import { ExportResult } from '@opentelemetry/core';
import { Context } from '@opentelemetry/api';
import { Span as OTelSpan, ReadableSpan as OTelReadableSpan, SpanProcessor, SpanExporter } from '@opentelemetry/sdk-trace-base';
import { MlflowClient } from '../clients';
import { UnityCatalogLocation } from '../core/entities/trace_location';
/**
 * Span processor for Databricks Unity Catalog backed traces.
 *
 * Mirrors Python's `DatabricksUCTableSpanProcessor`:
 *   - generates V4 trace IDs (`trace:/<location>/<hex>`)
 *   - constructs TraceInfo with the UC trace location, V4 schema version, and
 *     any context-injected metadata/tags so that `updateCurrentTrace({ tags })`
 *     can mutate the in-memory record before export
 *   - delegates to a UC-aware SpanExporter for trace info + OTLP span upload
 *
 * Wired up by `init({ trackingUri, experimentId })` when the linked Databricks
 * experiment carries `mlflow.experiment.databricksTrace*` tags pointing at a UC
 * destination.
 */
export declare class DatabricksUCTableSpanProcessor implements SpanProcessor {
    private _exporter;
    private _location;
    constructor(exporter: SpanExporter, location: UnityCatalogLocation);
    onStart(span: OTelSpan, _parentContext: Context): void;
    onEnd(span: OTelReadableSpan): void;
    private updateTraceInfo;
    shutdown(): Promise<void>;
    forceFlush(): Promise<void>;
}
/**
 * Exporter for Databricks Unity Catalog backed traces.
 *
 * Mirrors Python's `DatabricksUCTableSpanExporter`. Two-step export per trace:
 *   1. Call the V4 `CreateTraceInfo` REST endpoint with the full `TraceInfo`
 *      (including `tags`, `trace_metadata`, state, durations). This is what
 *      makes `updateCurrentTrace({ tags })` persist as trace-level tags in
 *      `_traces_unified`, `get_trace`, `search_traces`, and the UI.
 *   2. Upload the OTel spans for this trace to `/api/2.0/otel/v1/traces` with
 *      the `X-Databricks-UC-Table-Name` header pointing to the trace's UC
 *      spans table.
 *
 * Errors in either step are logged but do not surface to user code, matching
 * the non-fatal behavior of `_log_spans` on the Python side.
 */
export declare class DatabricksUCTableSpanExporter implements SpanExporter {
    private _client;
    private _pendingExports;
    private _hasRaisedMissingSpansTableWarning;
    private _hasRaisedSpanExportError;
    constructor(client: MlflowClient);
    export(spans: OTelReadableSpan[], resultCallback: (result: ExportResult) => void): void;
    private exportTraceToBackend;
    forceFlush(): Promise<void>;
    shutdown(): Promise<void>;
}