Persistent browser sessions

alembic/versions/2024_12_17_1841-282b0548d443_introduce_persistent_browser_sessions.py

@@ -0,0 +1,45 @@
+"""introduce persistent browser sessions
+
+Revision ID: 282b0548d443
+Revises: 411dd89f3df9
+Create Date: 2024-12-17 18:41:30.400052+00:00
+
+"""
+
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "282b0548d443"
+down_revision: Union[str, None] = "411dd89f3df9"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table(
+        "persistent_browser_sessions",
+        sa.Column("persistent_browser_session_id", sa.String(), nullable=False),
+        sa.Column("organization_id", sa.String(), nullable=False),
+        sa.Column("runnable_type", sa.String(), nullable=False),
+        sa.Column("runnable_id", sa.String(), nullable=False),
+        sa.Column("created_at", sa.DateTime(), nullable=False),
+        sa.Column("modified_at", sa.DateTime(), nullable=False),
+        sa.Column("deleted_at", sa.DateTime(), nullable=True),
+        sa.ForeignKeyConstraint(
+            ["organization_id"],
+            ["organizations.organization_id"],
+        ),
+        sa.PrimaryKeyConstraint("persistent_browser_session_id"),
+    )
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_table("persistent_browser_sessions")
+    # ### end Alembic commands ###

alembic/versions/2024_12_18_0630-065aca30d46d_make_runnable_type_and_runnable_id_.py

@@ -0,0 +1,40 @@
+"""make runnable_type and runnable_id nullable
+
+Revision ID: 065aca30d46d
+Revises: 282b0548d443
+Create Date: 2024-12-18 06:30:46.558730+00:00
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '065aca30d46d'
+down_revision: Union[str, None] = '282b0548d443'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.alter_column('persistent_browser_sessions', 'runnable_type',
+               existing_type=sa.VARCHAR(),
+               nullable=True)
+    op.alter_column('persistent_browser_sessions', 'runnable_id',
+               existing_type=sa.VARCHAR(),
+               nullable=True)
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.alter_column('persistent_browser_sessions', 'runnable_id',
+               existing_type=sa.VARCHAR(),
+               nullable=False)
+    op.alter_column('persistent_browser_sessions', 'runnable_type',
+               existing_type=sa.VARCHAR(),
+               nullable=False)
+    # ### end Alembic commands ###

run_tests.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+
+source "$(poetry env info --path)/bin/activate"
+poetry install
+
+# Run pytest with various options
+poetry run pytest \
+    -v \
+    --asyncio-mode=auto \
+    "$@"  
+
+# Exit with the pytest exit code
+exit $? 

skyvern-frontend/src/initDevCommands.ts

@@ -0,0 +1,148 @@
+import { apiBaseUrl, envCredential } from "./util/env";
+
+export type DevCommands = {
+  createBrowserSession: () => Promise<void>;
+  listBrowserSessions: () => Promise<void>;
+  getBrowserSession: (sessionId: string) => Promise<void>;
+  closeBrowserSessions: () => Promise<void>;
+  setValue: (key: string, value: unknown) => void;
+  getValue: (key: string) => unknown;
+};
+
+export function initDevCommands() {
+  if (!envCredential) {
+    console.warn("envCredential environment variable was not set");
+    return;
+  }
+
+  async function createBrowserSession() {
+    try {
+      const response = await fetch(`${apiBaseUrl}/browser_sessions`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          "X-API-Key": envCredential!,
+        },
+        credentials: "include",
+      });
+
+      if (!response.ok) {
+        throw new Error(
+          `Failed to create browser session: ${response.statusText}`,
+        );
+      }
+
+      const data = await response.json();
+      console.log("Created browser session:", data);
+      return undefined;
+    } catch (error) {
+      console.error("Error creating browser session:", error);
+      throw error;
+    }
+  }
+
+  async function listBrowserSessions() {
+    try {
+      const response = await fetch(`${apiBaseUrl}/browser_sessions`, {
+        method: "GET",
+        headers: {
+          "Content-Type": "application/json",
+          "X-API-Key": envCredential!,
+        },
+        credentials: "include",
+      });
+
+      if (!response.ok) {
+        throw new Error(
+          `Failed to list browser sessions: ${response.statusText}`,
+        );
+      }
+
+      const data = await response.json();
+      console.log("Browser sessions:", data);
+      return undefined;
+    } catch (error) {
+      console.error("Error listing browser sessions:", error);
+      throw error;
+    }
+  }
+
+  async function getBrowserSession(sessionId: string) {
+    try {
+      const response = await fetch(
+        `${apiBaseUrl}/browser_sessions/${sessionId}`,
+        {
+          method: "GET",
+          headers: {
+            "Content-Type": "application/json",
+            "X-API-Key": envCredential!,
+          },
+          credentials: "include",
+        },
+      );
+
+      if (!response.ok) {
+        throw new Error(
+          `Failed to get browser session: ${response.statusText}`,
+        );
+      }
+
+      const data = await response.json();
+      console.log("Browser session:", data);
+      return undefined;
+    } catch (error) {
+      console.error("Error getting browser session:", error);
+      throw error;
+    }
+  }
+
+  async function closeBrowserSessions() {
+    try {
+      const response = await fetch(`${apiBaseUrl}/browser_sessions/close`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          "X-API-Key": envCredential!,
+        },
+        credentials: "include",
+      });
+
+      if (!response.ok) {
+        throw new Error(
+          `Failed to close browser sessions: ${response.statusText}`,
+        );
+      }
+
+      const data = await response.json();
+      console.log("Browser sessions:", data);
+      return undefined;
+    } catch (error) {
+      console.error("Error closing browser sessions:", error);
+      throw error;
+    }
+  }
+
+  function setValue(key: string, value: unknown) {
+    localStorage.setItem(key, JSON.stringify(value));
+  }
+
+  function getValue(key: string) {
+    const value = localStorage.getItem(key);
+    return value ? JSON.parse(value) : null;
+  }
+
+  (window as unknown as Window).devCommands = {
+    createBrowserSession,
+    listBrowserSessions,
+    getBrowserSession,
+    closeBrowserSessions,
+    setValue,
+    getValue,
+  };
+
+  console.log("Dev commands initialized. Available commands:");
+  console.log("- window.devCommands.createBrowserSession()");
+  console.log("- window.devCommands.listBrowserSessions()");
+  console.log("- window.devCommands.getBrowserSession(sessionId)");
+  console.log("- window.devCommands.closeBrowserSessions()");
+}

skyvern-frontend/src/main.tsx

@@ -3,6 +3,10 @@ import ReactDOM from "react-dom/client";
 import App from "./App.tsx";
 import "./index.css";
 
+import { initDevCommands } from "./initDevCommands.ts";
+
+initDevCommands();
+
 ReactDOM.createRoot(document.getElementById("root")!).render(
   <React.StrictMode>
     <App />

skyvern-frontend/src/routes/tasks/create/SavedTaskForm.tsx

@@ -157,6 +157,13 @@ function SavedTaskForm({ initialValues }: Props) {
         })
         .then(() => {
           const taskRequest = createTaskRequestObject(formValues);
+
+          // if (window.devCommands?.getValue("browserSessionId")) {
+          //   taskRequest.browser_session_id = window.devCommands.getValue(
+          //     "browserSessionId",
+          //   ) as unknown as string;
+          // }
+
           const includeOverrideHeader =
             formValues.maxStepsOverride !== null &&
             formValues.maxStepsOverride !== MAX_STEPS_DEFAULT;

skyvern-frontend/src/routes/workflows/RunWorkflowForm.tsx

@@ -70,6 +70,7 @@ type RunWorkflowRequestBody = {
   data: Record<string, unknown>; // workflow parameters and values
   proxy_location: ProxyLocation | null;
   webhook_callback_url?: string | null;
+  browser_session_id?: string | null;
 };
 
 function getRunWorkflowRequestBody(
@@ -88,6 +89,12 @@ function getRunWorkflowRequestBody(
     proxy_location: proxyLocation,
   };
 
+  if (window.devCommands?.getValue("browserSessionId")) {
+    body.browser_session_id = window.devCommands.getValue(
+      "browserSessionId",
+    ) as unknown as string;
+  }
+
   if (webhookCallbackUrl) {
     body.webhook_callback_url = webhookCallbackUrl;
   }

skyvern-frontend/src/window.d.ts

@@ -0,0 +1,9 @@
+import { DevCommands } from "./initDevCommands";
+
+export {};
+
+declare global {
+  interface Window {
+    devCommands: DevCommands;
+  }
+}

skyvern/exceptions.py

@@ -552,3 +552,11 @@ def __init__(self, element_id: str):
         super().__init__(
             f"Select on the dropdown container instead of the option, try again with another element. element_id={element_id}"
         )
+
+
+class BrowserSessionNotFound(SkyvernHTTPException):
+    def __init__(self, browser_session_id: str):
+        super().__init__(
+            f"Browser session not found. browser_session_id={browser_session_id}",
+            status_code=status.HTTP_404_NOT_FOUND,
+        )

skyvern/forge/agent.py

@@ -115,6 +115,7 @@ async def create_task_and_step_from_block(
         workflow_run_context: WorkflowRunContext,
         task_order: int,
         task_retry: int,
+        browser_session_id: str | None = None,
     ) -> tuple[Task, Step]:
         task_block_parameters = task_block.parameters
         navigation_payload = {}
@@ -244,8 +245,10 @@ async def execute_step(
         api_key: str | None = None,
         close_browser_on_completion: bool = True,
         task_block: BaseTaskBlock | None = None,
+        browser_session_id: str | None = None,
     ) -> Tuple[Step, DetailedAgentStepOutput | None, Step | None]:
         workflow_run: WorkflowRun | None = None
+
         if task.workflow_run_id:
             workflow_run = await app.DATABASE.get_workflow_run(workflow_run_id=task.workflow_run_id)
             if workflow_run and workflow_run.status == WorkflowRunStatus.canceled:
@@ -286,6 +289,7 @@ async def execute_step(
                 last_step=step,
                 api_key=api_key,
                 need_call_webhook=True,
+                browser_session_id=browser_session_id,
             )
             return step, None, None
 
@@ -1099,14 +1103,23 @@ async def record_artifacts_after_action(self, task: Task, step: Step, browser_st
             )
 
     async def _initialize_execution_state(
-        self, task: Task, step: Step, workflow_run: WorkflowRun | None = None
+        self,
+        task: Task,
+        step: Step,
+        workflow_run: WorkflowRun | None = None,
+        browser_session_id: str | None = None,
     ) -> tuple[Step, BrowserState, DetailedAgentStepOutput]:
         if workflow_run:
             browser_state = await app.BROWSER_MANAGER.get_or_create_for_workflow_run(
-                workflow_run=workflow_run, url=task.url
+                workflow_run=workflow_run,
+                url=task.url,
+                browser_session_id=browser_session_id,
             )
         else:
-            browser_state = await app.BROWSER_MANAGER.get_or_create_for_task(task)
+            browser_state = await app.BROWSER_MANAGER.get_or_create_for_task(
+                task=task,
+                browser_session_id=browser_session_id,
+            )
         # Initialize video artifact for the task here, afterwards it'll only get updated
         if browser_state and browser_state.browser_artifacts:
             video_artifacts = await app.BROWSER_MANAGER.get_video_artifacts(

skyvern/forge/app.py

@@ -16,6 +16,7 @@
 from skyvern.forge.sdk.workflow.context_manager import WorkflowContextManager
 from skyvern.forge.sdk.workflow.service import WorkflowService
 from skyvern.webeye.browser_manager import BrowserManager
+from skyvern.webeye.persistent_sessions_manager import PersistentSessionsManager
 from skyvern.webeye.scraper.scraper import ScrapeExcludeFunc
 
 SETTINGS_MANAGER = SettingsManager.get_settings()
@@ -37,6 +38,7 @@
 WORKFLOW_CONTEXT_MANAGER = WorkflowContextManager()
 WORKFLOW_SERVICE = WorkflowService()
 AGENT_FUNCTION = AgentFunction()
+PERSISTENT_SESSIONS_MANAGER = PersistentSessionsManager(database=DATABASE)
 scrape_exclude: ScrapeExcludeFunc | None = None
 authentication_function: Callable[[str], Awaitable[Organization]] | None = None
 setup_api_app: Callable[[FastAPI], None] | None = None

skyvern/forge/sdk/core/skyvern_context.py

@@ -12,6 +12,7 @@ class SkyvernContext:
     workflow_run_id: str | None = None
     max_steps_override: int | None = None
     tz_info: ZoneInfo | None = None
+    browser_session_id: str | None = None
     totp_codes: dict[str, str | None] = field(default_factory=dict)
 
     def __repr__(self) -> str: