@nockdev/awf 6.2.0

package/.agent/core/ROUTER.yaml

@@ -0,0 +1,424 @@
+# Semantic Router Configuration
+# DOMYH Awesome Code
+# Selects top-K relevant skills AND routes to appropriate personas
+
+version: "6.1.2"
+
+# Routing Strategy
+strategy:
+  mode: semantic # semantic | pattern | hybrid
+  fallback: pattern # Use pattern matching if semantic fails
+
+# Embedding Configuration
+embedding:
+  model: hsa # hsa = HSA BM25+Vector hybrid, local = TF-IDF fallback
+  dimensions: 384
+  cache_enabled: true
+  cache_ttl: 3600
+
+  # HSA Integration (v1.0.0)
+  hsa:
+    enabled: true
+    prefer_hybrid: true # Use 70% vector + 30% BM25
+    fallback_to_local: true # Fallback to TF-IDF if HSA unavailable
+    engine_path: ".agent/scripts/hsa"
+
+# =============================================================================
+# P0: DYNAMIC THRESHOLD (v1.1)
+# =============================================================================
+# Research shows 0.30 is too low, industry best practice is 0.50-0.85
+# Source: Milvus, GitHub Copilot embeddings research
+
+selection:
+  top_k: 5 # Max skills to select
+  boost_active: 0.15 # Boost recently used skills
+
+  # Dynamic threshold strategy (replaces fixed min_score)
+  threshold:
+    strategy: "adaptive" # fixed | adaptive | percentile
+
+    # Threshold levels (higher = more precise, lower = more recall)
+    levels:
+      high_confidence: 0.70 # Definite match - use directly
+      medium_confidence: 0.50 # Likely match - include in results
+      low_confidence: 0.35 # Fallback only - require file pattern
+
+    # Default threshold for routing
+    default: 0.50
+
+    # Auto-adjustment rules
+    auto_adjust:
+      on_few_results: # If < 2 results above threshold
+        action: "lower"
+        delta: -0.10
+        min: 0.35
+      on_many_results: # If > 5 results above threshold
+        action: "raise"
+        delta: +0.10
+        max: 0.75
+      on_exact_match: # Exact keyword/file match
+        action: "boost"
+        boost: +0.20
+
+# Scoring Weights for Multi-Component Embedding
+weights:
+  description: 0.50 # Primary: skill description
+  keywords: 0.25 # Secondary: trigger keywords
+  examples: 0.15 # Usage examples in skill
+  detect_patterns: 0.10 # File detection patterns
+
+# Skill Priority Boost (added to similarity score)
+priority_boost:
+  core: 0.20 # Security, testing
+  language: 0.10 # Go, Python, etc.
+  framework: 0.05 # React, Vue, etc.
+  devops: 0.00 # Docker, K8s, etc.
+
+# Query Preprocessing
+preprocessing:
+  lowercase: true
+  remove_stopwords: true
+  stem: false # Disable for accuracy
+
+# =============================================================================
+# P1: QUERY EXPANSION
+# =============================================================================
+# Improves recall by expanding queries with synonyms and related terms
+# Source: GitHub Copilot RAG, Cursor semantic search
+
+query_expansion:
+  enabled: true
+  strategies:
+    # Synonym expansion for common abbreviations
+    synonyms:
+      # Languages
+      js: ["javascript", "nodejs", "ecmascript", "es6", "es2024"]
+      ts: ["typescript", "tsx"]
+      py: ["python", "python3"]
+      rb: ["ruby"]
+      rs: ["rust"]
+
+      # Groups (expand to multiple skills)
+      frontend: ["react", "vue", "angular", "svelte", "nextjs", "nuxt"]
+      backend: ["go", "python", "java", "rust", "csharp", "kotlin"]
+      mobile: ["react-native", "flutter", "swift", "kotlin"]
+      infra: ["docker", "kubernetes", "aws", "ci-cd"]
+
+      # Common phrases
+      api: ["rest", "graphql", "fastapi", "express"]
+      ui: ["react", "vue", "svelte", "tailwind", "ui-ux-pro-max"]
+      test: ["testing", "jest", "pytest", "vitest"]
+      db: ["database", "sql", "postgres", "mysql"]
+
+    # Context-based expansion (use signals from IDE)
+    context_expansion:
+      use_file_context: true # If .tsx open, boost React + TypeScript
+      use_recent_skills: true # Boost recently used skills
+      use_project_config: true # Detect from package.json, go.mod, etc.
+
+    # Expansion limits
+    max_expansions: 5 # Max synonyms to add per query
+    expansion_boost: 0.08 # Score boost for expanded matches
+
+# Caching Configuration
+cache:
+  query_cache: true # Cache query embeddings
+  skill_cache: true # Cache skill embeddings
+  cache_size: 1000 # Max cached queries
+  ttl: 3600 # 1 hour TTL
+
+# Fallback Pattern Matching
+patterns:
+  enabled: true # Enable as fallback
+  file_based: true # Use file detection
+  keyword_based: true # Use trigger keywords
+
+logging:
+  enabled: false # Disable for production
+  log_selections: false
+  log_scores: false
+
+# =============================================================================
+# P2: CONFIDENCE SCORING
+# =============================================================================
+# Returns confidence level and evidence for each skill match
+# Enables debugging and user trust
+
+confidence_scoring:
+  enabled: true
+
+  # Response includes these fields for each matched skill
+  response_format:
+    include_confidence: true # 0.0 - 1.0 score
+    include_match_type: true # How skill was matched
+    include_evidence: true # Why skill was matched
+    include_tier: true # high | medium | low confidence
+
+  # Match types (how skill was selected)
+  match_types:
+    exact_keyword: 0.95 # Query contains exact skill keyword
+    file_pattern: 0.90 # File extension matches (.py → python)
+    semantic: 0.80 # Semantic similarity match
+    synonym: 0.75 # Via query expansion synonym
+    related: 0.60 # Related skill (python → docker)
+    fallback: 0.40 # Low confidence fallback
+
+  # Evidence messages (auto-generated)
+  evidence_templates:
+    exact_keyword: "Query contains keyword '{keyword}'"
+    file_pattern: "Detected files: {patterns}"
+    semantic: "Semantic match: {score:.0%} similarity"
+    synonym: "Matched via synonym '{original}' → '{expanded}'"
+    related: "Related to {parent_skill} via {relationship}"
+    fallback: "Fallback match, low confidence"
+
+  # Confidence tier mapping
+  tiers:
+    high: { min: 0.70, action: "use directly" }
+    medium: { min: 0.50, action: "include in suggestions" }
+    low: { min: 0.35, action: "show if no better match" }
+
+# =============================================================================
+# EXAMPLE RESPONSE FORMAT (for consumers of this router)
+# =============================================================================
+# skills:
+#   - id: "python"
+#     confidence: 0.92
+#     tier: "high"
+#     match_type: "exact_keyword"
+#     evidence: ["Query contains keyword 'fastapi'", "Detected files: *.py"]
+#   - id: "docker"
+#     confidence: 0.65
+#     tier: "medium"
+#     match_type: "related"
+#     evidence: ["Related to python via related_skills"]
+# =============================================================================
+
+# =============================================================================
+# AUTO-SUGGEST CONFIGURATION (v6.2.0)
+# =============================================================================
+
+auto_suggest:
+  enabled: true
+  trigger: session_start # session_start | on_demand | always
+  display_top: 3 # Show top 3 relevant skills
+  format: compact # compact | detailed
+  show_reason: true # Show why skill was suggested
+
+# =============================================================================
+# CONTEXT-AWARE ROUTING (v6.2.0)
+# =============================================================================
+
+context_routing:
+  # Detect from currently open files
+  file_patterns: true
+
+  # Detect from recent git commits
+  git_history: true
+  git_history_depth: 5 # Last 5 commits
+
+  # Detect from error messages in terminal
+  error_context: true
+
+  # Detect from project structure
+  project_structure: true
+
+  # Weight for context signals
+  context_weights:
+    open_files: 0.40
+    git_history: 0.25
+    error_context: 0.20
+    project_structure: 0.15
+
+# =============================================================================
+# WORKFLOW ROUTING (v6.2.0)
+# =============================================================================
+
+workflow_routing:
+  # Route to appropriate workflow based on intent
+  enabled: true
+
+  # Intent → Workflow mapping
+  intents:
+    fix_project: modify
+    new_project: init
+    debug_error: debug
+    write_code: code
+    review_code: review
+    run_tests: test
+    deploy_app: deploy
+
+# =============================================================================
+# LIBRARY REGISTRY INTEGRATION ()
+# =============================================================================
+
+library_registry:
+  enabled: true
+  source: "core/LIBRARY_REGISTRY.yaml"
+  lazy_load: true # Only load relevant language
+
+  # Auto-suggest libraries based on detected intent
+  auto_suggest_libs:
+    on_new_file: true
+    on_import_error: true
+    on_feature_request: true
+
+  # Library selection criteria
+  selection:
+    prefer: tier_1
+    max_suggestions: 3
+    show_install_command: true
+
+# =============================================================================
+# INTENT DETECTION INTEGRATION ()
+# =============================================================================
+
+intent_detection:
+  enabled: true
+  source: "core/INTENT_DETECTION.yaml"
+
+  # Ask clarifying questions when confidence < threshold
+  clarify:
+    enabled: true
+    threshold: 0.7
+    max_questions: 2
+
+  # Apply coding style based on language
+  coding_style:
+    source: "core/CODING_STYLES.yaml"
+    apply_to_generation: true
+    check_in_review: true
+
+# =============================================================================
+# SMART RECOMMENDATION ENGINE ()
+# =============================================================================
+
+recommendations:
+  # When to trigger recommendations
+  triggers:
+    - new_project
+    - new_file
+    - feature_request
+    - error_context
+
+  # Recommendation format
+  format: |
+    📦 **Recommended**: {library}
+    💡 **Why**: {reason}
+    📋 **Install**: `{install_cmd}`
+
+  # Combine multiple signals
+  signal_weights:
+    user_request: 0.40
+    file_context: 0.25
+    project_type: 0.20
+    recent_activity: 0.15
+
+# _DOMYH Awesome Code
+
+# =============================================================================
+# PERSONA ROUTING (v6.0)
+# =============================================================================
+
+persona_routing:
+  enabled: true
+
+  # Default persona for general tasks
+  default: developer
+
+  # Command-based routing (explicit triggers)
+  command_triggers:
+    "/code": developer
+    "/refactor": developer
+    "/plan": planner
+    "/design": architect
+    "/debug": debugger
+    "/investigate": debugger
+    "/test": tester
+    "/coverage": tester
+    "/deploy": devops
+    "/monitor": devops
+    "/env": devops
+    "/infra": devops
+    "/doc": documenter
+    "/readme": documenter
+    "/ap": auditor
+    "/audit": auditor
+    "/security": security
+    "/scan": security
+    "/research": researcher
+    "/find": researcher
+
+  # Intent-based routing (inferred from request)
+  intent_triggers:
+    implement_feature: developer
+    fix_bug: debugger
+    review_code: auditor
+    design_architecture: architect
+    write_tests: tester
+    deploy_application: devops
+    document_code: documenter
+    security_check: security
+    research_topic: researcher
+    complex_multi_step: orchestrator
+
+  # Persona collaboration rules
+  collaboration:
+    orchestrator:
+      can_delegate: ["planner", "developer", "tester", "researcher"]
+      auto_activate_for:
+        - "tasks requiring multiple personas"
+        - "complex multi-step tasks"
+        - "tasks with unclear owner"
+
+    developer:
+      can_delegate: ["tester", "debugger"]
+      reports_to: ["architect"]
+
+    architect:
+      supervises: ["developer", "devops"]
+      consulted_by: ["*"]
+
+    debugger:
+      can_delegate: ["tester"]
+      reports_to: ["developer"]
+
+    security:
+      can_delegate: []
+      consulted_by: ["auditor", "developer"]
+
+  # Handoff conditions (when to switch personas)
+  handoff:
+    developer_to_tester: "code_complete"
+    debugger_to_developer: "root_cause_found"
+    planner_to_developer: "plan_approved"
+    auditor_to_developer: "findings_reported"
+    security_to_developer: "vulnerabilities_found"
+
+# =============================================================================
+# CONSTITUTIONAL RULES LOADING (v6.0)
+# =============================================================================
+
+constitutional:
+  enabled: true
+  tiers:
+    - source: "rules/constitutional/tier-0-core.yaml"
+      always_load: true
+    - source: "rules/constitutional/tier-1-safety.yaml"
+      always_load: true
+    - source: "rules/constitutional/tier-2-execution.yaml"
+      always_load: false # Load based on persona
+
+  modules:
+    reflection:
+      source: "rules/modules/reflection.yaml"
+      for_personas: ["*"]
+    context_management:
+      source: "rules/modules/context-management.yaml"
+      for_personas: ["*"]
+    evidence:
+      source: "rules/modules/evidence.yaml"
+      for_personas: ["auditor", "debugger", "security", "researcher"]
+    stop_conditions:
+      source: "rules/modules/stop-conditions.yaml"
+      for_personas: ["*"]

package/.agent/core/SCORING_FORMULA.yaml

@@ -0,0 +1,103 @@
+# DOMYH Awesome Code — Scoring Formula v6.2.0
+# Single Source of Truth for context priority scoring
+# Referenced by: CONTEXT_INJECTOR, CONTEXT_PRIORITY, ACTIVE_MEMORY
+
+version: "6.1.2"
+engine: scoring_formula
+
+# =============================================================================
+# CORE WEIGHTS (Single Definition)
+# =============================================================================
+
+weights:
+  recency: 0.25       # Newer content = higher priority
+  relevance: 0.40     # Semantic match with current task (highest weight)
+  importance: 0.25    # Critical > High > Medium > Low
+  frequency: 0.10     # Access count (capped at 10)
+
+# =============================================================================
+# MASTER FORMULA (Single Definition)
+# =============================================================================
+
+formula:
+  calculation: |
+    final_score = (recency * 0.25) + (relevance * 0.40) + 
+                  (importance * 0.25) + (frequency * 0.10)
+  
+  # Result range: 0.0 - 1.0
+
+# =============================================================================
+# THRESHOLDS (Single Definition)
+# =============================================================================
+
+thresholds:
+  excellent:    # 0.85 - 1.0
+    min: 0.85
+    action: always_include
+    token_budget: full
+  
+  good:         # 0.70 - 0.85
+    min: 0.70
+    action: include_if_budget
+    token_budget: full
+  
+  moderate:     # 0.50 - 0.70
+    min: 0.50
+    action: include_summarized
+    token_budget: 50%
+  
+  low:          # 0.30 - 0.50
+    min: 0.30
+    action: pointer_only
+    token_budget: minimal
+  
+  skip:         # 0.0 - 0.30
+    min: 0.0
+    action: do_not_include
+    token_budget: 0
+
+# =============================================================================
+# RECENCY SCORING
+# =============================================================================
+
+recency_scoring:
+  fresh:        { age_seconds: 3600,    score: 1.0 }  # < 1 hour
+  recent:       { age_seconds: 86400,   score: 0.8 }  # 1-24 hours
+  this_week:    { age_seconds: 604800,  score: 0.6 }  # 1-7 days
+  this_month:   { age_seconds: 2592000, score: 0.4 }  # 7-30 days
+  older:        { score: 0.2 }                        # > 30 days
+
+# =============================================================================
+# IMPORTANCE LEVELS
+# =============================================================================
+
+importance_levels:
+  critical: { score: 1.0, examples: [active_errors, security_vulns, user_corrections] }
+  high:     { score: 0.8, examples: [recent_decisions, current_context, active_warnings] }
+  medium:   { score: 0.6, examples: [project_state, recent_workflows, audit_findings] }
+  low:      { score: 0.4, examples: [historical_decisions, old_sessions, resolved_errors] }
+  minimal:  { score: 0.2, examples: [raw_logs, verbose_output, old_drafts] }
+
+# =============================================================================
+# FREQUENCY SCORING
+# =============================================================================
+
+frequency_scoring:
+  heavily_used:     { min_accesses: 20, score: 1.0 }
+  frequently_used:  { min_accesses: 10, score: 0.8 }
+  moderately_used:  { min_accesses: 5,  score: 0.6 }
+  occasionally_used:{ min_accesses: 2,  score: 0.4 }
+  rarely_used:      { min_accesses: 1,  score: 0.2 }
+  never_used:       { min_accesses: 0,  score: 0.1 }
+
+# =============================================================================
+# USAGE (How other files reference this)
+# =============================================================================
+
+usage:
+  reference: "core/SCORING_FORMULA.yaml"
+  import_pattern: |
+    # In CONTEXT_INJECTOR.yaml, CONTEXT_PRIORITY.yaml, ACTIVE_MEMORY.yaml:
+    scoring:
+      reference: core/SCORING_FORMULA.yaml
+      # Uses weights, formula, thresholds from this file

package/.agent/core/SEMANTIC_ENGINE.yaml

@@ -0,0 +1,162 @@
+# =============================================================================
+# SEMANTIC_ENGINE.yaml — HSA Semantic Memory Configuration
+# =============================================================================
+# DOMYH Awesome Code Agent — Vector-based Semantic Search
+# Version: 6.1.2
+# =============================================================================
+
+# Enable/Disable semantic memory
+config:
+  enabled: true # Enable semantic memory
+  auto_save: true # Auto-save on workflow completion
+  auto_cleanup: true # Auto-cleanup expired items
+
+# =============================================================================
+# LanceDB Configuration
+# =============================================================================
+
+database:
+  path: ".agent/memory/vectors/domyh_memory.lance"
+  max_size_mb: 500 # Maximum database size
+
+# =============================================================================
+# Embedding Configuration
+# =============================================================================
+
+embedding:
+  # Provider: "openai" | "local" | "auto"
+  # - openai: Uses OpenAI API (requires OPENAI_API_KEY)
+  # - local: Uses sentence-transformers (offline, slower)
+  # - auto: Tries openai first, falls back to local
+  provider: "auto"
+
+  # Model configurations
+  openai:
+    model: "text-embedding-3-small"
+    dimension: 1536
+    batch_size: 100
+
+  local:
+    model: "all-MiniLM-L6-v2" # Small, fast model
+    # Alternatives:
+    # - all-mpnet-base-v2 (better quality, slower)
+    # - paraphrase-multilingual-MiniLM-L12-v2 (multilingual)
+    dimension: 384
+
+# =============================================================================
+# Collection Configuration
+# =============================================================================
+
+collections:
+  audits:
+    description: "Full audit reports from /ap workflow"
+    retention_days: 365
+    min_keep: 10
+    auto_index_on:
+      - "/ap complete"
+      - "audit saved"
+
+  decisions:
+    description: "Architectural decisions log"
+    retention_days: 730 # 2 years
+    min_keep: 50
+    auto_index_on:
+      - "decision made"
+      - "architecture update"
+
+  errors:
+    description: "Error patterns and solutions"
+    retention_days: 180 # 6 months
+    min_keep: 20
+    auto_index_on:
+      - "error resolved"
+      - "bug fixed"
+
+  code_reviews:
+    description: "Code review findings"
+    retention_days: 365
+    min_keep: 15
+    auto_index_on:
+      - "/review complete"
+      - "PR reviewed"
+
+# =============================================================================
+# Search Configuration
+# =============================================================================
+
+search:
+  default_k: 5 # Default number of results
+  max_k: 20 # Maximum results allowed
+  min_score: 0.3 # Minimum relevance score (0-1)
+
+  # Token budget for search results
+  token_budget:
+    per_result: 500
+    total: 2500
+
+# =============================================================================
+# Caching
+# =============================================================================
+
+cache:
+  enabled: true
+  ttl_seconds: 300 # 5 minutes
+  max_entries: 100
+
+# =============================================================================
+# Fallback Behavior
+# =============================================================================
+
+fallback:
+  # When semantic search unavailable
+  strategy: "keyword" # "keyword" | "file_based" | "none"
+
+  # Silent fallback (no user warning)
+  silent: true
+
+# =============================================================================
+# Integration Triggers
+# =============================================================================
+
+triggers:
+  # Commands that trigger memory operations
+  save_triggers:
+    - workflow: "/ap"
+      collection: "audits"
+      content: "audit_summary"
+
+    - workflow: "/review"
+      collection: "code_reviews"
+      content: "review_findings"
+
+    - workflow: "/debug"
+      collection: "errors"
+      content: "error_resolution"
+
+  # Search triggers
+  search_triggers:
+    - command: "/search"
+      collection: "all"
+
+    - command: "/find-similar"
+      collection: "all"
+
+    - command: "/recall"
+      collection: "decisions"
+
+# =============================================================================
+# Metrics
+# =============================================================================
+
+metrics:
+  enabled: true
+  track:
+    - search_latency
+    - hit_rate
+    - index_size
+    - embedding_time
+
+# =============================================================================
+# Footer
+# =============================================================================
+# _DOMYH Awesome Code • Semantic Engine v6.2.0 • HSA Integration_