 * SettingsManager.get_settings().PROMPT_ACTION_HISTORY_WINDOW :]
-        actions_and_results: list[tuple[ActionTypeUnion, list[ActionResult]]] = []
+        actions_and_results: list[tuple[Action, list[ActionResult]]] = []
         for window_step in window_steps:
             if window_step.output and window_step.output.actions_and_results:
                 actions_and_results.extend(window_step.output.actions_and_results)

skyvern/forge/agent_functions.py

@@ -1,6 +1,6 @@
 from playwright.async_api import Page
 
-from skyvern.exceptions import StepUnableToExecuteError
+from skyvern.exceptions import StepTerminationError
 from skyvern.forge import app
 from skyvern.forge.async_operations import AsyncOperation
 from skyvern.forge.sdk.models import Organization, Step, StepStatus
@@ -34,7 +34,7 @@ class AgentFunction:
 
         can_execute = has_valid_task_status and has_valid_step_status and has_no_running_steps
         if not can_execute:
-            raise StepUnableToExecuteError(step_id=step.step_id, reason=f"Cannot execute step. Reasons: {reasons}")
+            raise StepTerminationError(step_id=step.step_id, reason="Cannot execute step. Reasons: {reasons}")
 
     def generate_async_operations(
         self,

skyvern/forge/sdk/db/client.py

@@ -309,7 +309,7 @@ class AgentDB:
                     if status is not None:
                         step.status = status
                     if output is not None:
-                        step.output = output.model_dump()
+                        step.output = output.model_dump(exclude_none=True)
                     if is_last is not None:
                         step.is_last = is_last
                     if retry_index is not None:

skyvern/forge/sdk/routes/agent_protocol.py

@@ -195,7 +195,7 @@ async def execute_agent_task_step(
             )
     step, _, _ = await app.agent.execute_step(current_org, task, step)
     return Response(
-        content=step.model_dump_json() if step else "",
+        content=step.model_dump_json(exclude_none=True) if step else "",
         status_code=200,
         media_type="application/json",
     )
@@ -402,7 +402,7 @@ async def get_agent_task_steps(
     """
     analytics.capture("skyvern-oss-agent-task-steps-get")
     steps = await app.DATABASE.get_task_steps(task_id, organization_id=current_org.organization_id)
-    return ORJSONResponse([step.model_dump() for step in steps])
+    return ORJSONResponse([step.model_dump(exclude_none=True) for step in steps])
 
 
 @base_router.get(

skyvern/webeye/actions/actions.py

@@ -1,6 +1,5 @@
-import abc
 from enum import StrEnum
-from typing import Any, Dict, List
+from typing import Any, Dict
 
 import structlog
 from deprecation import deprecated
@@ -27,17 +26,6 @@ class ActionType(StrEnum):
     SOLVE_CAPTCHA = "solve_captcha"
     TERMINATE = "terminate"
     COMPLETE = "complete"
-    # Note: Remember to update ActionTypeUnion with new actions
-
-
-class Action(BaseModel):
-    action_type: ActionType
-    description: str | None = None
-    reasoning: str | None = None
-
-
-class WebAction(Action, abc.ABC):
-    element_id: str
 
 
 class UserDefinedError(BaseModel):
@@ -46,8 +34,41 @@ class UserDefinedError(BaseModel):
     confidence_float: float = Field(..., ge=0, le=1)
 
 
-class DecisiveAction(Action, abc.ABC):
+class SelectOption(BaseModel):
-    errors: List[UserDefinedError] = []
+    label: str | None
+    value: str | None
+    index: int | None
+
+    def __repr__(self) -> str:
+        return f"SelectOption(label={self.label}, value={self.value}, index={self.index})"
+
+
+class Action(BaseModel):
+    action_type: ActionType
+    description: str | None = None
+    reasoning: str | None = None
+    element_id: str | None = None
+
+    # DecisiveAction (CompleteAction, TerminateAction) fields
+    errors: list[UserDefinedError] | None = None
+    data_extraction_goal: str | None = None
+
+    # WebAction fields
+    file_name: str | None = None
+    file_url: str | None = None
+    download: bool | None = None
+    is_upload_file_tag: bool | None = None
+    text: str | None = None
+    option: SelectOption | None = None
+    is_checked: bool | None = None
+
+
+class WebAction(Action):
+    element_id: str
+
+
+class DecisiveAction(Action):
+    errors: list[UserDefinedError] = []
 
 
 class ClickAction(WebAction):
@@ -55,6 +76,9 @@ class ClickAction(WebAction):
     file_url: str | None = None
     download: bool = False
 
+    def __repr__(self) -> str:
+        return f"ClickAction(element_id={self.element_id}, file_url={self.file_url}, download={self.download})"
+
 
 class InputTextAction(WebAction):
     action_type: ActionType = ActionType.INPUT_TEXT
@@ -90,15 +114,6 @@ class SolveCaptchaAction(Action):
     action_type: ActionType = ActionType.SOLVE_CAPTCHA
 
 
-class SelectOption(BaseModel):
-    label: str | None
-    value: str | None
-    index: int | None
-
-    def __repr__(self) -> str:
-        return f"SelectOption(label={self.label}, value={self.value}, index={self.index})"
-
-
 class SelectOptionAction(WebAction):
     action_type: ActionType = ActionType.SELECT_OPTION
     option: SelectOption
@@ -221,8 +236,8 @@ def parse_action(action: Dict[str, Any], data_extraction_goal: str | None = None
     raise UnsupportedActionType(action_type=action_type)
 
 
-def parse_actions(task: Task, json_response: List[Dict[str, Any]]) -> List[Action]:
+def parse_actions(task: Task, json_response: list[Dict[str, Any]]) -> list[Action]:
-    actions: List[Action] = []
+    actions: list[Action] = []
     for action in json_response:
         try:
             action_instance = parse_action(action=action, data_extraction_goal=task.data_extraction_goal)
@@ -257,7 +272,6 @@ def parse_actions(task: Task, json_response: List[Dict[str, Any]]) -> List[Actio
                 raw_action=action,
                 exc_info=True,
             )
-
     return actions
 
 
@@ -268,20 +282,3 @@ class ScrapeResult(BaseModel):
     """
 
     scraped_data: dict[str, Any] | list[dict[str, Any]]
-
-
-# https://blog.devgenius.io/deserialize-child-classes-with-pydantic-that-gonna-work-784230e1cf83
-ActionTypeUnion = (
-    ClickAction
-    | InputTextAction
-    | UploadFileAction
-    # Deprecated
-    # | DownloadFileAction
-    | SelectOptionAction
-    | CheckboxAction
-    | WaitAction
-    | NullAction
-    | SolveCaptchaAction
-    | TerminateAction
-    | CompleteAction
-)

skyvern/webeye/actions/models.py

@@ -5,7 +5,7 @@ from typing import Any
 from pydantic import BaseModel
 
 from skyvern.forge.sdk.settings_manager import SettingsManager
-from skyvern.webeye.actions.actions import Action, ActionTypeUnion, DecisiveAction, UserDefinedError
+from skyvern.webeye.actions.actions import Action, DecisiveAction, UserDefinedError
 from skyvern.webeye.actions.responses import ActionResult
 from skyvern.webeye.scraper.scraper import ScrapedPage
 
@@ -18,7 +18,7 @@ class AgentStepOutput(BaseModel):
     # Will be deprecated once we move to the new format below
     action_results: list[ActionResult] | None = None
     # Nullable for backwards compatibility, once backfill is done, this won't be nullable anymore
-    actions_and_results: list[tuple[ActionTypeUnion, list[ActionResult]]] | None = None
+    actions_and_results: list[tuple[Action, list[ActionResult]]] | None = None
     errors: list[UserDefinedError] = []
 
     def __repr__(self) -> str:
@@ -38,7 +38,7 @@ class DetailedAgentStepOutput(BaseModel):
     llm_response: dict[str, Any] | None
     actions: list[Action] | None
     action_results: list[ActionResult] | None
-    actions_and_results: list[tuple[ActionTypeUnion, list[ActionResult]]] | None
+    actions_and_results: list[tuple[Action, list[ActionResult]]] | None
 
     class Config:
         exclude = ["scraped_page", "extract_action_prompt"]