import type { BackendObservability, ChatMessage, EmbedRuntimeOptions, EngineExecutionMode, GenerateRequestId, GenerateResponse, NativeRuntimeConfig, PromptOptions, RequestObservabilityMetrics, TransportObservability } from '../engine/inference-types.js';
import type { ClassifiedAsset, InternalBundleDescriptor, ModelDetectionResult, PairingPlan, RegistryManifest, StagedModelBundle, StageModelBundleOptions } from '../models/types.js';
import type { ChatBoundaryInfo } from '../engine/chat-boundary-sanitizer.js';
import type { BrowserCacheLayout, GgufReadAtCallbacks, GgufSplitStreamCallbacks, RustLifecycleBridge } from '../wasm/wasm-bridge.js';
import type { RuntimeBackendOverride, WasmThreadingMode } from '../engine/runtime-assets.js';
export interface EngineRuntime {
    getExecutionMode(): EngineExecutionMode;
    getWasmThreadingMode(): WasmThreadingMode;
    getDefaultBackendOverride(): RuntimeBackendOverride | null;
    getTransportObservability(): TransportObservability;
    initModule(): Promise<void>;
    stageModelBundle(descriptor: InternalBundleDescriptor, options?: StageModelBundleOptions): Promise<StagedModelBundle>;
    loadRuntimeModel(modelPathOrBundle: string | StagedModelBundle, config?: NativeRuntimeConfig): Promise<void>;
    close(): void;
    getChatTemplate(): string | null;
    readMediaMarker(): string | null;
    /**
     * Returns the model's BOS token rendered as text, or '' if the model has
     * no BOS token. Used by the character-agent custom template builder to
     * emit the correct leading special token per model.
     */
    getBosText(): string;
    /** Returns the model's EOS token rendered as text, or '' if unavailable. */
    getEosText(): string;
    browserCacheLayout(sourceBytes: number, sourceBytesKnown: boolean, directLoadMaxBytes: number, shardMaxBytes: number): Promise<BrowserCacheLayout>;
    planGgufSplitCount(sourceBytes: number, shardMaxBytes: number, callbacks: GgufReadAtCallbacks): Promise<number>;
    splitGgufStream(sourceBytes: number, outputPrefix: string, shardMaxBytes: number, callbacks: GgufSplitStreamCallbacks): Promise<void>;
    detectModelFromGgufFile(file: Blob & {
        name?: string;
    }, signal?: AbortSignal): Promise<ModelDetectionResult>;
    resolvePairing(classified: readonly ClassifiedAsset[]): Promise<PairingPlan>;
    createRustLifecycleBridge(manifest: RegistryManifest): Promise<RustLifecycleBridge>;
    probeChatTemplateBoundaryInfo(): Promise<ChatBoundaryInfo>;
    enqueueChat(contextKey: string, messages: readonly ChatMessage[], options?: number | PromptOptions): Promise<GenerateRequestId>;
    cancelQuery(requestId: GenerateRequestId): Promise<boolean>;
    enqueueQuery(contextKey: string, promptText: string, options?: number | PromptOptions): Promise<GenerateRequestId>;
    enqueueEmbedding(contextKey: string, input: string, options?: EmbedRuntimeOptions): Promise<GenerateRequestId>;
    awaitQuery(requestId: GenerateRequestId, options?: {
        signal?: AbortSignal;
    }): Promise<GenerateResponse>;
    getRuntimeObservability(): RequestObservabilityMetrics | null;
    getBackendObservability(): Promise<BackendObservability | null>;
}