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

browser sessions v2 - backend (#4515)

Browse Source 
Signed-off-by: Benji Visser <benji@093b.org>
This commit is contained in:
Benji Visser
2026-01-21 22:27:16 -05:00
committed by GitHub
parent f781a6f0ef
commit b5ff547a3a
15 changed files with 273 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

66
skyvern/forge/sdk/artifact/storage/local.py
View File

@@ -289,14 +289,72 @@ class LocalStorage(BaseStorage):
return []
async def list_recordings_in_browser_session(self, organization_id: str, browser_session_id: str) -> list[str]:
"""List all recording files for a browser session (not implemented for local storage)."""
return []
"""List all recording files for a browser session from local storage.
Videos are synced to the browser_sessions storage path when the session closes.
"""
videos_base = (
Path(self.artifact_path)
/ settings.ENV
/ organization_id
/ "browser_sessions"
/ browser_session_id
/ "videos"
)
recording_files: list[str] = []
if videos_base.exists():
for root, _, files in os.walk(videos_base):
for file in files:
file_path = Path(root) / file
recording_files.append(f"file://{file_path}")
return recording_files
async def get_shared_recordings_in_browser_session(
self, organization_id: str, browser_session_id: str
) -> list[FileInfo]:
"""Get recording files with URLs for a browser session (not implemented for local storage)."""
return []
"""Get recording files with URLs for a browser session from local storage."""
file_uris = await self.list_recordings_in_browser_session(organization_id, browser_session_id)
if not file_uris:
return []
file_infos: list[FileInfo] = []
for uri in file_uris:
uri_lower = uri.lower()
if not (uri_lower.endswith(".webm") or uri_lower.endswith(".mp4")):
LOG.warning(
"Skipping recording file with unsupported extension",
uri=uri,
organization_id=organization_id,
browser_session_id=browser_session_id,
)
continue
file_path = parse_uri_to_path(uri)
path_obj = Path(file_path)
if not path_obj.exists():
continue
file_size = path_obj.stat().st_size
if file_size == 0:
continue
modified_at = datetime.fromtimestamp(path_obj.stat().st_mtime)
checksum = calculate_sha256_for_file(file_path)
filename = path_obj.name
file_info = FileInfo(
url=uri,
checksum=checksum,
filename=filename,
modified_at=modified_at,
)
file_infos.append(file_info)
file_infos.sort(key=lambda f: (f.modified_at is not None, f.modified_at), reverse=True)
return file_infos
async def get_downloaded_files(self, organization_id: str, run_id: str | None) -> list[FileInfo]:
download_dir = get_download_dir(run_id=run_id)

46
skyvern/forge/sdk/db/agent_db.py
View File

@@ -4705,6 +4705,7 @@ class AgentDB(BaseAlchemyDB):
status: str | None = None,
timeout_minutes: int | None = None,
organization_id: str | None = None,
completed_at: datetime | None = None,
) -> PersistentBrowserSession:
try:
async with self.Session() as session:
@@ -4723,6 +4724,8 @@ class AgentDB(BaseAlchemyDB):
persistent_browser_session.status = status
if timeout_minutes:
persistent_browser_session.timeout_minutes = timeout_minutes
if completed_at:
persistent_browser_session.completed_at = completed_at
await session.commit()
await session.refresh(persistent_browser_session)
@@ -4741,7 +4744,7 @@ class AgentDB(BaseAlchemyDB):
self,
browser_session_id: str,
browser_address: str | None,
ip_address: str,
ip_address: str | None,
ecs_task_arn: str | None,
organization_id: str | None = None,
) -> None:
@@ -4779,6 +4782,47 @@ class AgentDB(BaseAlchemyDB):
LOG.error("UnexpectedError", exc_info=True)
raise
async def update_persistent_browser_session_compute_cost(
self,
session_id: str,
organization_id: str,
instance_type: str,
vcpu_millicores: int,
memory_mb: int,
duration_ms: int,
compute_cost: float,
) -> None:
"""Update the compute cost fields for a persistent browser session"""
try:
async with self.Session() as session:
persistent_browser_session = (
await session.scalars(
select(PersistentBrowserSessionModel)
.filter_by(persistent_browser_session_id=session_id)
.filter_by(organization_id=organization_id)
.filter_by(deleted_at=None)
)
).first()
if persistent_browser_session:
persistent_browser_session.instance_type = instance_type
persistent_browser_session.vcpu_millicores = vcpu_millicores
persistent_browser_session.memory_mb = memory_mb
persistent_browser_session.duration_ms = duration_ms
persistent_browser_session.compute_cost = compute_cost
await session.commit()
await session.refresh(persistent_browser_session)
else:
raise NotFoundError(f"PersistentBrowserSession {session_id} not found")
except NotFoundError:
LOG.error("NotFoundError", exc_info=True)
raise
except SQLAlchemyError:
LOG.error("SQLAlchemyError", exc_info=True)
raise
except Exception:
LOG.error("UnexpectedError", exc_info=True)
raise
async def mark_persistent_browser_session_deleted(self, session_id: str, organization_id: str) -> None:
"""Mark a persistent browser session as deleted."""
try:

5
skyvern/forge/sdk/db/models.py
View File

@@ -856,6 +856,11 @@ class PersistentBrowserSessionModel(Base):
proxy_location = Column(String, nullable=True)
extensions = Column(JSON, nullable=True)
browser_type = Column(String, nullable=True)
instance_type = Column(String, nullable=True)
vcpu_millicores = Column(Integer, nullable=True)
memory_mb = Column(Integer, nullable=True)
duration_ms = Column(BigInteger, nullable=True)
compute_cost = Column(Numeric, nullable=True)
started_at = Column(DateTime, nullable=True)
completed_at = Column(DateTime, nullable=True)
created_at = Column(DateTime, default=datetime.datetime.utcnow, nullable=False, index=True)

2
skyvern/forge/sdk/db/polls.py
View File

@@ -33,7 +33,7 @@ async def await_browser_session(
if persistent_browser_session is None:
raise Exception(f"Persistent browser session not found for {session_id}")
LOG.info(
LOG.debug(
"Checking browser address",
session_id=session_id,
address=persistent_browser_session.browser_address,

80
skyvern/forge/sdk/routes/streaming/channels/vnc.py
View File

@@ -281,6 +281,43 @@ async def ask_for_clipboard(vnc_channel: VncChannel) -> None:
LOG.exception(f"{class_name} Failed to ask for clipboard via CDP", **vnc_channel.identity)
# TODO(benji): I hate this function. It's messy and gross. Once we remove v1,
# we should clean this up.
def _build_vnc_url_from_browser_address(browser_address: str) -> str | None:
"""
Build a routed VNC URL from a V2 K8s routed browser_address.
V2 K8s routed browser_address format:
wss://{domain}/{session_id}/{token}/devtools/browser/{browser_id}
Returns VNC URL in format:
wss://{domain}/vnc/{session_id}/{token}
Returns None if browser_address is not a V2 routed URL.
"""
if not browser_address:
return None
parsed = urlparse(browser_address)
# Check if this looks like a V2 routed URL (wss:// with token in path)
if parsed.scheme not in ("wss", "ws"):
return None
# Parse path: /{session_id}/{token}/devtools/browser/{browser_id}
path_parts = parsed.path.strip("/").split("/")
if len(path_parts) < 4 or path_parts[2] != "devtools":
return None
session_id = path_parts[0]
token = path_parts[1]
domain = parsed.netloc
# Build VNC URL with same domain and token
scheme = "wss" if parsed.scheme == "wss" else "ws"
return f"{scheme}://{domain}/vnc/{session_id}/{token}"
async def loop_stream_vnc(vnc_channel: VncChannel) -> None:
"""
Actually stream the VNC data between a frontend and a browser.
@@ -292,24 +329,28 @@ async def loop_stream_vnc(vnc_channel: VncChannel) -> None:
browser_session = vnc_channel.browser_session
class_name = vnc_channel.class_name
if browser_session:
if browser_session.ip_address:
if ":" in browser_session.ip_address:
ip, _ = browser_session.ip_address.split(":")
vnc_url = f"ws://{ip}:{vnc_channel.vnc_port}"
else:
vnc_url = f"ws://{browser_session.ip_address}:{vnc_channel.vnc_port}"
else:
browser_address = browser_session.browser_address
parsed_browser_address = urlparse(browser_address)
host = parsed_browser_address.hostname
vnc_url = f"ws://{host}:{vnc_channel.vnc_port}"
else:
if not browser_session:
raise Exception(f"{class_name} No browser session associated with vnc channel.")
# NOTE(jdo:streaming-local-dev)
# vnc_url = "ws://localhost:6080"
# First, check if this is a V2 K8s routed session by examining browser_address
# V2 sessions have browser_address like: wss://{domain}/{session_id}/{token}/devtools/...
# For these, we need to route VNC through the same nginx proxy
routed_vnc_url = _build_vnc_url_from_browser_address(browser_session.browser_address)
if routed_vnc_url:
vnc_url = routed_vnc_url
elif browser_session.ip_address:
# V1 ECS sessions: Direct IP connection (ip_address is a public/reachable IP)
if ":" in browser_session.ip_address:
ip, _ = browser_session.ip_address.split(":")
vnc_url = f"ws://{ip}:{vnc_channel.vnc_port}"
else:
vnc_url = f"ws://{browser_session.ip_address}:{vnc_channel.vnc_port}"
else:
# Last resort: parse browser_address hostname
browser_address = browser_session.browser_address
parsed_browser_address = urlparse(browser_address)
host = parsed_browser_address.hostname
vnc_url = f"ws://{host}:{vnc_channel.vnc_port}"
LOG.info(
f"{class_name} Connecting to vnc url.",
@@ -317,7 +358,12 @@ async def loop_stream_vnc(vnc_channel: VncChannel) -> None:
**vnc_channel.identity,
)
async with websockets.connect(vnc_url) as novnc_ws:
# For routed VNC URLs (wss://), we need to pass the x-api-key header for authentication
extra_headers: dict[str, str] = {}
if vnc_url.startswith("wss://") and vnc_channel.x_api_key:
extra_headers["x-api-key"] = vnc_channel.x_api_key
async with websockets.connect(vnc_url, additional_headers=extra_headers) as novnc_ws:
async def frontend_to_browser() -> None:
nonlocal class_name

8
skyvern/forge/sdk/routes/streaming/messages.py
View File

@@ -5,8 +5,10 @@ Provides WS endpoints for streaming messages to/from our frontend application.
import structlog
from fastapi import WebSocket
from skyvern.config import settings
from skyvern.forge.sdk.routes.routers import base_router, legacy_base_router
from skyvern.forge.sdk.routes.streaming.auth import auth
from skyvern.forge.sdk.routes.streaming.auth import _auth as local_auth
from skyvern.forge.sdk.routes.streaming.auth import auth as real_auth
from skyvern.forge.sdk.routes.streaming.channels.message import (
Loops,
MessageChannel,
@@ -60,6 +62,7 @@ async def messages(
client_id: str | None = None,
token: str | None = None,
) -> None:
auth = local_auth if settings.ENV == "local" else real_auth
organization_id = await auth(apikey=apikey, token=token, websocket=websocket)
if not organization_id:
@@ -98,7 +101,7 @@ async def messages(
)
else:
LOG.error(
"Message channel: no browser_session_id or workflow_run_id provided.",
"[WS] messages: no browser_session_id or workflow_run_id provided",
client_id=client_id,
organization_id=organization_id,
)
@@ -139,5 +142,4 @@ async def messages(
workflow_run_id=workflow_run_id,
organization_id=organization_id,
)
await message_channel.close(reason="message-stream-closed")

3
skyvern/forge/sdk/routes/streaming/verify.py
View File

@@ -45,12 +45,12 @@ async def verify_browser_session(
"""
Verify the browser session exists, and is usable.
"""
if settings.ENV == "local":
dummy_browser_session = AddressablePersistentBrowserSession(
persistent_browser_session_id=browser_session_id,
organization_id=organization_id,
browser_address="0.0.0.0:9223",
ip_address="localhost",
created_at=datetime.now(),
modified_at=datetime.now(),
)
@@ -199,6 +199,7 @@ async def verify_workflow_run(
persistent_browser_session_id=workflow_run_id,
organization_id=organization_id,
browser_address="0.0.0.0:9223",
ip_address="localhost",
created_at=datetime.now(),
modified_at=datetime.now(),
)

5
skyvern/forge/sdk/routes/streaming/vnc.py
View File

@@ -12,8 +12,10 @@ NOTE(jdo:streaming-local-dev)
import structlog
from fastapi import WebSocket
from skyvern.config import settings
from skyvern.forge.sdk.routes.routers import base_router, legacy_base_router
from skyvern.forge.sdk.routes.streaming.auth import auth
from skyvern.forge.sdk.routes.streaming.auth import _auth as local_auth
from skyvern.forge.sdk.routes.streaming.auth import auth as real_auth
from skyvern.forge.sdk.routes.streaming.channels.vnc import (
Loops,
VncChannel,
@@ -86,6 +88,7 @@ async def stream(
workflow_run_id=workflow_run_id,
)
auth = local_auth if settings.ENV == "local" else real_auth
organization_id = await auth(apikey=apikey, token=token, websocket=websocket)
if not organization_id:

6
skyvern/forge/sdk/schemas/persistent_browser_sessions.py
View File

@@ -1,4 +1,5 @@
from datetime import datetime
from decimal import Decimal
from enum import StrEnum
from pydantic import BaseModel, ConfigDict
@@ -48,6 +49,11 @@ class PersistentBrowserSession(BaseModel):
status: str | None = None
timeout_minutes: int | None = None
proxy_location: ProxyLocation | None = None
instance_type: str | None = None
vcpu_millicores: int | None = None
memory_mb: int | None = None
duration_ms: int | None = None
compute_cost: Decimal | None = None
started_at: datetime | None = None
completed_at: datetime | None = None
created_at: datetime