Open-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent.
import { invoke } from "@tauri-apps/api/core";
import { useEffect, useState } from "react";
import { isHttpEndpointUrl } from "./http-endpoint-url";
import { Mark } from "./Mark";
export type Harness = {
id: string;
name: string;
summary: string;
image: string | null;
health_path: string | null;
// Mirrors `Credential` in shared/harnesses.ts, which this cannot import: the desktop app is
// outside the workspace and reads the catalogue over the Tauri boundary as JSON. Adding a
// variant there means adding it here, and the tile below is what reads it.
credential: "any-provider" | "anthropic" | "google" | "their-endpoint";
maintainer: "first-party" | "partnership" | "community";
mark: string | null;
port: number | null;
};
export type HarnessChoice = {
id: string;
agentUrl?: string;
};
/** What OpenBot sets up unless somebody says otherwise. David's call. */
export const DEFAULT_HARNESS = "langgraph";
/**
* Which Bot, answered for them.
*
* ONE CHOICE IS MADE FOR THE PERSON, and that is the point of this screen rather than a limitation
* of it. The twelve rows are agent frameworks, and to anybody who is not a developer the difference
* between them is nil: they all take any model and they all answer the same questions. Asking a
* non-technical person to pick one is asking them to make a decision they cannot inform, at the
* start, which is where people leave.
*
* So the default is stated in one line and the list moves behind a disclosure. A developer who
* wants CrewAI opens it and picks CrewAI; everybody else presses Continue and never learns the word
* "harness". The address-your-own row lives in there too, because pasting a URL is the most
* developer thing on this screen.
*/
export function HarnessPicker({
chosen,
onChoose,
onContinue,
onBack,
}: {
chosen: HarnessChoice | null;
onChoose: (choice: HarnessChoice) => void;
onContinue: () => void;
onBack: () => void;
}) {
const [rows, setRows] = useState<Harness[]>([]);
const [failure, setFailure] = useState("");
// Open when the person has already chosen something other than the default, so coming back does
// not hide the choice they made.
const [open, setOpen] = useState(
chosen !== null && chosen.id !== DEFAULT_HARNESS,
);
useEffect(() => {
invoke<Harness[]>("harnesses")
.then(setRows)
.catch((error) => setFailure(String(error)));
}, []);
const chosenId = chosen?.id ?? DEFAULT_HARNESS;
const picked = rows.find((row) => row.id === chosenId);
const byoAgentUrl = chosenId === "byo-url" ? (chosen?.agentUrl ?? "") : "";
const byoReady =
(byoAgentUrl.trim().startsWith("http://") ||
byoAgentUrl.trim().startsWith("https://")) &&
isHttpEndpointUrl(byoAgentUrl);
if (failure) {
return (
<div className="sheet">
<div className="blocker" role="alert">
<h2>The list of Bots could not be read</h2>
<p>{failure}</p>
</div>
</div>
);
}
return (
<div className="sheet">
<p className="steps-of">Step 1 of 4</p>
<h1>Your first Bot</h1>
{/*
Written to the person who has to act, not about the situation.
An earlier version said the default "makes no difference unless you write code", which
describes a state and leaves a non-technical reader wondering what they were told. This
gives them the one thing to do — nothing — and puts the conditional where the only person it
applies to will read it.
*/}
<p className="lede">
OpenBot sets this up for you. If you write code, you can choose the
agent framework below.
</p>
<details open={open} onToggle={(e) => setOpen(e.currentTarget.open)}>
<summary>
{picked && picked.id !== DEFAULT_HARNESS
? `Using ${picked.name}`
: "Choose the agent framework"}
</summary>
<p className="footnote" style={{ margin: "0.6rem 0 0" }}>
Any of these works with any AI provider. Only the last one asks you
for an address.
</p>
<fieldset className="picker">
<legend className="sr-only">Bot</legend>
{rows.map((row) => (
<label
key={row.id}
className={`tile${chosenId === row.id ? " chosen" : ""}`}
>
<input
type="radio"
name="harness"
className="tile-input"
value={row.id}
checked={chosenId === row.id}
onChange={() =>
onChoose(
row.id === "byo-url"
? { id: row.id, agentUrl: byoAgentUrl }
: { id: row.id },
)
}
/>
<Mark id={row.mark} name={row.name} />
{/* The name is on every row, mark or no mark, so a person who does not recognise a
logo can still read it. */}
<span className="tile-name">{row.name}</span>
<span className="tile-summary">{row.summary}</span>
{row.credential === "anthropic" && (
<span className="tile-note">No API key needed</span>
)}
{row.credential === "google" && (
<span className="tile-note">Needs a Gemini key</span>
)}
{row.credential === "their-endpoint" && (
<span className="tile-note">Uses your running agent</span>
)}
</label>
))}
</fieldset>
{chosenId === "byo-url" && (
<div className="field" style={{ marginTop: "0.75rem" }}>
<label htmlFor="agent-url">AG-UI endpoint</label>
<input
id="agent-url"
value={byoAgentUrl}
onChange={(event) =>
onChoose({
id: "byo-url",
agentUrl: event.target.value,
})
}
placeholder="https://your-agent.example/ag-ui"
spellCheck={false}
/>
</div>
)}
</details>
<div className="row">
<button type="button" className="quiet" onClick={onBack}>
Back
</button>
<button
type="button"
disabled={chosenId === "byo-url" && !byoReady}
onClick={onContinue}
>
Continue
</button>
</div>
</div>
);
}