node-llama-cpp 3.0.0-beta.4 → 3.0.0-beta.41

package/dist/bindings/Llama.d.ts

@@ -0,0 +1,47 @@
+import { EventRelay } from "lifecycle-utils";
+import { LlamaModel, LlamaModelOptions } from "../evaluator/LlamaModel/LlamaModel.js";
+import { GbnfJsonSchema } from "../utils/gbnfJson/types.js";
+import { LlamaJsonSchemaGrammar } from "../evaluator/LlamaJsonSchemaGrammar.js";
+import { LlamaGrammar, LlamaGrammarOptions } from "../evaluator/LlamaGrammar.js";
+import { LlamaGpuType, LlamaLogLevel } from "./types.js";
+export declare class Llama {
+    readonly onDispose: EventRelay<void>;
+    private constructor();
+    dispose(): Promise<void>;
+    /** @hidden */
+    [Symbol.asyncDispose](): Promise<void>;
+    get disposed(): boolean;
+    get gpu(): LlamaGpuType;
+    get supportsGpuOffloading(): boolean;
+    get supportsMmap(): boolean;
+    get supportsMlock(): boolean;
+    get logLevel(): LlamaLogLevel;
+    set logLevel(value: LlamaLogLevel);
+    get logger(): (level: LlamaLogLevel, message: string) => void;
+    set logger(value: (level: LlamaLogLevel, message: string) => void);
+    get buildType(): "localBuild" | "prebuilt";
+    get cmakeOptions(): Readonly<Record<string, string>>;
+    get llamaCppRelease(): {
+        readonly repo: string;
+        readonly release: string;
+    };
+    get systemInfo(): string;
+    /**
+     * VRAM padding used for memory size calculations, as these calculations are not always accurate.
+     * This is set by default to ensure stability, but can be configured when you call `getLlama`.
+     *
+     * See `vramPadding` on `getLlama` for more information.
+     */
+    get vramPaddingSize(): number;
+    getVramState(): Promise<{
+        total: number;
+        used: number;
+        free: number;
+    }>;
+    getGpuDeviceNames(): Promise<string[]>;
+    loadModel(options: LlamaModelOptions): Promise<LlamaModel>;
+    createGrammarForJsonSchema<const T extends Readonly<GbnfJsonSchema>>(schema: T): Promise<LlamaJsonSchemaGrammar<T>>;
+    getGrammarFor(type: Parameters<typeof LlamaGrammar.getFor>[1]): Promise<LlamaGrammar>;
+    createGrammar(options: LlamaGrammarOptions): Promise<LlamaGrammar>;
+    static defaultConsoleLogger(level: LlamaLogLevel, message: string): void;
+}

package/dist/bindings/Llama.js

@@ -0,0 +1,356 @@
+import chalk from "chalk";
+import { DisposedError, EventRelay, withLock } from "lifecycle-utils";
+import { getConsoleLogPrefix } from "../utils/getConsoleLogPrefix.js";
+import { LlamaModel } from "../evaluator/LlamaModel/LlamaModel.js";
+import { DisposeGuard } from "../utils/DisposeGuard.js";
+import { LlamaJsonSchemaGrammar } from "../evaluator/LlamaJsonSchemaGrammar.js";
+import { LlamaGrammar } from "../evaluator/LlamaGrammar.js";
+import { LlamaLocks, LlamaLogLevel } from "./types.js";
+import { MemoryOrchestrator } from "./utils/MemoryOrchestrator.js";
+const LlamaLogLevelToAddonLogLevel = new Map([
+    [LlamaLogLevel.disabled, 0],
+    [LlamaLogLevel.fatal, 1],
+    [LlamaLogLevel.error, 2],
+    [LlamaLogLevel.warn, 3],
+    [LlamaLogLevel.info, 4],
+    [LlamaLogLevel.debug, 5]
+]);
+const addonLogLevelToLlamaLogLevel = new Map([...LlamaLogLevelToAddonLogLevel.entries()].map(([key, value]) => [value, key]));
+const defaultLogLevel = 5;
+export class Llama {
+    /** @internal */ _bindings;
+    /** @internal */ _backendDisposeGuard = new DisposeGuard();
+    /** @internal */ _memoryLock = {};
+    /** @internal */ _consts;
+    /** @internal */ _vramOrchestrator;
+    /** @internal */ _vramPadding;
+    /** @internal */ _debug;
+    /** @internal */ _gpu;
+    /** @internal */ _buildType;
+    /** @internal */ _cmakeOptions;
+    /** @internal */ _supportsGpuOffloading;
+    /** @internal */ _supportsMmap;
+    /** @internal */ _supportsMlock;
+    /** @internal */ _llamaCppRelease;
+    /** @internal */ _logger;
+    /** @internal */ _logLevel;
+    /** @internal */ _pendingLog = null;
+    /** @internal */ _pendingLogLevel = null;
+    /** @internal */ _logDispatchQueuedMicrotasks = 0;
+    /** @internal */ _previousLog = null;
+    /** @internal */ _previousLogLevel = null;
+    /** @internal */ _nextLogNeedNewLine = false;
+    /** @internal */ _disposed = false;
+    onDispose = new EventRelay();
+    constructor({ bindings, logLevel, logger, buildType, cmakeOptions, llamaCppRelease, debug, gpu, vramOrchestrator, vramPadding }) {
+        this._bindings = bindings;
+        this._gpu = gpu;
+        this._supportsGpuOffloading = bindings.getSupportsGpuOffloading();
+        this._supportsMmap = bindings.getSupportsMmap();
+        this._supportsMlock = bindings.getSupportsMlock();
+        this._consts = bindings.getConsts();
+        this._debug = debug;
+        this._vramOrchestrator = vramOrchestrator;
+        this._vramPadding = vramPadding;
+        this._logLevel = this._debug
+            ? LlamaLogLevel.debug
+            : (logLevel ?? LlamaLogLevel.debug);
+        this._logger = logger;
+        this._buildType = buildType;
+        this._cmakeOptions = Object.freeze({ ...cmakeOptions });
+        this._llamaCppRelease = Object.freeze({
+            repo: llamaCppRelease.repo,
+            release: llamaCppRelease.release
+        });
+        this._dispatchPendingLogMicrotask = this._dispatchPendingLogMicrotask.bind(this);
+        this._onAddonLog = this._onAddonLog.bind(this);
+        if (!this._debug) {
+            this._bindings.setLogger(this._onAddonLog);
+            this._bindings.setLoggerLogLevel(LlamaLogLevelToAddonLogLevel.get(this._logLevel) ?? defaultLogLevel);
+        }
+        this._onExit = this._onExit.bind(this);
+        process.on("exit", this._onExit);
+    }
+    async dispose() {
+        if (this._disposed)
+            return;
+        this._disposed = true;
+        this.onDispose.dispatchEvent();
+        await this._backendDisposeGuard.acquireDisposeLock();
+        await this._bindings.dispose();
+    }
+    /** @hidden */
+    async [Symbol.asyncDispose]() {
+        await this.dispose();
+    }
+    get disposed() {
+        return this._disposed;
+    }
+    get gpu() {
+        return this._gpu;
+    }
+    get supportsGpuOffloading() {
+        return this._supportsGpuOffloading;
+    }
+    get supportsMmap() {
+        return this._supportsMmap;
+    }
+    get supportsMlock() {
+        return this._supportsMlock;
+    }
+    get logLevel() {
+        return this._logLevel;
+    }
+    set logLevel(value) {
+        this._ensureNotDisposed();
+        if (value === this._logLevel || this._debug)
+            return;
+        this._bindings.setLoggerLogLevel(LlamaLogLevelToAddonLogLevel.get(value) ?? defaultLogLevel);
+        this._logLevel = value;
+    }
+    get logger() {
+        return this._logger;
+    }
+    set logger(value) {
+        this._logger = value;
+        if (value !== Llama.defaultConsoleLogger)
+            this._nextLogNeedNewLine = false;
+    }
+    get buildType() {
+        return this._buildType;
+    }
+    get cmakeOptions() {
+        return this._cmakeOptions;
+    }
+    get llamaCppRelease() {
+        return this._llamaCppRelease;
+    }
+    get systemInfo() {
+        this._ensureNotDisposed();
+        return this._bindings.systemInfo();
+    }
+    /**
+     * VRAM padding used for memory size calculations, as these calculations are not always accurate.
+     * This is set by default to ensure stability, but can be configured when you call `getLlama`.
+     *
+     * See `vramPadding` on `getLlama` for more information.
+     */
+    get vramPaddingSize() {
+        return this._vramPadding.size;
+    }
+    async getVramState() {
+        this._ensureNotDisposed();
+        const { total, used } = this._bindings.getGpuVramInfo();
+        return {
+            total,
+            used,
+            free: Math.max(0, total - used)
+        };
+    }
+    async getGpuDeviceNames() {
+        this._ensureNotDisposed();
+        const { deviceNames } = this._bindings.getGpuDeviceInfo();
+        return deviceNames;
+    }
+    async loadModel(options) {
+        this._ensureNotDisposed();
+        return await withLock(this._memoryLock, LlamaLocks.loadToMemory, options.loadSignal, async () => {
+            this._ensureNotDisposed();
+            const preventDisposalHandle = this._backendDisposeGuard.createPreventDisposalHandle();
+            try {
+                return await LlamaModel._create(options, { _llama: this });
+            }
+            finally {
+                preventDisposalHandle.dispose();
+            }
+        });
+    }
+    async createGrammarForJsonSchema(schema) {
+        return new LlamaJsonSchemaGrammar(this, schema);
+    }
+    async getGrammarFor(type) {
+        return await LlamaGrammar.getFor(this, type);
+    }
+    async createGrammar(options) {
+        return new LlamaGrammar(this, options);
+    }
+    /** @internal */
+    async _init() {
+        await this._bindings.init();
+    }
+    /**
+     * Log messages related to the Llama instance
+     * @internal
+     */
+    _log(level, message) {
+        this._onAddonLog(LlamaLogLevelToAddonLogLevel.get(level) ?? defaultLogLevel, message + "\n");
+    }
+    /** @internal */
+    _onAddonLog(level, message) {
+        const llamaLogLevel = addonLogLevelToLlamaLogLevel.get(level) ?? LlamaLogLevel.fatal;
+        if (this._pendingLog != null && this._pendingLogLevel != null && this._pendingLogLevel != llamaLogLevel) {
+            this._callLogger(this._pendingLogLevel, this._pendingLog);
+            this._pendingLog = null;
+        }
+        const sourceMessage = (this._pendingLog ?? "") + message;
+        const lastNewLineIndex = sourceMessage.lastIndexOf("\n");
+        const currentLog = lastNewLineIndex < 0
+            ? sourceMessage
+            : sourceMessage.slice(0, lastNewLineIndex);
+        const nextLog = lastNewLineIndex < 0
+            ? ""
+            : sourceMessage.slice(lastNewLineIndex + 1);
+        if (currentLog !== "")
+            this._callLogger(llamaLogLevel, currentLog);
+        if (nextLog !== "") {
+            this._pendingLog = nextLog;
+            this._pendingLogLevel = llamaLogLevel;
+            queueMicrotask(this._dispatchPendingLogMicrotask);
+            this._logDispatchQueuedMicrotasks++;
+        }
+        else
+            this._pendingLog = null;
+    }
+    /** @internal */
+    _dispatchPendingLogMicrotask() {
+        this._logDispatchQueuedMicrotasks--;
+        if (this._logDispatchQueuedMicrotasks !== 0)
+            return;
+        if (this._pendingLog != null && this._pendingLogLevel != null) {
+            this._callLogger(this._pendingLogLevel, this._pendingLog);
+            this._pendingLog = null;
+        }
+    }
+    /** @internal */
+    _callLogger(level, message) {
+        // llama.cpp uses dots to indicate progress, so we don't want to print them as different lines,
+        // and instead, append to the same log line
+        if (logMessageIsOnlyDots(message) && this._logger === Llama.defaultConsoleLogger) {
+            if (logMessageIsOnlyDots(this._previousLog) && level === this._previousLogLevel) {
+                process.stdout.write(message);
+            }
+            else {
+                this._nextLogNeedNewLine = true;
+                process.stdout.write(prefixAndColorMessage(message, getColorForLogLevel(level)));
+            }
+        }
+        else {
+            if (this._nextLogNeedNewLine) {
+                process.stdout.write("\n");
+                this._nextLogNeedNewLine = false;
+            }
+            try {
+                this._logger(level, message);
+            }
+            catch (err) {
+                // the native addon code calls this function, so there's no use to throw an error here
+            }
+        }
+        this._previousLog = message;
+        this._previousLogLevel = level;
+    }
+    /** @internal */
+    _onExit() {
+        if (this._pendingLog != null && this._pendingLogLevel != null) {
+            this._callLogger(this._pendingLogLevel, this._pendingLog);
+            this._pendingLog = null;
+        }
+    }
+    /** @internal */
+    _ensureNotDisposed() {
+        if (this._disposed)
+            throw new DisposedError();
+    }
+    /** @internal */
+    static async _create({ bindings, buildType, buildMetadata, logLevel, logger, vramPadding, skipLlamaInit = false, debug }) {
+        const gpu = bindings.getGpuType() ?? false;
+        const vramOrchestrator = new MemoryOrchestrator(() => {
+            const { total, used } = bindings.getGpuVramInfo();
+            return {
+                total,
+                free: Math.max(0, total - used)
+            };
+        });
+        let resolvedVramPadding;
+        if (gpu === false || vramPadding === 0)
+            resolvedVramPadding = vramOrchestrator.reserveMemory(0);
+        else if (vramPadding instanceof Function)
+            resolvedVramPadding = vramOrchestrator.reserveMemory(vramPadding((await vramOrchestrator.getMemoryState()).total));
+        else
+            resolvedVramPadding = vramOrchestrator.reserveMemory(vramPadding);
+        const llama = new Llama({
+            bindings,
+            buildType,
+            cmakeOptions: buildMetadata.buildOptions.customCmakeOptions,
+            llamaCppRelease: {
+                repo: buildMetadata.buildOptions.llamaCpp.repo,
+                release: buildMetadata.buildOptions.llamaCpp.release
+            },
+            logLevel,
+            logger,
+            debug,
+            gpu,
+            vramOrchestrator,
+            vramPadding: resolvedVramPadding
+        });
+        if (!skipLlamaInit)
+            await llama._init();
+        return llama;
+    }
+    static defaultConsoleLogger(level, message) {
+        switch (level) {
+            case LlamaLogLevel.disabled:
+                break;
+            case LlamaLogLevel.fatal:
+                // we don't use console.error here because it prints the stack trace
+                console.warn(prefixAndColorMessage(message, getColorForLogLevel(level)));
+                break;
+            case LlamaLogLevel.error:
+                // we don't use console.error here because it prints the stack trace
+                console.warn(prefixAndColorMessage(message, getColorForLogLevel(level)));
+                break;
+            case LlamaLogLevel.warn:
+                console.warn(prefixAndColorMessage(message, getColorForLogLevel(level)));
+                break;
+            case LlamaLogLevel.info:
+                console.info(prefixAndColorMessage(message, getColorForLogLevel(level)));
+                break;
+            case LlamaLogLevel.debug:
+                console.debug(prefixAndColorMessage(message, getColorForLogLevel(level)));
+                break;
+            default:
+                void (level);
+                console.warn(getConsoleLogPrefix() + getColorForLogLevel(LlamaLogLevel.warn)(`Unknown log level: ${level}`));
+                console.log(prefixAndColorMessage(message, getColorForLogLevel(level)));
+        }
+    }
+}
+function getColorForLogLevel(level) {
+    switch (level) {
+        case LlamaLogLevel.disabled: return chalk.whiteBright;
+        case LlamaLogLevel.fatal: return chalk.redBright;
+        case LlamaLogLevel.error: return chalk.red;
+        case LlamaLogLevel.warn: return chalk.yellow;
+        case LlamaLogLevel.info: return chalk.whiteBright;
+        case LlamaLogLevel.debug: return chalk.white;
+        default:
+            void (level);
+            return chalk.whiteBright;
+    }
+}
+function prefixAndColorMessage(message, color) {
+    return getConsoleLogPrefix() + (message
+        .split("\n")
+        .map(line => color(line))
+        .join("\n" + getConsoleLogPrefix()));
+}
+function logMessageIsOnlyDots(message) {
+    if (message == null)
+        return false;
+    for (let i = 0; i < message.length; i++) {
+        if (message[i] !== ".")
+            return false;
+    }
+    return true;
+}
+//# sourceMappingURL=Llama.js.map

package/dist/bindings/Llama.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"Llama.js","sourceRoot":"","sources":["../../src/bindings/Llama.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,MAAM,OAAO,CAAC;AAC1B,OAAO,EAAC,aAAa,EAAE,UAAU,EAAE,QAAQ,EAAC,MAAM,iBAAiB,CAAC;AACpE,OAAO,EAAC,mBAAmB,EAAC,MAAM,iCAAiC,CAAC;AACpE,OAAO,EAAC,UAAU,EAAoB,MAAM,uCAAuC,CAAC;AACpF,OAAO,EAAC,YAAY,EAAC,MAAM,0BAA0B,CAAC;AAEtD,OAAO,EAAC,sBAAsB,EAAC,MAAM,wCAAwC,CAAC;AAC9E,OAAO,EAAC,YAAY,EAAsB,MAAM,8BAA8B,CAAC;AAE/E,OAAO,EAA4C,UAAU,EAAE,aAAa,EAAC,MAAM,YAAY,CAAC;AAChG,OAAO,EAAC,kBAAkB,EAAoB,MAAM,+BAA+B,CAAC;AAEpF,MAAM,4BAA4B,GAAuC,IAAI,GAAG,CAAC;IAC7E,CAAC,aAAa,CAAC,QAAQ,EAAE,CAAC,CAAC;IAC3B,CAAC,aAAa,CAAC,KAAK,EAAE,CAAC,CAAC;IACxB,CAAC,aAAa,CAAC,KAAK,EAAE,CAAC,CAAC;IACxB,CAAC,aAAa,CAAC,IAAI,EAAE,CAAC,CAAC;IACvB,CAAC,aAAa,CAAC,IAAI,EAAE,CAAC,CAAC;IACvB,CAAC,aAAa,CAAC,KAAK,EAAE,CAAC,CAAC;CAC3B,CAAC,CAAC;AACH,MAAM,4BAA4B,GAAuC,IAAI,GAAG,CAC5E,CAAC,GAAG,4BAA4B,CAAC,OAAO,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,GAAG,EAAE,KAAK,CAAC,EAAE,EAAE,CAAC,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,CAClF,CAAC;AACF,MAAM,eAAe,GAAG,CAAC,CAAC;AAE1B,MAAM,OAAO,KAAK;IACd,gBAAgB,CAAiB,SAAS,CAAgB;IAC1D,gBAAgB,CAAiB,oBAAoB,GAAG,IAAI,YAAY,EAAE,CAAC;IAC3E,gBAAgB,CAAiB,WAAW,GAAG,EAAE,CAAC;IAClD,gBAAgB,CAAiB,OAAO,CAAyC;IACjF,gBAAgB,CAAiB,iBAAiB,CAAqB;IACvE,gBAAgB,CAAiB,YAAY,CAAoB;IACjE,gBAAgB,CAAiB,MAAM,CAAU;IACjD,gBAAgB,CAAkB,IAAI,CAAe;IACrD,gBAAgB,CAAkB,UAAU,CAA4B;IACxE,gBAAgB,CAAkB,aAAa,CAAmC;IAClF,gBAAgB,CAAkB,sBAAsB,CAAU;IAClE,gBAAgB,CAAkB,aAAa,CAAU;IACzD,gBAAgB,CAAkB,cAAc,CAAU;IAC1D,gBAAgB,CAAkB,gBAAgB,CAGhD;IACF,gBAAgB,CAAS,OAAO,CAAoD;IACpF,gBAAgB,CAAS,SAAS,CAAgB;IAClD,gBAAgB,CAAS,WAAW,GAAkB,IAAI,CAAC;IAC3D,gBAAgB,CAAS,gBAAgB,GAAyB,IAAI,CAAC;IACvE,gBAAgB,CAAS,4BAA4B,GAAW,CAAC,CAAC;IAClE,gBAAgB,CAAS,YAAY,GAAkB,IAAI,CAAC;IAC5D,gBAAgB,CAAS,iBAAiB,GAAyB,IAAI,CAAC;IACxE,gBAAgB,CAAS,mBAAmB,GAAY,KAAK,CAAC;IAC9D,gBAAgB,CAAS,SAAS,GAAY,KAAK,CAAC;IAEpC,SAAS,GAAG,IAAI,UAAU,EAAQ,CAAC;IAEnD,YAAoB,EAChB,QAAQ,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,EAAE,YAAY,EAAE,eAAe,EAAE,KAAK,EAAE,GAAG,EAAE,gBAAgB,EAAE,WAAW,EAelH;QACG,IAAI,CAAC,SAAS,GAAG,QAAQ,CAAC;QAC1B,IAAI,CAAC,IAAI,GAAG,GAAG,CAAC;QAChB,IAAI,CAAC,sBAAsB,GAAG,QAAQ,CAAC,wBAAwB,EAAE,CAAC;QAClE,IAAI,CAAC,aAAa,GAAG,QAAQ,CAAC,eAAe,EAAE,CAAC;QAChD,IAAI,CAAC,cAAc,GAAG,QAAQ,CAAC,gBAAgB,EAAE,CAAC;QAClD,IAAI,CAAC,OAAO,GAAG,QAAQ,CAAC,SAAS,EAAE,CAAC;QACpC,IAAI,CAAC,MAAM,GAAG,KAAK,CAAC;QACpB,IAAI,CAAC,iBAAiB,GAAG,gBAAgB,CAAC;QAC1C,IAAI,CAAC,YAAY,GAAG,WAAW,CAAC;QAEhC,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC,MAAM;YACxB,CAAC,CAAC,aAAa,CAAC,KAAK;YACrB,CAAC,CAAC,CAAC,QAAQ,IAAI,aAAa,CAAC,KAAK,CAAC,CAAC;QACxC,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC;QACtB,IAAI,CAAC,UAAU,GAAG,SAAS,CAAC;QAC5B,IAAI,CAAC,aAAa,GAAG,MAAM,CAAC,MAAM,CAAC,EAAC,GAAG,YAAY,EAAC,CAAC,CAAC;QACtD,IAAI,CAAC,gBAAgB,GAAG,MAAM,CAAC,MAAM,CAAC;YAClC,IAAI,EAAE,eAAe,CAAC,IAAI;YAC1B,OAAO,EAAE,eAAe,CAAC,OAAO;SACnC,CAAC,CAAC;QAEH,IAAI,CAAC,4BAA4B,GAAG,IAAI,CAAC,4BAA4B,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACjF,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAE/C,IAAI,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC;YACf,IAAI,CAAC,SAAS,CAAC,SAAS,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;YAC3C,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAAC,4BAA4B,CAAC,GAAG,CAAC,IAAI,CAAC,SAAS,CAAC,IAAI,eAAe,CAAC,CAAC;QAC1G,CAAC;QAED,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAEvC,OAAO,CAAC,EAAE,CAAC,MAAM,EAAE,IAAI,CAAC,OAAO,CAAC,CAAC;IACrC,CAAC;IAEM,KAAK,CAAC,OAAO;QAChB,IAAI,IAAI,CAAC,SAAS;YACd,OAAO;QAEX,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;QACtB,IAAI,CAAC,SAAS,CAAC,aAAa,EAAE,CAAC;QAC/B,MAAM,IAAI,CAAC,oBAAoB,CAAC,kBAAkB,EAAE,CAAC;QACrD,MAAM,IAAI,CAAC,SAAS,CAAC,OAAO,EAAE,CAAC;IACnC,CAAC;IAED,cAAc;IACP,KAAK,CAAC,CAAC,MAAM,CAAC,YAAY,CAAC;QAC9B,MAAM,IAAI,CAAC,OAAO,EAAE,CAAC;IACzB,CAAC;IAED,IAAW,QAAQ;QACf,OAAO,IAAI,CAAC,SAAS,CAAC;IAC1B,CAAC;IAED,IAAW,GAAG;QACV,OAAO,IAAI,CAAC,IAAI,CAAC;IACrB,CAAC;IAED,IAAW,qBAAqB;QAC5B,OAAO,IAAI,CAAC,sBAAsB,CAAC;IACvC,CAAC;IAED,IAAW,YAAY;QACnB,OAAO,IAAI,CAAC,aAAa,CAAC;IAC9B,CAAC;IAED,IAAW,aAAa;QACpB,OAAO,IAAI,CAAC,cAAc,CAAC;IAC/B,CAAC;IAED,IAAW,QAAQ;QACf,OAAO,IAAI,CAAC,SAAS,CAAC;IAC1B,CAAC;IAED,IAAW,QAAQ,CAAC,KAAoB;QACpC,IAAI,CAAC,kBAAkB,EAAE,CAAC;QAE1B,IAAI,KAAK,KAAK,IAAI,CAAC,SAAS,IAAI,IAAI,CAAC,MAAM;YACvC,OAAO;QAEX,IAAI,CAAC,SAAS,CAAC,iBAAiB,CAAC,4BAA4B,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,eAAe,CAAC,CAAC;QAC7F,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC;IAC3B,CAAC;IAED,IAAW,MAAM;QACb,OAAO,IAAI,CAAC,OAAO,CAAC;IACxB,CAAC;IAED,IAAW,MAAM,CAAC,KAAsD;QACpE,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC;QAErB,IAAI,KAAK,KAAK,KAAK,CAAC,oBAAoB;YACpC,IAAI,CAAC,mBAAmB,GAAG,KAAK,CAAC;IACzC,CAAC;IAED,IAAW,SAAS;QAChB,OAAO,IAAI,CAAC,UAAU,CAAC;IAC3B,CAAC;IAED,IAAW,YAAY;QACnB,OAAO,IAAI,CAAC,aAAa,CAAC;IAC9B,CAAC;IAED,IAAW,eAAe;QACtB,OAAO,IAAI,CAAC,gBAAgB,CAAC;IACjC,CAAC;IAED,IAAW,UAAU;QACjB,IAAI,CAAC,kBAAkB,EAAE,CAAC;QAE1B,OAAO,IAAI,CAAC,SAAS,CAAC,UAAU,EAAE,CAAC;IACvC,CAAC;IAED;;;;;OAKG;IACH,IAAW,eAAe;QACtB,OAAO,IAAI,CAAC,YAAY,CAAC,IAAI,CAAC;IAClC,CAAC;IAEM,KAAK,CAAC,YAAY;QACrB,IAAI,CAAC,kBAAkB,EAAE,CAAC;QAE1B,MAAM,EAAC,KAAK,EAAE,IAAI,EAAC,GAAG,IAAI,CAAC,SAAS,CAAC,cAAc,EAAE,CAAC;QAEtD,OAAO;YACH,KAAK;YACL,IAAI;YACJ,IAAI,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,GAAG,IAAI,CAAC;SAClC,CAAC;IACN,CAAC;IAEM,KAAK,CAAC,iBAAiB;QAC1B,IAAI,CAAC,kBAAkB,EAAE,CAAC;QAE1B,MAAM,EAAC,WAAW,EAAC,GAAG,IAAI,CAAC,SAAS,CAAC,gBAAgB,EAAE,CAAC;QAExD,OAAO,WAAW,CAAC;IACvB,CAAC;IAEM,KAAK,CAAC,SAAS,CAAC,OAA0B;QAC7C,IAAI,CAAC,kBAAkB,EAAE,CAAC;QAE1B,OAAO,MAAM,QAAQ,CAAC,IAAI,CAAC,WAAW,EAAE,UAAU,CAAC,YAAY,EAAE,OAAO,CAAC,UAAU,EAAE,KAAK,IAAI,EAAE;YAC5F,IAAI,CAAC,kBAAkB,EAAE,CAAC;YAE1B,MAAM,qBAAqB,GAAG,IAAI,CAAC,oBAAoB,CAAC,2BAA2B,EAAE,CAAC;YACtF,IAAI,CAAC;gBACD,OAAO,MAAM,UAAU,CAAC,OAAO,CAAC,OAAO,EAAE,EAAC,MAAM,EAAE,IAAI,EAAC,CAAC,CAAC;YAC7D,CAAC;oBAAS,CAAC;gBACP,qBAAqB,CAAC,OAAO,EAAE,CAAC;YACpC,CAAC;QACL,CAAC,CAAC,CAAC;IACP,CAAC;IAEM,KAAK,CAAC,0BAA0B,CAA2C,MAAS;QACvF,OAAO,IAAI,sBAAsB,CAAI,IAAI,EAAE,MAAM,CAAC,CAAC;IACvD,CAAC;IAEM,KAAK,CAAC,aAAa,CAAC,IAA+C;QACtE,OAAO,MAAM,YAAY,CAAC,MAAM,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC;IACjD,CAAC;IAEM,KAAK,CAAC,aAAa,CAAC,OAA4B;QACnD,OAAO,IAAI,YAAY,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC;IAC3C,CAAC;IAED,gBAAgB;IACT,KAAK,CAAC,KAAK;QACd,MAAM,IAAI,CAAC,SAAS,CAAC,IAAI,EAAE,CAAC;IAChC,CAAC;IAED;;;OAGG;IACI,IAAI,CAAC,KAAoB,EAAE,OAAe;QAC7C,IAAI,CAAC,WAAW,CAAC,4BAA4B,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,eAAe,EAAE,OAAO,GAAG,IAAI,CAAC,CAAC;IACjG,CAAC;IAED,gBAAgB;IACR,WAAW,CAAC,KAAa,EAAE,OAAe;QAC9C,MAAM,aAAa,GAAG,4BAA4B,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,aAAa,CAAC,KAAK,CAAC;QAErF,IAAI,IAAI,CAAC,WAAW,IAAI,IAAI,IAAI,IAAI,CAAC,gBAAgB,IAAI,IAAI,IAAI,IAAI,CAAC,gBAAgB,IAAI,aAAa,EAAE,CAAC;YACtG,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,gBAAgB,EAAE,IAAI,CAAC,WAAW,CAAC,CAAC;YAC1D,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC;QAC5B,CAAC;QAED,MAAM,aAAa,GAAG,CAAC,IAAI,CAAC,WAAW,IAAI,EAAE,CAAC,GAAG,OAAO,CAAC;QAEzD,MAAM,gBAAgB,GAAG,aAAa,CAAC,WAAW,CAAC,IAAI,CAAC,CAAC;QACzD,MAAM,UAAU,GAAG,gBAAgB,GAAG,CAAC;YACnC,CAAC,CAAC,aAAa;YACf,CAAC,CAAC,aAAa,CAAC,KAAK,CAAC,CAAC,EAAE,gBAAgB,CAAC,CAAC;QAC/C,MAAM,OAAO,GAAG,gBAAgB,GAAG,CAAC;YAChC,CAAC,CAAC,EAAE;YACJ,CAAC,CAAC,aAAa,CAAC,KAAK,CAAC,gBAAgB,GAAG,CAAC,CAAC,CAAC;QAEhD,IAAI,UAAU,KAAK,EAAE;YACjB,IAAI,CAAC,WAAW,CAAC,aAAa,EAAE,UAAU,CAAC,CAAC;QAEhD,IAAI,OAAO,KAAK,EAAE,EAAE,CAAC;YACjB,IAAI,CAAC,WAAW,GAAG,OAAO,CAAC;YAC3B,IAAI,CAAC,gBAAgB,GAAG,aAAa,CAAC;YAEtC,cAAc,CAAC,IAAI,CAAC,4BAA4B,CAAC,CAAC;YAClD,IAAI,CAAC,4BAA4B,EAAE,CAAC;QACxC,CAAC;;YACG,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC;IAChC,CAAC;IAED,gBAAgB;IACR,4BAA4B;QAChC,IAAI,CAAC,4BAA4B,EAAE,CAAC;QACpC,IAAI,IAAI,CAAC,4BAA4B,KAAK,CAAC;YACvC,OAAO;QAEX,IAAI,IAAI,CAAC,WAAW,IAAI,IAAI,IAAI,IAAI,CAAC,gBAAgB,IAAI,IAAI,EAAE,CAAC;YAC5D,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,gBAAgB,EAAE,IAAI,CAAC,WAAW,CAAC,CAAC;YAC1D,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC;QAC5B,CAAC;IACL,CAAC;IAED,gBAAgB;IACR,WAAW,CAAC,KAAoB,EAAE,OAAe;QACrD,+FAA+F;QAC/F,2CAA2C;QAC3C,IAAI,oBAAoB,CAAC,OAAO,CAAC,IAAI,IAAI,CAAC,OAAO,KAAK,KAAK,CAAC,oBAAoB,EAAE,CAAC;YAC/E,IAAI,oBAAoB,CAAC,IAAI,CAAC,YAAY,CAAC,IAAI,KAAK,KAAK,IAAI,CAAC,iBAAiB,EAAE,CAAC;gBAC9E,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;YAClC,CAAC;iBAAM,CAAC;gBACJ,IAAI,CAAC,mBAAmB,GAAG,IAAI,CAAC;gBAChC,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,qBAAqB,CAAC,OAAO,EAAE,mBAAmB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YACrF,CAAC;QACL,CAAC;aAAM,CAAC;YACJ,IAAI,IAAI,CAAC,mBAAmB,EAAE,CAAC;gBAC3B,OAAO,CAAC,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;gBAC3B,IAAI,CAAC,mBAAmB,GAAG,KAAK,CAAC;YACrC,CAAC;YAED,IAAI,CAAC;gBACD,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;YACjC,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACX,sFAAsF;YAC1F,CAAC;QACL,CAAC;QAED,IAAI,CAAC,YAAY,GAAG,OAAO,CAAC;QAC5B,IAAI,CAAC,iBAAiB,GAAG,KAAK,CAAC;IACnC,CAAC;IAED,gBAAgB;IACR,OAAO;QACX,IAAI,IAAI,CAAC,WAAW,IAAI,IAAI,IAAI,IAAI,CAAC,gBAAgB,IAAI,IAAI,EAAE,CAAC;YAC5D,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,gBAAgB,EAAE,IAAI,CAAC,WAAW,CAAC,CAAC;YAC1D,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC;QAC5B,CAAC;IACL,CAAC;IAED,gBAAgB;IACR,kBAAkB;QACtB,IAAI,IAAI,CAAC,SAAS;YACd,MAAM,IAAI,aAAa,EAAE,CAAC;IAClC,CAAC;IAED,gBAAgB;IACT,MAAM,CAAC,KAAK,CAAC,OAAO,CAAC,EACxB,QAAQ,EAAE,SAAS,EAAE,aAAa,EAAE,QAAQ,EAAE,MAAM,EAAE,WAAW,EAAE,aAAa,GAAG,KAAK,EAAE,KAAK,EAUlG;QACG,MAAM,GAAG,GAAG,QAAQ,CAAC,UAAU,EAAE,IAAI,KAAK,CAAC;QAC3C,MAAM,gBAAgB,GAAG,IAAI,kBAAkB,CAAC,GAAG,EAAE;YACjD,MAAM,EAAC,KAAK,EAAE,IAAI,EAAC,GAAG,QAAQ,CAAC,cAAc,EAAE,CAAC;YAEhD,OAAO;gBACH,KAAK;gBACL,IAAI,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,GAAG,IAAI,CAAC;aAClC,CAAC;QACN,CAAC,CAAC,CAAC;QAEH,IAAI,mBAAsC,CAAC;QAC3C,IAAI,GAAG,KAAK,KAAK,IAAI,WAAW,KAAK,CAAC;YAClC,mBAAmB,GAAG,gBAAgB,CAAC,aAAa,CAAC,CAAC,CAAC,CAAC;aACvD,IAAI,WAAW,YAAY,QAAQ;YACpC,mBAAmB,GAAG,gBAAgB,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC,MAAM,gBAAgB,CAAC,cAAc,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC;;YAEnH,mBAAmB,GAAG,gBAAgB,CAAC,aAAa,CAAC,WAAW,CAAC,CAAC;QAEtE,MAAM,KAAK,GAAI,IAAI,KAAK,CAAC;YACrB,QAAQ;YACR,SAAS;YACT,YAAY,EAAE,aAAa,CAAC,YAAY,CAAC,kBAAkB;YAC3D,eAAe,EAAE;gBACb,IAAI,EAAE,aAAa,CAAC,YAAY,CAAC,QAAQ,CAAC,IAAI;gBAC9C,OAAO,EAAE,aAAa,CAAC,YAAY,CAAC,QAAQ,CAAC,OAAO;aACvD;YACD,QAAQ;YACR,MAAM;YACN,KAAK;YACL,GAAG;YACH,gBAAgB;YAChB,WAAW,EAAE,mBAAmB;SACnC,CAAC,CAAC;QAEH,IAAI,CAAC,aAAa;YACd,MAAM,KAAK,CAAC,KAAK,EAAE,CAAC;QAExB,OAAO,KAAK,CAAC;IACjB,CAAC;IAEM,MAAM,CAAC,oBAAoB,CAAC,KAAoB,EAAE,OAAe;QACpE,QAAQ,KAAK,EAAE,CAAC;YACZ,KAAK,aAAa,CAAC,QAAQ;gBACvB,MAAM;YACV,KAAK,aAAa,CAAC,KAAK;gBACpB,oEAAoE;gBACpE,OAAO,CAAC,IAAI,CAAC,qBAAqB,CAAC,OAAO,EAAE,mBAAmB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBACzE,MAAM;YACV,KAAK,aAAa,CAAC,KAAK;gBACpB,oEAAoE;gBACpE,OAAO,CAAC,IAAI,CAAC,qBAAqB,CAAC,OAAO,EAAE,mBAAmB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBACzE,MAAM;YACV,KAAK,aAAa,CAAC,IAAI;gBACnB,OAAO,CAAC,IAAI,CAAC,qBAAqB,CAAC,OAAO,EAAE,mBAAmB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBACzE,MAAM;YACV,KAAK,aAAa,CAAC,IAAI;gBACnB,OAAO,CAAC,IAAI,CAAC,qBAAqB,CAAC,OAAO,EAAE,mBAAmB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBACzE,MAAM;YACV,KAAK,aAAa,CAAC,KAAK;gBACpB,OAAO,CAAC,KAAK,CAAC,qBAAqB,CAAC,OAAO,EAAE,mBAAmB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBAC1E,MAAM;YACV;gBACI,KAAK,CAAC,KAAqB,CAAC,CAAC;gBAC7B,OAAO,CAAC,IAAI,CAAC,mBAAmB,EAAE,GAAG,mBAAmB,CAAC,aAAa,CAAC,IAAI,CAAC,CAAC,sBAAsB,KAAK,EAAE,CAAC,CAAC,CAAC;gBAC7G,OAAO,CAAC,GAAG,CAAC,qBAAqB,CAAC,OAAO,EAAE,mBAAmB,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAChF,CAAC;IAEL,CAAC;CACJ;AAED,SAAS,mBAAmB,CAAC,KAAoB;IAC7C,QAAQ,KAAK,EAAE,CAAC;QACZ,KAAK,aAAa,CAAC,QAAQ,CAAC,CAAC,OAAO,KAAK,CAAC,WAAW,CAAC;QACtD,KAAK,aAAa,CAAC,KAAK,CAAC,CAAC,OAAO,KAAK,CAAC,SAAS,CAAC;QACjD,KAAK,aAAa,CAAC,KAAK,CAAC,CAAC,OAAO,KAAK,CAAC,GAAG,CAAC;QAC3C,KAAK,aAAa,CAAC,IAAI,CAAC,CAAC,OAAO,KAAK,CAAC,MAAM,CAAC;QAC7C,KAAK,aAAa,CAAC,IAAI,CAAC,CAAC,OAAO,KAAK,CAAC,WAAW,CAAC;QAClD,KAAK,aAAa,CAAC,KAAK,CAAC,CAAC,OAAO,KAAK,CAAC,KAAK,CAAC;QAC7C;YACI,KAAK,CAAC,KAAqB,CAAC,CAAC;YAC7B,OAAO,KAAK,CAAC,WAAW,CAAC;IACjC,CAAC;AACL,CAAC;AAED,SAAS,qBAAqB,CAAC,OAAe,EAAE,KAAkC;IAC9E,OAAO,mBAAmB,EAAE,GAAG,CAC3B,OAAO;SACF,KAAK,CAAC,IAAI,CAAC;SACX,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;SACxB,IAAI,CAAC,IAAI,GAAG,mBAAmB,EAAE,CAAC,CAC1C,CAAC;AACN,CAAC;AAED,SAAS,oBAAoB,CAAC,OAAsB;IAChD,IAAI,OAAO,IAAI,IAAI;QACf,OAAO,KAAK,CAAC;IAEjB,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QACtC,IAAI,OAAO,CAAC,CAAC,CAAC,KAAK,GAAG;YAClB,OAAO,KAAK,CAAC;IACrB,CAAC;IAED,OAAO,IAAI,CAAC;AAChB,CAAC"}

package/dist/bindings/consts.d.ts

@@ -0,0 +1,2 @@
+import { BuildGpu } from "./types.js";
+export declare function getPrettyBuildGpuName(gpu: BuildGpu): string;

package/dist/bindings/consts.js

@@ -0,0 +1,11 @@
+const prettyBuildGpuNames = {
+    metal: "Metal",
+    cuda: "CUDA",
+    vulkan: "Vulkan"
+};
+export function getPrettyBuildGpuName(gpu) {
+    if (gpu == false)
+        return "no GPU";
+    return prettyBuildGpuNames[gpu] ?? ('"' + gpu + '"');
+}
+//# sourceMappingURL=consts.js.map

package/dist/bindings/consts.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"consts.js","sourceRoot":"","sources":["../../src/bindings/consts.ts"],"names":[],"mappings":"AAEA,MAAM,mBAAmB,GAA6C;IAClE,KAAK,EAAE,OAAO;IACd,IAAI,EAAE,MAAM;IACZ,MAAM,EAAE,QAAQ;CACnB,CAAC;AAEF,MAAM,UAAU,qBAAqB,CAAC,GAAa;IAC/C,IAAI,GAAG,IAAI,KAAK;QACZ,OAAO,QAAQ,CAAC;IAEpB,OAAO,mBAAmB,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,GAAG,GAAG,GAAG,GAAG,CAAC,CAAC;AACzD,CAAC"}

package/dist/bindings/getLlama.d.ts

@@ -0,0 +1,152 @@
+import { LlamaGpuType, LlamaLogLevel } from "./types.js";
+import { Llama } from "./Llama.js";
+export type LlamaOptions = {
+    /**
+     * The compute layer implementation type to use for llama.cpp.
+     * - **`"auto"`**: Automatically detect and use the best GPU available (Metal on macOS, and CUDA or Vulkan on Windows and Linux)
+     * - **`"metal"`**: Use Metal.
+     *   Only supported on macOS.
+     *   Enabled by default on Apple Silicon Macs.
+     * - **`"cuda"`**: Use CUDA.
+     * - **`"vulkan"`**: Use Vulkan.
+     * - **`false`**: Disable any GPU support and only use the CPU.
+     *
+     * `"auto"` by default.
+     */
+    gpu?: "auto" | LlamaGpuType | {
+        type: "auto";
+        exclude?: LlamaGpuType[];
+    };
+    /**
+     * Set the minimum log level for llama.cpp.
+     * Defaults to `"warn"`.
+     */
+    logLevel?: LlamaLogLevel;
+    /**
+     * Set a custom logger for llama.cpp logs.
+     */
+    logger?: (level: LlamaLogLevel, message: string) => void;
+    /**
+     * Set what build method to use.
+     * - **`"auto"`**: If a local build is found, use it.
+     * Otherwise, if a prebuilt binary is found, use it.
+     * Otherwise, build from source.
+     * - **`"never"`**: If a local build is found, use it.
+     * Otherwise, if a prebuilt binary is found, use it.
+     * Otherwise, throw a `NoBinaryFoundError` error.
+     * - **`"forceRebuild"`**: Always build from source.
+     * Be cautious with this option, as it will cause the build to fail on Windows when the binaries are in use by another process.
+     *
+     * If running from inside an Asar archive in Electron, building from source is not possible, so it'll never build from source.
+     * To allow building from source in Electron apps, make sure you ship `node-llama-cpp` as an unpacked module.
+     *
+     * Defaults to `"auto"`.
+     * On Electron, defaults to `"never"`.
+     */
+    build?: "auto" | "never" | "forceRebuild";
+    /**
+     * Set custom CMake options for llama.cpp
+     */
+    cmakeOptions?: Record<string, string>;
+    /**
+     * When a prebuilt binary is found, only use it if it was built with the same build options as the ones specified in `buildOptions`.
+     * Disabled by default.
+     */
+    existingPrebuiltBinaryMustMatchBuildOptions?: boolean;
+    /**
+     * Use prebuilt binaries if they match the build options.
+     * Enabled by default.
+     */
+    usePrebuiltBinaries?: boolean;
+    /**
+     * Print binary compilation progress logs.
+     * Enabled by default.
+     */
+    progressLogs?: boolean;
+    /**
+     * Don't download llama.cpp source if it's not found.
+     * When set to `true`, and llama.cpp source is not found, a `NoBinaryFoundError` error will be thrown.
+     * Disabled by default.
+     */
+    skipDownload?: boolean;
+    /**
+     * Pad the available VRAM for the memory size calculations, as these calculations are not always accurate.
+     * Recommended to ensure stability.
+     * This only affects the calculations of `"auto"` in function options and is not reflected in the `getVramState` function.
+     *
+     * Defaults to `6%` of the total VRAM or 1GB, whichever is lower.
+     * Set to `0` to disable.
+     */
+    vramPadding?: number | ((totalVram: number) => number);
+    /**
+     * Enable debug mode to find issues with llama.cpp.
+     * Makes logs print directly to the console from `llama.cpp` and not through the provided logger.
+     *
+     * Defaults to `false`.
+     *
+     * The default can be set using the `NODE_LLAMA_CPP_DEBUG` environment variable.
+     */
+    debug?: boolean;
+};
+export type LastBuildOptions = {
+    /**
+     * Set the minimum log level for llama.cpp.
+     * Defaults to "debug".
+     */
+    logLevel?: LlamaLogLevel;
+    /**
+     * Set a custom logger for llama.cpp logs.
+     */
+    logger?: (level: LlamaLogLevel, message: string) => void;
+    /**
+     * If a local build is not found, use prebuilt binaries.
+     * Enabled by default.
+     */
+    usePrebuiltBinaries?: boolean;
+    /**
+     * If a local build is not found, and prebuilt binaries are not found, when building from source,
+     * print binary compilation progress logs.
+     * Enabled by default.
+     */
+    progressLogs?: boolean;
+    /**
+     * If a local build is not found, and prebuilt binaries are not found, don't download llama.cpp source if it's not found.
+     * When set to `true`, and llama.cpp source is needed but is not found, a `NoBinaryFoundError` error will be thrown.
+     * Disabled by default.
+     */
+    skipDownload?: boolean;
+    /**
+     * Pad the available VRAM for the memory size calculations, as these calculations are not always accurate.
+     * Recommended to ensure stability.
+     * This only affects the calculations of `"auto"` in function options and is not reflected in the `getVramState` function.
+     *
+     * Defaults to `6%` of the total VRAM or 1GB, whichever is lower.
+     * Set to `0` to disable.
+     */
+    vramPadding?: number | ((totalVram: number) => number);
+    /**
+     * Enable debug mode to find issues with llama.cpp.
+     * Makes logs print directly to the console from `llama.cpp` and not through the provided logger.
+     *
+     * Defaults to `false`.
+     *
+     * The default can be set using the `NODE_LLAMA_CPP_DEBUG` environment variable.
+     */
+    debug?: boolean;
+};
+export declare const getLlamaFunctionName = "getLlama";
+export declare const defaultLlamaVramPadding: (totalVram: number) => number;
+/**
+ * Get a `llama.cpp` binding.
+ *
+ * Defaults to use a local binary built using the `download` or `build` CLI commands if one exists,
+ * otherwise, uses a prebuilt binary, and fallbacks to building from source if a prebuilt binary is not found.
+ *
+ * Pass `"lastBuild"` to default to use the last successful build created using the `download` or `build` CLI commands if one exists.
+ */
+export declare function getLlama(options?: LlamaOptions): Promise<Llama>;
+export declare function getLlama(type: "lastBuild", lastBuildOptions?: LastBuildOptions): Promise<Llama>;
+export declare function getLlamaForOptions({ gpu, logLevel, logger, build, cmakeOptions, existingPrebuiltBinaryMustMatchBuildOptions, usePrebuiltBinaries, progressLogs, skipDownload, vramPadding, debug }: LlamaOptions, { updateLastBuildInfoOnCompile, skipLlamaInit }?: {
+    updateLastBuildInfoOnCompile?: boolean;
+    skipLlamaInit?: boolean;
+}): Promise<Llama>;