The all-in-one local AI studio for your desktop: chat, image and video generation and a coding agent in one free, open source app. Windows and Linux. No Docker, no terminal, no cloud required.
TypeScript agentai-chatai-studioall-in-onecoding-agentdesktop-appfluximage-generationlinuxllmlocal-ailocal-llmopen-sourceprivacyself-hostedstable-diffusionuncensoreduncensored-aivideo-generationwindows
/**
* Did the backend ANSWER, or did the call simply not get through.
*
* A14 second review: the boot resume gets one shot per app session, and the
* flag that spends it was raised before the call. That is right for a machine
* that has no sidecar and will never have one, and wrong for the moment this
* actually runs: right after launch, with the antivirus still scanning the
* fresh install and the Tauri command layer coming up behind the window. A
* first call that loses that race used to spend the only shot, and the engine
* the user left running yesterday stayed dead until he re-picked the model by
* hand. That is GH #118 again, one door further along.
*
* So the shot is spent on an ANSWER, and a refusal counts as one: a backend
* that says "I do not have that command" has answered, and waiting will not
* change its mind. A timeout or a dead transport has said nothing at all.
*/
/** Rejections that mean "this build does not have that command". Lower-cased
* fragments, matched against the message Tauri hands across as a string. */
const NO_SUCH_COMMAND = [
// Tauri's own wording for a command that is not in the invoke handler, and
// the shapes the HTTP bridge and the web build produce for the same thing.
'not found',
'unknown command',
// A14 third review: the one refusal this module was written for was the one
// it did not recognise. `backendCall` outside Tauri routes through the HTTP
// bridge's endpoint table and throws `Unknown backend command: <name>` for
// anything not in it, which matches none of the fragments above, so the web
// and remote-bridge builds re-asked the same dead question on every single
// model refresh. The literal lives in src/api/backend.ts and is pinned by
// engine-command-availability.test.ts.
'unknown backend command',
'no such command',
'invalid command',
'command not allowed',
'not allowed by scope',
'not implemented',
// No Tauri at all: `backendCall` outside the desktop build.
'not running in tauri',
'is not a function',
'not available in this build',
]
/** The message behind an unknown thrown value. */
export function commandFailureText(err: unknown): string {
if (err instanceof Error) return err.message
if (typeof err === 'string') return err
return String(err ?? '')
}
/**
* True when the failure is the backend saying it does not have this command.
*
* Deliberately the narrow half. Everything unrecognised counts as "no answer"
* and is retried, because the cost of retrying is one extra command on the
* next model refresh and the cost of not retrying is a dead engine for the
* rest of the session.
*/
export function commandIsUnavailable(err: unknown): boolean {
const msg = commandFailureText(err).toLowerCase()
if (!msg) return false
return NO_SUCH_COMMAND.some((m) => msg.includes(m))
}