ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.
Python ai-agentsamdcomfyuidockerllama-cppllmlocal-ain8nnvidiaopen-webuiragself-hostedspeech-to-textstrix-halotext-to-speechworkflow-automation
export function readSharing(value) {
const config = value?.configuration
if (!config || config.schemaVersion !== 1 || !Number.isSafeInteger(config.revision) || config.revision < 0
|| typeof config.enabled !== 'boolean' || !Array.isArray(config.devices) || config.devices.length > 64
|| !['not-probed', 'starting', 'ready', 'stopped', 'error', 'unavailable'].includes(value.runtime?.status)
|| value.transport?.mode !== 'loopback-only' || !Number.isInteger(value.transport.port)
|| value.transport.port < 1024 || value.transport.port > 65535) throw new Error('Invalid sharing response')
return { configuration: config, activeRoute: value.activeRoute, transport: value.transport, runtime: value.runtime }
}
export function connectionBaseUrl(value) {
const url = new URL(value)
const local = ['127.0.0.1', 'localhost', '[::1]'].includes(url.hostname)
if (url.username || url.password || url.search || url.hash || !url.hostname
|| (url.protocol !== 'https:' && !(url.protocol === 'http:' && local))
|| !['/v1', '/v1/'].includes(url.pathname)) throw new Error('Use an HTTPS or loopback /v1 URL')
return url.href.replace(/\/$/, '')
}
export function connectionBundle(issued, baseUrl) {
const snapshot = readSharing(issued)
const credential = issued.credential
const device = snapshot.configuration.devices.find(item => item.id === credential?.id && !item.revoked)
if (!device || !/^ods_infer_[a-f0-9]{64}$/.test(credential?.key || '') || issued.model !== 'ods/shared') {
throw new Error('Invalid issued credential')
}
return {
schemaVersion: 1, kind: 'ods-inference-connection', label: device.label,
baseUrl: connectionBaseUrl(baseUrl), model: 'ods/shared', deviceId: device.id, expiresAt: device.expiresAt,
expected: { catalogId: device.catalogId, runtimeModelId: device.runtimeModelId },
credential: { apiKey: credential.key }, execution: 'client-owned',
}
}