vodorod/Dorod-Sky
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Fail task if FailedToNavigateToUrl (#209)

Browse Source
This commit is contained in:
Kerem Yilmaz
2024-04-19 00:12:02 -07:00
committed by GitHub
parent d059619e5e
commit 393b3fba87
2 changed files with 91 additions and 52 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
skyvern/exceptions.py
View File

@@ -126,6 +126,8 @@ class WorkflowParameterNotFound(SkyvernException):
class FailedToNavigateToUrl(SkyvernException): class FailedToNavigateToUrl(SkyvernException):
def __init__(self, url: str, error_message: str) -> None: def __init__(self, url: str, error_message: str) -> None:
self.url = url
self.error_message = error_message
super().__init__(f"Failed to navigate to url {url}. Error message: {error_message}") super().__init__(f"Failed to navigate to url {url}. Error message: {error_message}")

141
skyvern/forge/agent.py
View File

@@ -12,6 +12,7 @@ from playwright.async_api import Page
from skyvern import analytics from skyvern import analytics
from skyvern.exceptions import ( from skyvern.exceptions import (
BrowserStateMissingPage, BrowserStateMissingPage,
FailedToNavigateToUrl,
FailedToSendWebhook, FailedToSendWebhook,
InvalidWorkflowTaskURLState, InvalidWorkflowTaskURLState,
MissingBrowserStatePage, MissingBrowserStatePage,
@@ -309,6 +310,28 @@ class ForgeAgent(Agent):
step=step, step=step,
) )
return step, detailed_output, next_step return step, detailed_output, next_step
except FailedToNavigateToUrl as e:
# Fail the task if we can't navigate to the URL and send the response
LOG.error(
"Failed to navigate to URL, marking task as failed, and sending webhook response",
task_id=task.task_id,
step_id=step.step_id,
url=e.url,
error_message=e.error_message,
)
task = await self.update_task(
task,
status=TaskStatus.failed,
failure_reason=f"Failed to navigate to URL. URL:{e.url}, Error:{e.error_message}",
)
await self.send_task_response(
task=task,
last_step=step,
api_key=api_key,
close_browser_on_completion=close_browser_on_completion,
skip_artifacts=True,
)
return step, detailed_output, next_step
async def agent_step( async def agent_step(
self, self,
@@ -711,6 +734,7 @@ class ForgeAgent(Agent):
last_step: Step, last_step: Step,
api_key: str | None = None, api_key: str | None = None,
close_browser_on_completion: bool = True, close_browser_on_completion: bool = True,
skip_artifacts: bool = False,
) -> None: ) -> None:
""" """
send the task response to the webhook callback url send the task response to the webhook callback url
@@ -727,6 +751,14 @@ class ForgeAgent(Agent):
task = refreshed_task task = refreshed_task
# log the task status as an event # log the task status as an event
analytics.capture("skyvern-oss-agent-task-status", {"status": task.status}) analytics.capture("skyvern-oss-agent-task-status", {"status": task.status})
# We skip the artifacts and send the webhook response directly only when there is an issue with the browser
# initialization. In this case, we don't have any artifacts to send and we can't take final screenshots etc.
# since the browser is not initialized properly or the proxy is not working.
if skip_artifacts:
await app.ARTIFACT_MANAGER.wait_for_upload_aiotasks_for_task(task.task_id)
await self.execute_task_webhook(task=task, last_step=last_step, api_key=api_key)
return
# Take one last screenshot and create an artifact before closing the browser to see the final state # Take one last screenshot and create an artifact before closing the browser to see the final state
browser_state: BrowserState = await app.BROWSER_MANAGER.get_or_create_for_task(task) browser_state: BrowserState = await app.BROWSER_MANAGER.get_or_create_for_task(task)
await browser_state.get_or_create_page() await browser_state.get_or_create_page()
@@ -760,7 +792,9 @@ class ForgeAgent(Agent):
await self.execute_task_webhook(task=task, last_step=last_step, api_key=api_key) await self.execute_task_webhook(task=task, last_step=last_step, api_key=api_key)
async def execute_task_webhook(self, task: Task, last_step: Step, api_key: str | None) -> None: async def execute_task_webhook(
self, task: Task, last_step: Step, api_key: str | None, skip_artifacts: bool = False
) -> None:
if not api_key: if not api_key:
LOG.warning( LOG.warning(
"Request has no api key. Not sending task response", "Request has no api key. Not sending task response",
@@ -775,65 +809,68 @@ class ForgeAgent(Agent):
) )
return return
# get the artifact of the screenshot and get the screenshot_url if not skip_artifacts:
screenshot_artifact = await app.DATABASE.get_artifact( # get the artifact of the screenshot and get the screenshot_url
task_id=task.task_id, screenshot_artifact = await app.DATABASE.get_artifact(
step_id=last_step.step_id, task_id=task.task_id,
artifact_type=ArtifactType.SCREENSHOT_FINAL, step_id=last_step.step_id,
organization_id=task.organization_id, artifact_type=ArtifactType.SCREENSHOT_FINAL,
) organization_id=task.organization_id,
screenshot_url = None )
if screenshot_artifact: screenshot_url = None
screenshot_url = await app.ARTIFACT_MANAGER.get_share_link(screenshot_artifact) if screenshot_artifact:
screenshot_url = await app.ARTIFACT_MANAGER.get_share_link(screenshot_artifact)
recording_artifact = await app.DATABASE.get_artifact( recording_artifact = await app.DATABASE.get_artifact(
task_id=task.task_id, task_id=task.task_id,
step_id=last_step.step_id, step_id=last_step.step_id,
artifact_type=ArtifactType.RECORDING, artifact_type=ArtifactType.RECORDING,
organization_id=task.organization_id, organization_id=task.organization_id,
) )
recording_url = None recording_url = None
if recording_artifact: if recording_artifact:
recording_url = await app.ARTIFACT_MANAGER.get_share_link(recording_artifact) recording_url = await app.ARTIFACT_MANAGER.get_share_link(recording_artifact)
# get the artifact of the last TASK_RESPONSE_ACTION_SCREENSHOT_COUNT screenshots and get the screenshot_url # get the artifact of the last TASK_RESPONSE_ACTION_SCREENSHOT_COUNT screenshots and get the screenshot_url
latest_action_screenshot_artifacts = await app.DATABASE.get_latest_n_artifacts( latest_action_screenshot_artifacts = await app.DATABASE.get_latest_n_artifacts(
task_id=task.task_id, task_id=task.task_id,
organization_id=task.organization_id, organization_id=task.organization_id,
artifact_types=[ArtifactType.SCREENSHOT_ACTION], artifact_types=[ArtifactType.SCREENSHOT_ACTION],
n=SettingsManager.get_settings().TASK_RESPONSE_ACTION_SCREENSHOT_COUNT, n=SettingsManager.get_settings().TASK_RESPONSE_ACTION_SCREENSHOT_COUNT,
) )
latest_action_screenshot_urls = [] latest_action_screenshot_urls = []
if latest_action_screenshot_artifacts: if latest_action_screenshot_artifacts:
for artifact in latest_action_screenshot_artifacts: for artifact in latest_action_screenshot_artifacts:
screenshot_url = await app.ARTIFACT_MANAGER.get_share_link(artifact) screenshot_url = await app.ARTIFACT_MANAGER.get_share_link(artifact)
if screenshot_url: if screenshot_url:
latest_action_screenshot_urls.append(screenshot_url) latest_action_screenshot_urls.append(screenshot_url)
else: else:
LOG.error( LOG.error(
"Failed to get share link for action screenshot", "Failed to get share link for action screenshot",
artifact_id=artifact.artifact_id, artifact_id=artifact.artifact_id,
) )
else:
LOG.error("Failed to get latest action screenshots")
# get the latest task from the db to get the latest status, extracted_information, and failure_reason
task_from_db = await app.DATABASE.get_task(task_id=task.task_id, organization_id=task.organization_id)
if not task_from_db:
LOG.error("Failed to get task from db when sending task response")
raise TaskNotFound(task_id=task.task_id)
task = task_from_db
task_response = task.to_task_response(
action_screenshot_urls=latest_action_screenshot_urls,
screenshot_url=screenshot_url,
recording_url=recording_url,
)
else: else:
LOG.error("Failed to get latest action screenshots") task_response = task.to_task_response()
# get the latest task from the db to get the latest status, extracted_information, and failure_reason
task_from_db = await app.DATABASE.get_task(task_id=task.task_id, organization_id=task.organization_id)
if not task_from_db:
LOG.error("Failed to get task from db when sending task response")
raise TaskNotFound(task_id=task.task_id)
task = task_from_db
if not task.webhook_callback_url: if not task.webhook_callback_url:
LOG.info("Task has no webhook callback url. Not sending task response") LOG.info("Task has no webhook callback url. Not sending task response")
return return
task_response = task.to_task_response(
action_screenshot_urls=latest_action_screenshot_urls,
screenshot_url=screenshot_url,
recording_url=recording_url,
)
# send task_response to the webhook callback url # send task_response to the webhook callback url
# TODO: use async requests (httpx) # TODO: use async requests (httpx)
timestamp = str(int(datetime.utcnow().timestamp())) timestamp = str(int(datetime.utcnow().timestamp()))