@openplait/core

Portable OpenPlait definitions with no database client, network execution, renderer, or application dependency.

Included in v1alpha1

• Query resources in semantic and native modes
• Dashboard resources with datasources, variables, panels, and grid layouts
• Time series, stat, and table visualization specifications
• Standalone and panel-local Transformation IR with execution preferences
• Independently executable AlertRule resources and lifecycle policy types
• Recursive boolean filters and arithmetic value expressions
• Select, aggregate, group, time bucket, order, limit, and basic join types
• Normalized QueryResult, DataFrame, and DataField types
• JSON Schemas for Query, Dashboard, TransformationPipeline, and AlertRule resources and normalized results
• Structural and cross-field validation
• JSON and YAML parsing/serialization
• Specification and API version constants

Usage

import {
  OPENPLAIT_API_VERSION,
  parseDashboard,
  parseQuery,
  validateQueryResult,
  type SemanticQuery,
} from "@openplait/core";

const query: SemanticQuery = {
  apiVersion: OPENPLAIT_API_VERSION,
  kind: "Query",
  metadata: { name: "requests-by-model" },
  spec: {
    mode: "semantic",
    datasource: { kind: "ClickHouseDatasource", name: "primary" },
    input: { signal: "traces", entity: "otel.spans" },
    select: [
      { field: "gen_ai.request.model", as: "model" },
      { aggregate: { function: "count" }, as: "requests" },
    ],
    groupBy: [{ field: "gen_ai.request.model" }],
  },
};

const parsed = parseQuery(queryYamlSource);
const dashboard = parseDashboard(dashboardYamlSource);
const result = validateQueryResult(adapterOutput);

The schemas are exported from JavaScript as schemas and as package files under @openplait/core/schemas/v1alpha1/*.

JSON Schema verifies document structure. The runtime validator additionally checks invariants JSON Schema cannot express cleanly, including unique query output names, named dashboard datasource bindings, dashboard reference integrity, grid bounds, and dataframe field lengths matching their frame length. Datasource plugin kinds remain open-ended; ClickHouse is simply the first implemented adapter.