/** * @file This file exports a class that implements the InferenceExtension interface from the @janhq/core package. * The class provides methods for initializing and stopping a model, and for making inference requests. * It also subscribes to events emitted by the @janhq/core package and handles new message requests. * @version 1.0.0 * @module inference-extension/src/index */ import { ChatCompletionRole, ContentType, EventName, MessageRequest, MessageStatus, ExtensionType, ThreadContent, ThreadMessage, events, executeOnMain, fs, Model, joinPath, InferenceExtension, log, } from "@janhq/core"; import { requestInference } from "./helpers/sse"; import { ulid } from "ulid"; import { join } from "path"; /** * A class that implements the InferenceExtension interface from the @janhq/core package. * The class provides methods for initializing and stopping a model, and for making inference requests. * It also subscribes to events emitted by the @janhq/core package and handles new message requests. */ export default class JanInferenceNitroExtension implements InferenceExtension { private static readonly _homeDir = "file://engines"; private static readonly _settingsDir = "file://settings"; private static readonly _engineMetadataFileName = "nitro.json"; /** * Checking the health for Nitro's process each 5 secs. */ private static readonly _intervalHealthCheck = 5 * 1000; private _currentModel: Model; private _engineSettings: EngineSettings = { ctx_len: 2048, ngl: 100, cpu_threads: 1, cont_batching: false, embedding: false, }; controller = new AbortController(); isCancelled = false; /** * The interval id for the health check. Used to stop the health check. */ private getNitroProcesHealthIntervalId: NodeJS.Timeout | undefined = undefined; /** * Tracking the current state of nitro process. */ private nitroProcessInfo: any = undefined; /** * Returns the type of the extension. * @returns {ExtensionType} The type of the extension. */ type(): ExtensionType { return ExtensionType.Inference; } /** * Subscribes to events emitted by the @janhq/core package. */ async onLoad() { if (!(await fs.existsSync(JanInferenceNitroExtension._homeDir))) { await fs .mkdirSync(JanInferenceNitroExtension._homeDir) .catch((err) => console.debug(err)); } if (!(await fs.existsSync(JanInferenceNitroExtension._settingsDir))) await fs.mkdirSync(JanInferenceNitroExtension._settingsDir); this.writeDefaultEngineSettings(); // Events subscription events.on(EventName.OnMessageSent, (data) => this.onMessageRequest(data)); events.on(EventName.OnModelInit, (model: Model) => this.onModelInit(model)); events.on(EventName.OnModelStop, (model: Model) => this.onModelStop(model)); events.on(EventName.OnInferenceStopped, () => this.onInferenceStopped()); // Attempt to fetch nvidia info await executeOnMain(MODULE, "updateNvidiaInfo", {}); } /** * Stops the model inference. */ onUnload(): void {} private async writeDefaultEngineSettings() { try { const engineFile = join( JanInferenceNitroExtension._homeDir, JanInferenceNitroExtension._engineMetadataFileName ); if (await fs.existsSync(engineFile)) { const engine = await fs.readFileSync(engineFile, "utf-8"); this._engineSettings = typeof engine === "object" ? engine : JSON.parse(engine); } else { await fs.writeFileSync( engineFile, JSON.stringify(this._engineSettings, null, 2) ); } } catch (err) { console.error(err); } } private async onModelInit(model: Model) { if (model.engine !== "nitro") return; const modelFullPath = await joinPath(["models", model.id]); const nitroInitResult = await executeOnMain(MODULE, "initModel", { modelFullPath: modelFullPath, model: model, }); if (nitroInitResult.error === null) { events.emit(EventName.OnModelFail, model); return; } this._currentModel = model; events.emit(EventName.OnModelReady, model); this.getNitroProcesHealthIntervalId = setInterval( () => this.periodicallyGetNitroHealth(), JanInferenceNitroExtension._intervalHealthCheck ); } private async onModelStop(model: Model) { if (model.engine !== "nitro") return; await executeOnMain(MODULE, "stopModel"); events.emit(EventName.OnModelStopped, {}); // stop the periocally health check if (this.getNitroProcesHealthIntervalId) { console.debug("Stop calling Nitro process health check"); clearInterval(this.getNitroProcesHealthIntervalId); this.getNitroProcesHealthIntervalId = undefined; } } /** * Periodically check for nitro process's health. */ private async periodicallyGetNitroHealth(): Promise { const health = await executeOnMain(MODULE, "getCurrentNitroProcessInfo"); const isRunning = this.nitroProcessInfo?.isRunning ?? false; if (isRunning && health.isRunning === false) { console.debug("Nitro process is stopped"); events.emit(EventName.OnModelStopped, {}); } this.nitroProcessInfo = health; } private async onInferenceStopped() { this.isCancelled = true; this.controller?.abort(); } /** * Makes a single response inference request. * @param {MessageRequest} data - The data for the inference request. * @returns {Promise} A promise that resolves with the inference response. */ async inference(data: MessageRequest): Promise { const timestamp = Date.now(); const message: ThreadMessage = { thread_id: data.threadId, created: timestamp, updated: timestamp, status: MessageStatus.Ready, id: "", role: ChatCompletionRole.Assistant, object: "thread.message", content: [], }; return new Promise(async (resolve, reject) => { requestInference(data.messages ?? [], this._currentModel).subscribe({ next: (_content) => {}, complete: async () => { resolve(message); }, error: async (err) => { reject(err); }, }); }); } /** * Handles a new message request by making an inference request and emitting events. * Function registered in event manager, should be static to avoid binding issues. * Pass instance as a reference. * @param {MessageRequest} data - The data for the new message request. */ private async onMessageRequest(data: MessageRequest) { if (data.model.engine !== "nitro") return; const timestamp = Date.now(); const message: ThreadMessage = { id: ulid(), thread_id: data.threadId, assistant_id: data.assistantId, role: ChatCompletionRole.Assistant, content: [], status: MessageStatus.Pending, created: timestamp, updated: timestamp, object: "thread.message", }; events.emit(EventName.OnMessageResponse, message); this.isCancelled = false; this.controller = new AbortController(); requestInference( data.messages ?? [], { ...this._currentModel, ...data.model }, this.controller ).subscribe({ next: (content) => { const messageContent: ThreadContent = { type: ContentType.Text, text: { value: content.trim(), annotations: [], }, }; message.content = [messageContent]; events.emit(EventName.OnMessageUpdate, message); }, complete: async () => { message.status = message.content.length ? MessageStatus.Ready : MessageStatus.Error; events.emit(EventName.OnMessageUpdate, message); }, error: async (err) => { if (this.isCancelled || message.content.length) { message.status = MessageStatus.Stopped; events.emit(EventName.OnMessageUpdate, message); return; } message.status = MessageStatus.Error; events.emit(EventName.OnMessageUpdate, message); log(`[APP]::Error: ${err.message}`); }, }); } }