|
| 1 | +/** |
| 2 | + * @license |
| 3 | + * Copyright 2025 BOSS Ghost MCP |
| 4 | + * SPDX-License-Identifier: Apache-2.0 |
| 5 | + */ |
| 6 | + |
| 7 | +import {zod} from '../third_party/index.js'; |
| 8 | + |
| 9 | +/** |
| 10 | + * Zod Schema Type Definitions |
| 11 | + * |
| 12 | + * Provides type-safe wrappers for Zod schema operations to replace `any` casts |
| 13 | + * with properly typed alternatives. This module bridges schema definition and |
| 14 | + * extraction logic. |
| 15 | + */ |
| 16 | + |
| 17 | +/** |
| 18 | + * Zod Schema Shape - Type representation of object shape |
| 19 | + * Replaces Record<string, any> for schema definitions |
| 20 | + */ |
| 21 | +export type ZodSchemaShape = Record<string, zod.ZodTypeAny>; |
| 22 | + |
| 23 | +/** |
| 24 | + * Zod Object Schema - Type-safe wrapper for ZodObject |
| 25 | + * Replaces zod.ZodObject<any> throughout the codebase |
| 26 | + */ |
| 27 | +export type ZodObjectSchema = zod.ZodObject<ZodSchemaShape>; |
| 28 | + |
| 29 | +/** |
| 30 | + * Schema Field Definition - Represents a single field in a schema |
| 31 | + * Supports: string, number, boolean, array types |
| 32 | + */ |
| 33 | +export interface SchemaFieldDef { |
| 34 | + type: |
| 35 | + | 'string' |
| 36 | + | 'number' |
| 37 | + | 'boolean' |
| 38 | + | 'string[]' |
| 39 | + | 'number[]' |
| 40 | + | 'boolean[]' |
| 41 | + | 'date' |
| 42 | + | 'unknown'; |
| 43 | + optional: boolean; |
| 44 | +} |
| 45 | + |
| 46 | +/** |
| 47 | + * Extraction Result Type - Generic extraction output |
| 48 | + * Represents extracted data from LLM or DOM extraction |
| 49 | + */ |
| 50 | +export type ExtractionResult = Record<string, unknown>; |
| 51 | + |
| 52 | +/** |
| 53 | + * Zod Schema Definition - Internal schema type representation |
| 54 | + * Used for accessing _def properties safely |
| 55 | + */ |
| 56 | +export interface ZodSchemaDef { |
| 57 | + type: string; |
| 58 | + element?: {_def?: {type?: string}}; |
| 59 | + innerType?: {_def?: {type?: string}}; |
| 60 | +} |
| 61 | + |
| 62 | +/** |
| 63 | + * Type guard for checking if a value is a ZodTypeAny |
| 64 | + */ |
| 65 | +export function isZodType(value: unknown): value is zod.ZodTypeAny { |
| 66 | + return ( |
| 67 | + value !== null && |
| 68 | + typeof value === 'object' && |
| 69 | + '_def' in value && |
| 70 | + typeof (value as any)._def === 'object' |
| 71 | + ); |
| 72 | +} |
| 73 | + |
| 74 | +/** |
| 75 | + * Safe schema shape accessor |
| 76 | + * Retrieves the shape property from a ZodObject schema safely |
| 77 | + */ |
| 78 | +export function getZodSchemaShape( |
| 79 | + schema: ZodObjectSchema |
| 80 | +): ZodSchemaShape | null { |
| 81 | + if (!isZodType(schema)) { |
| 82 | + return null; |
| 83 | + } |
| 84 | + |
| 85 | + // eslint-disable-next-line @typescript-eslint/no-explicit-any |
| 86 | + const def = (schema as any)._def; |
| 87 | + return def?.shape || null; |
| 88 | +} |
| 89 | + |
| 90 | +/** |
| 91 | + * Safe schema definition accessor |
| 92 | + * Retrieves the _def property from any Zod type |
| 93 | + * Note: We use 'any' here because Zod's _def is an internal property |
| 94 | + * not exposed in public type definitions |
| 95 | + */ |
| 96 | +export function getZodTypeDef(type: zod.ZodTypeAny): ZodSchemaDef | null { |
| 97 | + if (!isZodType(type)) { |
| 98 | + return null; |
| 99 | + } |
| 100 | + |
| 101 | + // eslint-disable-next-line @typescript-eslint/no-explicit-any |
| 102 | + return (type as any)._def as ZodSchemaDef; |
| 103 | +} |
| 104 | + |
| 105 | +/** |
| 106 | + * Build a Zod object schema from field definitions |
| 107 | + * Type-safe replacement for schema builder functions |
| 108 | + * |
| 109 | + * Note: Uses 'any' cast for zod.object() shape parameter because Zod's |
| 110 | + * ZodObject constructor expects Record<string, any> internally, even though |
| 111 | + * we're providing properly typed ZodTypeAny values. This is a bridge between |
| 112 | + * Zod's internal type system and our type-safe wrapper. |
| 113 | + */ |
| 114 | +export function buildZodObjectSchema( |
| 115 | + fields: Record<string, SchemaFieldDef> |
| 116 | +): ZodObjectSchema { |
| 117 | + const shape: Record<string, zod.ZodTypeAny> = {}; |
| 118 | + |
| 119 | + for (const [fieldName, fieldDef] of Object.entries(fields)) { |
| 120 | + let fieldSchema: zod.ZodTypeAny; |
| 121 | + |
| 122 | + // Build base schema based on type |
| 123 | + switch (fieldDef.type) { |
| 124 | + case 'string': |
| 125 | + fieldSchema = zod.string(); |
| 126 | + break; |
| 127 | + case 'number': |
| 128 | + fieldSchema = zod.number(); |
| 129 | + break; |
| 130 | + case 'boolean': |
| 131 | + fieldSchema = zod.boolean(); |
| 132 | + break; |
| 133 | + case 'string[]': |
| 134 | + fieldSchema = zod.array(zod.string()); |
| 135 | + break; |
| 136 | + case 'number[]': |
| 137 | + fieldSchema = zod.array(zod.number()); |
| 138 | + break; |
| 139 | + case 'boolean[]': |
| 140 | + fieldSchema = zod.array(zod.boolean()); |
| 141 | + break; |
| 142 | + case 'date': |
| 143 | + // Store dates as ISO strings |
| 144 | + fieldSchema = zod.string(); |
| 145 | + break; |
| 146 | + default: |
| 147 | + // Default to string for unknown types |
| 148 | + fieldSchema = zod.string(); |
| 149 | + } |
| 150 | + |
| 151 | + // Make optional if specified |
| 152 | + if (fieldDef.optional) { |
| 153 | + fieldSchema = fieldSchema.optional(); |
| 154 | + } |
| 155 | + |
| 156 | + shape[fieldName] = fieldSchema; |
| 157 | + } |
| 158 | + |
| 159 | + // eslint-disable-next-line @typescript-eslint/no-explicit-any |
| 160 | + return zod.object(shape as any); |
| 161 | +} |
| 162 | + |
| 163 | +/** |
| 164 | + * Parse field type string into SchemaFieldDef |
| 165 | + * Handles suffixes: ?, [], etc. |
| 166 | + * |
| 167 | + * @example |
| 168 | + * parseFieldType('string') => { type: 'string', optional: false } |
| 169 | + * parseFieldType('string?') => { type: 'string', optional: true } |
| 170 | + * parseFieldType('number[]') => { type: 'number[]', optional: false } |
| 171 | + * parseFieldType('boolean[]?') => { type: 'boolean[]', optional: true } |
| 172 | + */ |
| 173 | +export function parseFieldType(typeString: string): SchemaFieldDef { |
| 174 | + const isOptional = typeString.endsWith('?'); |
| 175 | + const baseType = isOptional ? typeString.slice(0, -1) : typeString; |
| 176 | + |
| 177 | + // Validate known types |
| 178 | + const knownTypes = [ |
| 179 | + 'string', |
| 180 | + 'number', |
| 181 | + 'boolean', |
| 182 | + 'string[]', |
| 183 | + 'number[]', |
| 184 | + 'boolean[]', |
| 185 | + 'date', |
| 186 | + ]; |
| 187 | + const type = ( |
| 188 | + knownTypes.includes(baseType) ? baseType : 'unknown' |
| 189 | + ) as SchemaFieldDef['type']; |
| 190 | + |
| 191 | + return { |
| 192 | + type, |
| 193 | + optional: isOptional, |
| 194 | + }; |
| 195 | +} |
| 196 | + |
| 197 | +/** |
| 198 | + * Extract schema description from Zod schema |
| 199 | + * Converts internal Zod structure to human-readable format for LLMs |
| 200 | + * |
| 201 | + * Note: Accesses _def property which is internal to Zod, but is the only way |
| 202 | + * to introspect schema structure for LLM-friendly descriptions |
| 203 | + */ |
| 204 | +export function describeZodSchema(schema: ZodObjectSchema): string { |
| 205 | + // eslint-disable-next-line @typescript-eslint/no-explicit-any |
| 206 | + const shape = (schema as any)?._def?.shape; |
| 207 | + if (!shape) { |
| 208 | + return '(Unable to describe schema)'; |
| 209 | + } |
| 210 | + |
| 211 | + const fields: string[] = []; |
| 212 | + |
| 213 | + for (const [fieldName, fieldSchema] of Object.entries(shape)) { |
| 214 | + const def = getZodTypeDef(fieldSchema as zod.ZodTypeAny); |
| 215 | + if (!def) { |
| 216 | + fields.push(` - ${fieldName}: (unknown type)`); |
| 217 | + continue; |
| 218 | + } |
| 219 | + |
| 220 | + const typeString = def.type || 'unknown'; |
| 221 | + let typeDesc = typeString; |
| 222 | + |
| 223 | + // Handle optional types |
| 224 | + if (typeString === 'optional') { |
| 225 | + const innerDef = |
| 226 | + // eslint-disable-next-line @typescript-eslint/no-explicit-any |
| 227 | + def.innerType ? getZodTypeDef(def.innerType as any) : null; |
| 228 | + const innerType = innerDef?.type || 'unknown'; |
| 229 | + typeDesc = `optional ${innerType}`; |
| 230 | + } |
| 231 | + // Handle array types |
| 232 | + else if (typeString === 'array') { |
| 233 | + const elementDef = |
| 234 | + // eslint-disable-next-line @typescript-eslint/no-explicit-any |
| 235 | + def.element ? getZodTypeDef(def.element as any) : null; |
| 236 | + const elementType = elementDef?.type || 'unknown'; |
| 237 | + typeDesc = `array of ${elementType}s`; |
| 238 | + } |
| 239 | + |
| 240 | + fields.push(` - ${fieldName}: ${typeDesc}`); |
| 241 | + } |
| 242 | + |
| 243 | + return fields.length > 0 |
| 244 | + ? fields.join('\n') |
| 245 | + : '(No fields in schema)'; |
| 246 | +} |
| 247 | + |
| 248 | +/** |
| 249 | + * Configuration Validation Schemas |
| 250 | + * |
| 251 | + * Provides Zod schemas for runtime validation of configuration objects. |
| 252 | + * These schemas ensure configuration is valid at entry points before use. |
| 253 | + */ |
| 254 | + |
| 255 | +/** |
| 256 | + * ExplorationConfig Schema - Validates autonomous exploration configuration |
| 257 | + * |
| 258 | + * Ensures all configuration values are within valid ranges and correct types. |
| 259 | + * Applied at exploration initialization time. |
| 260 | + */ |
| 261 | +export const ExplorationConfigSchema = zod.object({ |
| 262 | + maxDepth: zod |
| 263 | + .number() |
| 264 | + .int() |
| 265 | + .min(1, 'maxDepth must be at least 1') |
| 266 | + .max(20, 'maxDepth must be at most 20') |
| 267 | + .default(3), |
| 268 | + maxPages: zod |
| 269 | + .number() |
| 270 | + .int() |
| 271 | + .min(1, 'maxPages must be at least 1') |
| 272 | + .max(1000, 'maxPages must be at most 1000') |
| 273 | + .default(50), |
| 274 | + followExternal: zod.boolean().default(false), |
| 275 | + ignorePatterns: zod |
| 276 | + .array(zod.string()) |
| 277 | + .default([]) |
| 278 | + .refine( |
| 279 | + patterns => { |
| 280 | + // Validate that all patterns are valid regex strings |
| 281 | + return patterns.every(pattern => { |
| 282 | + try { |
| 283 | + new RegExp(pattern); |
| 284 | + return true; |
| 285 | + } catch { |
| 286 | + return false; |
| 287 | + } |
| 288 | + }); |
| 289 | + }, |
| 290 | + {message: 'All ignorePatterns must be valid regex strings'}, |
| 291 | + ), |
| 292 | + respectRobotsTxt: zod.boolean().default(true), |
| 293 | + captureScreenshots: zod.boolean().default(false), |
| 294 | + detectErrors: zod.boolean().default(true), |
| 295 | + timeout: zod |
| 296 | + .number() |
| 297 | + .int() |
| 298 | + .min(1000, 'timeout must be at least 1000ms') |
| 299 | + .max(300000, 'timeout must be at most 300000ms') |
| 300 | + .default(30000), |
| 301 | +}); |
| 302 | + |
| 303 | +export type ValidatedExplorationConfig = zod.infer< |
| 304 | + typeof ExplorationConfigSchema |
| 305 | +>; |
| 306 | + |
| 307 | +/** |
| 308 | + * Request Parameters Schema - Validates MCP tool request parameters |
| 309 | + * |
| 310 | + * Ensures tool requests have valid structure before processing. |
| 311 | + */ |
| 312 | +export const McpToolRequestSchema = zod.object({ |
| 313 | + method: zod |
| 314 | + .string() |
| 315 | + .min(1, 'method is required') |
| 316 | + .regex(/^[a-zA-Z_][a-zA-Z0-9_]*$/, 'method must be a valid identifier'), |
| 317 | + params: zod.record(zod.string(), zod.unknown()).optional(), |
| 318 | + timeout: zod |
| 319 | + .number() |
| 320 | + .int() |
| 321 | + .min(100, 'timeout must be at least 100ms') |
| 322 | + .max(300000, 'timeout must be at most 300000ms') |
| 323 | + .optional(), |
| 324 | +}); |
| 325 | + |
| 326 | +export type ValidatedMcpToolRequest = zod.infer<typeof McpToolRequestSchema>; |
0 commit comments