import { AuthProvider } from '../auth';
/**
 * User-facing shape for specifying a Databricks Unity Catalog trace location.
 *
 * All three fields are required. Unlike Python's
 * `mlflow.set_experiment(trace_location=UnityCatalog(...))`, the TS SDK does
 * not upsert the underlying UC trace location, so it can't default
 * `tablePrefix` to anything sensible - the customer must point at a
 * trace location that is already provisioned in the workspace.
 */
export interface UnityCatalogLocationOptions {
    catalogName: string;
    schemaName: string;
    tablePrefix: string;
}
/**
 * Configuration options for the MLflow tracing SDK
 */
export interface MLflowTracingConfig {
    /**
     * The host where traces will be logged.
     * Can be:
     * - HTTP URI for MLflow trace server backend
     * - "databricks" (uses default profile)
     * - "databricks://profile" (uses specific profile)
     */
    trackingUri: string;
    /**
     * The experiment ID where traces will be logged
     */
    experimentId: string;
    /**
     * The location of the Databricks config file, default to ~/.databrickscfg
     */
    databricksConfigPath?: string;
    /**
     * The Databricks host. If not provided, the host will be read from the Databricks config file.
     * @deprecated For Databricks auth, prefer using environment variables or config profiles.
     * The auth system will automatically resolve credentials.
     */
    host?: string;
    /**
     * The Databricks token. If not provided, the token will be read from the Databricks config file.
     * @deprecated For Databricks auth, prefer using environment variables or config profiles.
     * The auth system will automatically resolve credentials.
     */
    databricksToken?: string;
    /**
     * The tracking server username for basic auth.
     */
    trackingServerUsername?: string;
    /**
     * The tracking server password for basic auth.
     */
    trackingServerPassword?: string;
    /**
     * Bearer token for OSS MLflow tracking servers.
     * Can also be set via MLFLOW_TRACKING_TOKEN environment variable.
     */
    trackingServerToken?: string;
    /**
     * Workspace name for MLflow servers with workspaces enabled (OSS/non-Databricks only).
     * Sets the X-MLFLOW-WORKSPACE header on requests to self-hosted MLflow.
     * Can also be set via MLFLOW_WORKSPACE environment variable (takes precedence).
     */
    workspace?: string;
    /**
     * Optional Databricks Unity Catalog trace location. When provided, the SDK
     * generates V4 trace IDs and persists trace tags / metadata via the V4
     * `CreateTraceInfo` endpoint, mirroring Python's
     * `mlflow.set_experiment(trace_location=UnityCatalog(...))`. When omitted,
     * traces use the V3 experiment-backed path.
     *
     * The UC trace location must already be provisioned in the workspace; the
     * TS SDK does not upsert it.
     */
    traceLocation?: UnityCatalogLocationOptions;
}
/**
 * Initialization options for the MLflow tracing SDK. The trackingUri and experimentId
 * can be omitted and will be resolved from environment variables when available. Since
 * this is used on the client side, we need to make sure the trackingUri and experimentId
 * are required.
 */
export type MLflowTracingInitOptions = Partial<MLflowTracingConfig>;
/**
 * Configure the MLflow tracing SDK with tracking location settings.
 * This must be called before using other tracing functions.
 *
 * Call once per process. Calling `init()` again will tear down the existing
 * OpenTelemetry SDK and start a fresh one, but the underlying NodeSDK
 * shutdown is asynchronous and may race with subsequent tracer registration;
 * configuration is intended to be set up once at startup, not reconfigured
 * mid-process.
 *
 * ## Authentication
 *
 * ### Databricks (trackingUri = "databricks" or "databricks://profile")
 *
 * Authentication is handled automatically by the Databricks SDK. It supports:
 * - Personal Access Tokens (DATABRICKS_TOKEN env var)
 * - OAuth M2M service principals (DATABRICKS_CLIENT_ID/DATABRICKS_CLIENT_SECRET)
 * - Azure CLI / Azure MSI / Azure Client Secret
 * - Google Cloud credentials
 * - Config file profiles (~/.databrickscfg)
 *
 * For Databricks Apps, credentials are automatically injected by the runtime.
 *
 * ### OSS MLflow (trackingUri = HTTP URL)
 *
 * Authentication is resolved in priority order:
 * 1. Basic Auth (trackingServerUsername/Password or MLFLOW_TRACKING_USERNAME/PASSWORD)
 * 2. Bearer Token (trackingServerToken or MLFLOW_TRACKING_TOKEN)
 * 3. No Auth
 *
 * @param config Configuration object with trackingUri and experimentId
 *
 * @example
 * ```typescript
 * import { init, withSpan } from '@mlflow/core';
 *
 * // Option 1: Use MLflow Tracking Server
 * init({
 *   trackingUri: "http://localhost:5000",
 *   experimentId: "123456789"
 * });
 *
 * // Option 2: Use default Databricks profile from ~/.databrickscfg
 * init({
 *   trackingUri: "databricks",
 *   experimentId: "123456789"
 * });
 *
 * // Option 3: Use specific Databricks profile
 * init({
 *   trackingUri: "databricks://my-profile",
 *   experimentId: "123456789"
 * });
 *
 * // Option 4: Custom config file path
 * init({
 *   trackingUri: "databricks",
 *   experimentId: "123456789",
 *   databricksConfigPath: "/path/to/my/databrickscfg"
 * });
 *
 * // Option 5: Override with explicit host/token (backwards compatibility)
 * init({
 *   trackingUri: "databricks",
 *   experimentId: "123456789",
 *   host: "https://my-workspace.databricks.com",
 *   databricksToken: "my-token"
 * });
 *
 * // Option 6: OSS MLflow with basic auth
 * init({
 *   trackingUri: "http://localhost:5000",
 *   experimentId: "123456789",
 *   trackingServerUsername: "user",
 *   trackingServerPassword: "pass"
 * });
 *
 * // Option 7: OSS MLflow with bearer token
 * init({
 *   trackingUri: "http://localhost:5000",
 *   experimentId: "123456789",
 *   trackingServerToken: "my-token"
 * });
 *
 * // Option 8: Databricks Unity Catalog trace location.
 * // Mirrors Python's `mlflow.set_experiment(..., trace_location=UnityCatalog(...))`.
 * // The UC trace location must already be provisioned for this workspace.
 * init({
 *   trackingUri: "databricks",
 *   experimentId: "123456789",
 *   traceLocation: {
 *     catalogName: "my_catalog",
 *     schemaName: "my_schema",
 *     tablePrefix: "my_prefix",
 *   },
 * });
 *
 * // Now you can use tracing functions
 * function add(a: number, b: number) {
 *   return withSpan(
 *     { name: 'add', inputs: { a, b } },
 *     (span) => {
 *       const result = a + b;
 *       span.setOutputs({ result });
 *       return result;
 *     }
 *   );
 * }
 * ```
 */
export declare function init(config: MLflowTracingInitOptions): void;
/**
 * Get the current configuration. Throws an error if not configured.
 * @returns The current MLflow tracing configuration
 */
export declare function getConfig(): MLflowTracingConfig;
/**
 * Get the current authentication provider. Throws an error if not configured.
 * @returns The current authentication provider
 */
export declare function getAuthProvider(): AuthProvider;
/**
 * Reset the global configuration. For testing purposes only.
 * @internal
 */
export declare function resetConfig(): void;