@elevasis/core 0.2.1 → 0.4.0

package/src/execution/calibration/types.ts

@@ -1,190 +0,0 @@
-/**
- * Calibration Lab Type Definitions
- * Core types for AI model configuration optimization with cost/performance comparison
- */
-
-import type { AgentDefinition } from '../engine/agent/core/types'
-import type { WorkflowDefinition } from '../engine/workflow/types'
-import type { ExecutionMetricsDetail } from '../../operations/observability/types'
-
-// ============================================================================
-// Calibration Project Types
-// ============================================================================
-
-/**
- * Calibration project - groups related optimization runs
- * Provides organizational structure for iterative testing
- */
-export interface CalibrationProject {
-  id: string
-  organizationId: string
-  resourceId: string
-  resourceType: 'agent' | 'workflow'
-  name: string
-  description?: string | null
-  createdAt: Date
-  updatedAt: Date
-}
-
-// ============================================================================
-// Configuration Variant Types
-// ============================================================================
-
-/**
- * Configuration variant for testing
- * Defines what to override in the base agent/workflow definition
- */
-export interface ConfigVariant {
-  variantName: string
-  definitionOverrides?: AgentCalibrationOverrides | WorkflowCalibrationOverrides
-}
-
-/**
- * Agent definition overrides for calibration
- * Excludes identity/metadata fields that should not be modified
- */
-export type AgentCalibrationOverrides = Omit<
-  Partial<AgentDefinition>,
-  'config' | 'contract' | 'tools' | 'metricsConfig' | 'interface'
->
-
-/**
- * Workflow definition overrides for calibration
- * Excludes identity/metadata fields that should not be modified
- */
-export type WorkflowCalibrationOverrides = Omit<
-  Partial<WorkflowDefinition>,
-  'config' | 'contract' | 'steps' | 'entryPoint' | 'metricsConfig' | 'interface'
->
-
-// ============================================================================
-// Grading System Types
-// ============================================================================
-
-/**
- * Grading rubric for LLM-as-judge evaluation
- * Defines criteria and passing threshold
- */
-export interface GradingRubric {
-  passingThreshold: number // 0-1, e.g., 0.7 means 70% to pass
-  criteria: GradingCriterion[]
-}
-
-/**
- * Individual grading criterion
- * Weight should be 0-1, and all weights should sum to 1
- */
-export interface GradingCriterion {
-  name: string
-  weight: number // 0-1, weights should sum to 1
-  description: string
-  scoringGuide: string
-}
-
-/**
- * Grading result for a single execution or session
- * Contains overall score and per-criterion breakdown
- */
-export interface GradeResult {
-  score: number // 0-1
-  passed: boolean
-  details: Record<string, { score: number; justification: string }>
-}
-
-// ============================================================================
-// Result Types
-// ============================================================================
-
-/**
- * Single-turn calibration result
- * One result per (variant × input) combination
- */
-export interface SingleCalibrationResult {
-  executionId: string // Reference to execution_logs
-  variantName: string
-  inputIndex: number // Which input from testInputs array (0-based)
-  appliedOverrides?: AgentCalibrationOverrides | WorkflowCalibrationOverrides
-  status: 'pending' | 'running' | 'completed' | 'failed'
-  errorMessage?: string
-  grade?: GradeResult
-  gradeError?: string
-}
-
-/**
- * Multi-turn session calibration result
- * Leverages existing sessions infrastructure
- */
-export interface SessionCalibrationResult {
-  sessionId: string // Reference to sessions table
-  variantName: string
-  appliedOverrides?: AgentCalibrationOverrides | WorkflowCalibrationOverrides
-  status: 'pending' | 'running' | 'completed' | 'failed'
-  errorMessage?: string
-  turnCount: number
-  grade?: GradeResult
-  gradeError?: string
-}
-
-// ============================================================================
-// Calibration Run Types
-// ============================================================================
-
-/**
- * Calibration run - individual test execution within a project
- * Contains configuration, results, and grading information
- */
-export interface CalibrationRun {
-  id: string
-  organizationId: string
-  projectId: string
-  name: string
-  description?: string | null
-  executionMode: 'single' | 'session'
-  testInputs: unknown[]
-  configVariants: ConfigVariant[]
-  gradingRubric?: GradingRubric | null
-  graderModel?: string | null
-  results: (SingleCalibrationResult | SessionCalibrationResult)[]
-  status: 'pending' | 'running' | 'completed' | 'partial' | 'failed'
-  createdAt: Date
-  completedAt?: Date | null
-}
-
-// ============================================================================
-// Combined Response Types
-// ============================================================================
-
-/**
- * Execution log subset for comparison view
- * Contains essential data without full execution context
- */
-export interface ExecutionLog {
-  executionId: string
-  status: 'running' | 'completed' | 'failed'
-  input: unknown
-  output: unknown | null
-  error: CalibrationExecutionError | null
-  startedAt: string
-  completedAt: string | null
-}
-
-/**
- * Execution error information for calibration display
- * Subset of full execution error for UI display purposes
- * Named differently from ExecutionError class to avoid type conflicts
- */
-export interface CalibrationExecutionError {
-  message: string
-  category?: string
-  type?: string
-}
-
-/**
- * Combined calibration run data for comparison view
- * Single API call returns run + execution logs + metrics
- */
-export interface CalibrationRunWithFullData {
-  run: CalibrationRun
-  logs: Record<string, ExecutionLog> // keyed by executionId
-  metrics: Record<string, ExecutionMetricsDetail> // keyed by executionId
-}