Jon/browser stream component (#2808)

This commit is contained in:
Shuchang Zheng
2025-06-28 02:46:20 +09:00
committed by GitHub
parent 11ca5b7250
commit db40abde6f
3 changed files with 148 additions and 73 deletions

View File

@@ -0,0 +1,352 @@
import { Status } from "@/api/types";
import { useEffect, useState, useRef, useCallback } from "react";
import { HandIcon, PlayIcon } from "@radix-ui/react-icons";
import { Button } from "@/components/ui/button";
import { Skeleton } from "@/components/ui/skeleton";
import { statusIsNotFinalized } from "@/routes/tasks/types";
import { useCredentialGetter } from "@/hooks/useCredentialGetter";
import { envCredential } from "@/util/env";
import { toast } from "@/components/ui/use-toast";
import RFB from "@novnc/novnc/lib/rfb.js";
import { environment } from "@/util/env";
import { cn } from "@/util/utils";
import { useClientIdStore } from "@/store/useClientIdStore";
import type {
TaskApiResponse,
WorkflowRunStatusApiResponse,
} from "@/api/types";
import "./browser-stream.css";
const wssBaseUrl = import.meta.env.VITE_WSS_BASE_URL;
interface CommandTakeControl {
kind: "take-control";
}
interface CommandCedeControl {
kind: "cede-control";
}
type Command = CommandTakeControl | CommandCedeControl;
type Props = {
task?: {
run: TaskApiResponse;
};
workflow?: {
run: WorkflowRunStatusApiResponse;
};
// --
onClose?: () => void;
};
function BrowserStream({
task = undefined,
workflow = undefined,
// --
onClose,
}: Props) {
let showStream: boolean = false;
let runId: string;
let entity: "task" | "workflow";
if (task) {
runId = task.run.task_id;
showStream = statusIsNotFinalized(task.run);
entity = "task";
} else if (workflow) {
runId = workflow.run.workflow_run_id;
showStream = statusIsNotFinalized(workflow.run);
entity = "workflow";
} else {
throw new Error("No task or workflow provided");
}
const [commandSocket, setCommandSocket] = useState<WebSocket | null>(null);
const [userIsControlling, setUserIsControlling] = useState<boolean>(false);
const [vncDisconnectedTrigger, setVncDisconnectedTrigger] = useState(0);
const prevVncConnectedRef = useRef<boolean>(false);
const [isVncConnected, setIsVncConnected] = useState<boolean>(false);
const [commandDisconnectedTrigger, setCommandDisconnectedTrigger] =
useState(0);
const prevCommandConnectedRef = useRef<boolean>(false);
const [isCommandConnected, setIsCommandConnected] = useState<boolean>(false);
// goes up a level
// const queryClient = useQueryClient();
const [canvasContainer, setCanvasContainer] = useState<HTMLDivElement | null>(
null,
);
const setCanvasContainerRef = useCallback((node: HTMLDivElement | null) => {
setCanvasContainer(node);
}, []);
const rfbRef = useRef<RFB | null>(null);
const clientId = useClientIdStore((state) => state.clientId);
const credentialGetter = useCredentialGetter();
const getWebSocketParams = useCallback(async () => {
const clientIdQueryParam = `client_id=${clientId}`;
let credentialQueryParam = "";
if (environment === "local") {
credentialQueryParam = `apikey=${envCredential}`;
} else {
if (credentialGetter) {
const token = await credentialGetter();
credentialQueryParam = `token=Bearer ${token}`;
} else {
credentialQueryParam = `apikey=${envCredential}`;
}
}
const params = [credentialQueryParam, clientIdQueryParam].join("&");
return `${params}`;
}, [clientId, credentialGetter]);
// effect for vnc disconnects only
useEffect(() => {
if (prevVncConnectedRef.current && !isVncConnected) {
setVncDisconnectedTrigger((x) => x + 1);
onClose?.();
}
prevVncConnectedRef.current = isVncConnected;
}, [isVncConnected, onClose]);
// effect for command disconnects only
useEffect(() => {
if (prevCommandConnectedRef.current && !isCommandConnected) {
setCommandDisconnectedTrigger((x) => x + 1);
onClose?.();
}
prevCommandConnectedRef.current = isCommandConnected;
}, [isCommandConnected, onClose]);
// vnc socket
useEffect(
() => {
if (!showStream || !canvasContainer || !runId) {
if (rfbRef.current) {
rfbRef.current.disconnect();
rfbRef.current = null;
setIsVncConnected(false);
}
return;
}
async function setupVnc() {
if (rfbRef.current && isVncConnected) {
return;
}
const wsParams = await getWebSocketParams();
const vncUrl =
entity === "task"
? `${wssBaseUrl}/stream/vnc/task/${runId}?${wsParams}`
: entity === "workflow"
? `${wssBaseUrl}/stream/vnc/workflow_run/${runId}?${wsParams}`
: null;
if (!vncUrl) {
throw new Error("No vnc url");
}
if (rfbRef.current) {
rfbRef.current.disconnect();
}
const canvas = canvasContainer;
if (!canvas) {
throw new Error("Canvas element not found");
}
const rfb = new RFB(canvas, vncUrl);
rfb.scaleViewport = true;
rfbRef.current = rfb;
rfb.addEventListener("connect", () => {
setIsVncConnected(true);
});
rfb.addEventListener("disconnect", async (/* e: RfbEvent */) => {
setIsVncConnected(false);
});
}
setupVnc();
return () => {
if (rfbRef.current) {
rfbRef.current.disconnect();
rfbRef.current = null;
}
setIsVncConnected(false);
};
},
// cannot include isVncConnected in deps as it will cause infinite loop
// eslint-disable-next-line react-hooks/exhaustive-deps
[
canvasContainer,
showStream,
vncDisconnectedTrigger, // will re-run on disconnects
runId,
entity,
],
);
// command socket
useEffect(() => {
if (!showStream || !canvasContainer || !runId) {
return;
}
let ws: WebSocket | null = null;
const connect = async () => {
const wsParams = await getWebSocketParams();
const commandUrl =
entity === "task"
? `${wssBaseUrl}/stream/commands/task/${runId}?${wsParams}`
: entity === "workflow"
? `${wssBaseUrl}/stream/commands/workflow_run/${runId}?${wsParams}`
: null;
if (!commandUrl) {
throw new Error("No command url");
}
ws = new WebSocket(commandUrl);
ws.onopen = () => {
setIsCommandConnected(true);
setCommandSocket(ws);
};
ws.onclose = () => {
setIsCommandConnected(false);
setCommandSocket(null);
};
};
connect();
return () => {
try {
ws && ws.close();
} catch (e) {
// pass
}
};
}, [
canvasContainer,
commandDisconnectedTrigger,
entity,
getWebSocketParams,
runId,
showStream,
]);
// effect to send a command when the user is controlling, vs not controlling
useEffect(() => {
if (!isCommandConnected) {
return;
}
const sendCommand = (command: Command) => {
if (!commandSocket) {
console.warn("Cannot send command, as command socket is closed.");
console.warn(command);
return;
}
commandSocket.send(JSON.stringify(command));
};
if (userIsControlling) {
sendCommand({ kind: "take-control" });
} else {
sendCommand({ kind: "cede-control" });
}
// eslint-disable-next-line react-hooks/exhaustive-deps
}, [userIsControlling, isCommandConnected]);
// Effect to show toast when task or workflow reaches a final state based on hook updates
useEffect(() => {
const run = task ? task.run : workflow ? workflow.run : null;
if (!run) {
return;
}
const name = task ? "task" : workflow ? "workflow" : null;
if (!name) {
return;
}
if (run.status === Status.Failed || run.status === Status.Terminated) {
// Only show toast if VNC is not connected or was never connected,
// to avoid double toasting if disconnect handler also triggers similar logic.
// However, the disconnect handler now primarily invalidates queries.
toast({
title: "Run Ended",
description: `The ${name} run has ${run.status}.`,
variant: "destructive",
});
} else if (run.status === Status.Completed) {
toast({
title: "Run Completed",
description: `The ${name} run has been completed.`,
variant: "success",
});
}
}, [task, workflow]);
return (
<div
className={cn("browser-stream", {
"user-is-controlling": userIsControlling,
})}
ref={setCanvasContainerRef}
>
{isVncConnected && (
<div className="overlay-container">
<div className="overlay">
<Button
// className="take-control"
className={cn("take-control", { hide: userIsControlling })}
type="button"
onClick={() => setUserIsControlling(true)}
>
<HandIcon className="mr-2 h-4 w-4" />
take control
</Button>
<div className="absolute bottom-[-1rem] right-[1rem]">
<Button
className={cn("relinquish-control", {
hide: !userIsControlling,
})}
type="button"
onClick={() => setUserIsControlling(false)}
>
<PlayIcon className="mr-2 h-4 w-4" />
run agent
</Button>
</div>
</div>
</div>
)}
{!isVncConnected && (
<div className="absolute left-0 top-0 flex h-full w-full items-center justify-center bg-black">
<Skeleton className="aspect-[16/9] h-auto max-h-full w-full max-w-full rounded-lg object-cover" />
</div>
)}
</div>
);
}
export { BrowserStream };

View File

@@ -0,0 +1,96 @@
.browser-stream {
position: relative;
width: 100%;
height: 100%;
min-height: 0;
padding: 0.5rem;
overflow: visible;
transition: padding 0.2s ease-in-out;
}
.browser-stream.user-is-controlling {
padding: 0rem;
}
.browser-stream .overlay-container {
position: absolute;
top: 0;
left: 0;
width: 100%;
height: 100%;
display: flex;
align-items: center;
justify-content: center;
}
.browser-stream .overlay {
position: relative;
height: auto;
width: 100%;
max-height: 100%;
aspect-ratio: 16 / 9;
object-fit: cover;
display: flex;
align-items: center;
justify-content: center;
}
.browser-stream.user-is-controlling .overlay {
pointer-events: none;
}
.browser-stream.user-is-controlling .overlay-container {
pointer-events: none;
}
.browser-stream .take-control {
transform: translateY(0);
transition:
transform 0.2s ease-in-out,
opacity 0.2s ease-in-out;
opacity: 0.3;
}
.browser-stream .take-control:not(.hide):hover {
opacity: 1;
}
.browser-stream .take-control.hide {
transform: translateY(100%);
opacity: 0;
pointer-events: none;
}
.browser-stream .relinquish-control {
transform: translateY(0);
transition:
transform 0.2s ease-in-out,
opacity 0.2s ease-in-out;
opacity: 0.3;
pointer-events: all;
}
.browser-stream .relinquish-control:not(.hide):hover {
opacity: 1;
}
.browser-stream .relinquish-control.hide {
transform: translateY(100%);
opacity: 0;
pointer-events: none;
}
.browser-stream > div > canvas {
opacity: 0;
animation: skyvern-anim-fadeIn 1s ease-in forwards;
}
@keyframes skyvern-anim-fadeIn {
from {
opacity: 0;
}
to {
opacity: 1;
}
}