prjct-cli 1.22.0 → 1.24.0

package/core/agentic/index.ts

@@ -1,193 +0,0 @@
-/**
- * Agentic Module - Central exports for agentic capabilities
- *
- * Organized into logical groups:
- * - Memory: Pattern learning, semantic memories, history
- * - Planning: Plan mode, approval flows
- * - Execution: Command executor, loop detection
- * - Context: Context building, smart context, prompt generation
- * - Routing: Agent router, ground truth verification
- * - Tools: Tool registry, template loader
- */
-
-// ============ Types ============
-// All types re-exported from ../types (canonical source)
-export type {
-  // Agent router types
-  Agent,
-  AgentInfo,
-  ApprovalContext,
-  ApprovalOperation,
-  ApprovalOption,
-  ApprovalPrompt,
-  AssignmentContext,
-  AttemptInfo,
-  AttemptRecord,
-  AttemptResult,
-  // Chain of Thought types
-  ChainOfThoughtContext,
-  ChainOfThoughtResult,
-  ChainOfThoughtState,
-  ChangedFile,
-  // Smart context types
-  ContextDomain,
-  // Context builder types
-  ContextPaths,
-  ContextState,
-  Decision,
-  DomainAnalysis,
-  // Loop Detector types
-  ErrorEntry,
-  ErrorPattern,
-  EscalationInfo,
-  // Execution types
-  ExecutionResult,
-  ExecutionToolsFn,
-  FeatureInfo,
-  FilteredContext,
-  FilterMetrics,
-  // Prompt builder types
-  Frontmatter,
-  FullContext,
-  GatheredAnalysisData,
-  GatheredFileData,
-  GatheredInfo,
-  GatheredInfoType,
-  // Ground Truth types
-  GroundTruthContext,
-  HallucinationPattern,
-  HallucinationResult,
-  HistoryEntry,
-  HistoryEventType,
-  LearnedPatterns,
-  LoadedAgent,
-  LoadedSkill,
-  // Memory types
-  Memory,
-  MemoryContext,
-  MemoryContextParams,
-  MemoryDatabase,
-  MemoryTag,
-  // Orchestrator types
-  OrchestratorContext,
-  OrchestratorSubtask,
-  OutputAnalysis,
-  PatternInfo,
-  Patterns,
-  Plan,
-  PlanAnalysis,
-  PlanInfo,
-  // Plan Mode types
-  PlanParams,
-  PlanStatus,
-  PlanStep,
-  PlanStepDefinition,
-  PlanStepResult,
-  Preference,
-  ProjectContext,
-  ProposedPlan,
-  ReasoningResult,
-  ReasoningStep,
-  SimpleExecutionResult,
-  StackInfo,
-  Template,
-  ThinkBlock,
-  // Tool registry types
-  ToolFunction,
-  ToolRegistryInterface,
-  VerificationResult,
-  Verifier,
-  Workflow,
-} from '../types'
-// ============ Routing ============
-// Agent routing and ground truth verification
-export { default as AgentRouter } from './agent-router'
-// ============ Utilities ============
-// Chain of thought, services
-export {
-  default as chainOfThought,
-  formatPlan,
-  REASONING_REQUIRED_COMMANDS,
-  reason,
-  requiresReasoning,
-} from './chain-of-thought'
-// ============ Execution ============
-// Command execution, loop detection
-export {
-  CommandExecutor,
-  default as commandExecutor,
-  signalEnd,
-  signalStart,
-} from './command-executor'
-
-// ============ Context ============
-// Context building and prompt generation
-export { default as contextBuilder } from './context-builder'
-export { DomainClassifier, default as domainClassifier } from './domain-classifier'
-export {
-  default as groundTruth,
-  escapeRegex,
-  formatDuration,
-  formatWarnings,
-  prepareCommand,
-  requiresVerification,
-  verifiers,
-  verify,
-  verifyAnalyze,
-  verifyDone,
-  verifyFeature,
-  verifyInit,
-  verifyNow,
-  verifyShip,
-  verifySpec,
-  verifySync,
-} from './ground-truth'
-export {
-  analyzeErrorPattern,
-  default as loopDetector,
-  detectHallucination,
-  generateEscalationMessage,
-  generateSuggestion,
-  getHallucinationSuggestion,
-  HALLUCINATION_PATTERNS,
-  isSimilarError,
-  LoopDetector,
-} from './loop-detector'
-// ============ Memory ============
-// Pattern learning, semantic memories, session history
-export {
-  CachedStore,
-  default as memorySystem,
-  HistoryStore,
-  MEMORY_TAGS,
-  MemorySystem,
-  PatternStore,
-  SemanticMemories,
-  SessionStore,
-} from './memory-system'
-export { default as orchestratorExecutor, OrchestratorExecutor } from './orchestrator-executor'
-// ============ Planning ============
-// Plan mode for complex tasks
-export {
-  DESTRUCTIVE_COMMANDS,
-  default as planMode,
-  generateApprovalPrompt,
-  PLAN_REQUIRED_COMMANDS,
-  PLAN_STATUS,
-  PLANNING_TOOLS,
-  PlanMode,
-} from './plan-mode'
-export { default as promptBuilder } from './prompt-builder'
-export { default as services } from './services'
-export { default as smartContext } from './smart-context'
-export { default as templateExecutor, TemplateExecutor } from './template-executor'
-export {
-  clearCache,
-  default as templateLoader,
-  getAllowedTools,
-  load,
-  parseFrontmatter,
-} from './template-loader'
-// ============ Tools ============
-// Tool and template management
-export { default as toolRegistry } from './tool-registry'

package/core/agentic/injection-validator.ts

@@ -1,208 +0,0 @@
-/**
- * Injection Validator
- *
- * Validates data before auto-injection into LLM prompts.
- * Corrupted or oversized data gets safe fallbacks instead of broken context.
- *
- * @module agentic/injection-validator
- */
-
-import type { z } from 'zod'
-import type { TokenBudgetCoordinator } from './token-budget'
-
-// =============================================================================
-// Token Budget Configuration
-// =============================================================================
-
-/** Configurable token budgets per injection section */
-export interface InjectionBudgets {
-  /** Auto-injected context (task + queue + patterns) */
-  autoContext: number
-  /** Per-agent content in orchestrator */
-  agentContent: number
-  /** Per-skill content in orchestrator */
-  skillContent: number
-  /** State data section */
-  stateData: number
-  /** Memories section */
-  memories: number
-  /** Total prompt ceiling (all sections combined) */
-  totalPrompt: number
-}
-
-/** Default budgets (in estimated tokens, ~4 chars per token) */
-export const DEFAULT_BUDGETS: InjectionBudgets = {
-  autoContext: 500,
-  agentContent: 400,
-  skillContent: 500,
-  stateData: 1000,
-  memories: 600,
-  totalPrompt: 8000,
-}
-
-/**
- * Create injection budgets from a TokenBudgetCoordinator.
- * Uses the coordinator's injection allocation as the totalPrompt ceiling,
- * keeping per-section budgets at their defaults.
- *
- * @see PRJ-266
- */
-export function budgetsFromCoordinator(coordinator: TokenBudgetCoordinator): InjectionBudgets {
-  const injectionBudget = coordinator.getAllocationFor('injection')
-  return {
-    ...DEFAULT_BUDGETS,
-    totalPrompt: injectionBudget,
-  }
-}
-
-// Approximate chars-per-token ratio
-const CHARS_PER_TOKEN = 4
-
-// =============================================================================
-// Safe Injection
-// =============================================================================
-
-/**
- * Validate data against a Zod schema before injection.
- * Returns validated data on success, or the fallback on failure.
- */
-export function safeInject<T>(data: unknown, schema: z.ZodType<T>, fallback: T): T {
-  const result = schema.safeParse(data)
-  if (result.success) {
-    return result.data
-  }
-  return fallback
-}
-
-/**
- * Validate and stringify data for prompt injection.
- * Returns formatted string on success, or fallback string on failure.
- */
-export function safeInjectString<T>(
-  data: unknown,
-  schema: z.ZodType<T>,
-  formatter: (valid: T) => string,
-  fallbackString: string
-): string {
-  const result = schema.safeParse(data)
-  if (result.success) {
-    return formatter(result.data)
-  }
-  return fallbackString
-}
-
-// =============================================================================
-// Token-Aware Truncation
-// =============================================================================
-
-/**
- * Truncate text to fit within a token budget.
- * Uses char-based estimation (~4 chars/token).
- */
-export function truncateToTokenBudget(text: string, maxTokens: number): string {
-  const maxChars = maxTokens * CHARS_PER_TOKEN
-  if (text.length <= maxChars) return text
-  return `${text.substring(0, maxChars)}\n... (truncated to ~${maxTokens} tokens)`
-}
-
-/**
- * Estimate token count for a string.
- */
-export function estimateTokens(text: string): number {
-  return Math.ceil(text.length / CHARS_PER_TOKEN)
-}
-
-// =============================================================================
-// Skill Filtering
-// =============================================================================
-
-/** Domain keywords for matching skills to task domains */
-const DOMAIN_KEYWORDS: Record<string, string[]> = {
-  frontend: ['react', 'vue', 'svelte', 'css', 'html', 'ui', 'component', 'frontend', 'web', 'dom'],
-  backend: ['api', 'server', 'backend', 'endpoint', 'route', 'middleware', 'database', 'sql'],
-  testing: ['test', 'spec', 'jest', 'vitest', 'cypress', 'playwright', 'coverage', 'assert'],
-  devops: ['docker', 'ci', 'cd', 'deploy', 'kubernetes', 'terraform', 'pipeline', 'github-actions'],
-  docs: ['documentation', 'readme', 'guide', 'tutorial', 'markdown'],
-  design: ['design', 'ux', 'ui', 'figma', 'wireframe', 'layout', 'accessibility'],
-}
-
-/**
- * Filter skills by relevance to detected task domains.
- * Returns only skills whose content matches the task's domains.
- */
-export function filterSkillsByDomains(
-  skills: { name: string; content: string }[],
-  detectedDomains: string[]
-): { name: string; content: string }[] {
-  if (detectedDomains.length === 0 || skills.length === 0) return skills
-
-  // Build keyword set from detected domains
-  const relevantKeywords = new Set<string>()
-  for (const domain of detectedDomains) {
-    const keywords = DOMAIN_KEYWORDS[domain.toLowerCase()]
-    if (keywords) {
-      for (const kw of keywords) relevantKeywords.add(kw)
-    }
-    // Also add the domain name itself
-    relevantKeywords.add(domain.toLowerCase())
-  }
-
-  return skills.filter((skill) => {
-    const text = `${skill.name} ${skill.content}`.toLowerCase()
-    // Keep skill if any relevant keyword appears in its name or content
-    for (const kw of relevantKeywords) {
-      if (text.includes(kw)) return true
-    }
-    return false
-  })
-}
-
-// =============================================================================
-// Section Budget Tracker
-// =============================================================================
-
-/**
- * Tracks cumulative token usage across injection sections.
- * Allows checking remaining budget before adding more content.
- */
-export class InjectionBudgetTracker {
-  private used = 0
-  private budgets: InjectionBudgets
-
-  constructor(budgets: Partial<InjectionBudgets> = {}) {
-    this.budgets = { ...DEFAULT_BUDGETS, ...budgets }
-  }
-
-  /** Add content and return it (possibly truncated to fit budget) */
-  addSection(content: string, sectionBudget: number): string {
-    const truncated = truncateToTokenBudget(content, sectionBudget)
-    const tokens = estimateTokens(truncated)
-
-    // Check total budget
-    if (this.used + tokens > this.budgets.totalPrompt) {
-      const remaining = this.budgets.totalPrompt - this.used
-      if (remaining <= 0) return ''
-      const fitted = truncateToTokenBudget(truncated, remaining)
-      this.used += estimateTokens(fitted)
-      return fitted
-    }
-
-    this.used += tokens
-    return truncated
-  }
-
-  /** Get remaining token budget */
-  get remaining(): number {
-    return Math.max(0, this.budgets.totalPrompt - this.used)
-  }
-
-  /** Get total tokens used */
-  get totalUsed(): number {
-    return this.used
-  }
-
-  /** Get the budgets config */
-  get config(): InjectionBudgets {
-    return this.budgets
-  }
-}