PromptTuner MCP

// Zod output schemas for PromptTuner MCP import { z } from 'zod'; import { OPTIMIZATION_TECHNIQUES, TARGET_FORMATS } from '../config/types.js'; /** Schema for error details in responses */ const ErrorSchema = z.object({ code: z.string().describe('Error code (e.g., E_INVALID_INPUT)'), message: z.string().describe('Human-readable error message'), context: z.string().optional().describe('Additional context (truncated)'), details: z .record(z.unknown()) .optional() .describe('Additional error details'), }); /** Schema for prompt analysis scores (0-100 for each dimension) */ const ScoreSchema = z.object({ clarity: z.number().min(0).max(100).describe('How clear and unambiguous'), specificity: z.number().min(0).max(100).describe('How specific and detailed'), completeness: z .number() .min(0) .max(100) .describe('How complete with context'), structure: z.number().min(0).max(100).describe('How well-organized'), effectiveness: z.number().min(0).max(100).describe('Predicted effectiveness'), overall: z.number().min(0).max(100).describe('Weighted overall score'), }); /** Schema for detected prompt characteristics */ const CharacteristicsSchema = z.object({ detectedFormat: z.enum(TARGET_FORMATS).describe('Detected target format'), hasExamples: z.boolean().describe('Contains examples'), hasRoleContext: z.boolean().describe('Has role/persona defined'), hasStructure: z.boolean().describe('Has XML/Markdown structure'), hasStepByStep: z.boolean().describe('Has step-by-step guidance'), wordCount: z.number().describe('Total word count'), estimatedComplexity: z .enum(['simple', 'moderate', 'complex']) .describe('Estimated complexity'), }); const techniqueSchema = z.enum(OPTIMIZATION_TECHNIQUES); export const RefinePromptOutputSchema = { ok: z.boolean(), original: z.string().optional(), refined: z.string().optional(), corrections: z.array(z.string()).optional(), technique: techniqueSchema.optional(), targetFormat: z.enum(TARGET_FORMATS).optional(), usedFallback: z.boolean().optional(), fromCache: z.boolean().optional(), error: ErrorSchema.optional(), }; export const AnalyzePromptOutputSchema = { ok: z.boolean(), hasTypos: z.boolean().optional(), isVague: z.boolean().optional(), missingContext: z.boolean().optional(), suggestions: z.array(z.string()).optional(), score: ScoreSchema.optional(), characteristics: CharacteristicsSchema.optional(), error: ErrorSchema.optional(), }; export const OptimizePromptOutputSchema = { ok: z.boolean(), original: z.string().optional(), optimized: z.string().optional(), techniquesApplied: z.array(techniqueSchema).optional(), targetFormat: z.enum(TARGET_FORMATS).optional(), beforeScore: ScoreSchema.optional(), afterScore: ScoreSchema.optional(), improvements: z.array(z.string()).optional(), usedFallback: z.boolean().optional(), error: ErrorSchema.optional(), }; export const DetectFormatOutputSchema = { ok: z.boolean(), detectedFormat: z.enum(TARGET_FORMATS).optional(), confidence: z.number().min(0).max(100).optional(), characteristics: CharacteristicsSchema.optional(), recommendation: z.string().optional(), error: ErrorSchema.optional(), }; export const ComparePromptsOutputSchema = { ok: z.boolean(), promptA: z.string().optional(), promptB: z.string().optional(), scoreA: ScoreSchema.optional(), scoreB: ScoreSchema.optional(), scoreDelta: ScoreSchema.optional(), winner: z.enum(['A', 'B', 'tie']).optional(), improvements: z.array(z.string()).optional(), regressions: z.array(z.string()).optional(), recommendation: z.string().optional(), error: ErrorSchema.optional(), }; export const ValidatePromptOutputSchema = { ok: z.boolean(), isValid: z.boolean().optional(), issues: z .array( z.object({ type: z.enum(['error', 'warning', 'info']), message: z.string(), suggestion: z.string().optional(), }) ) .optional(), tokenEstimate: z.number().optional(), securityFlags: z.array(z.string()).optional(), error: ErrorSchema.optional(), };