parrot Public
Commits
trunk
Commits on November 6, 2025
-
Tier 4 Intelligence: ML-like learning & dynamic generation
Implemented FULL Tier 4 with advanced learning and generation: PERSISTENT LEARNING SYSTEM: - Failure history tracked to ~/.parrot/failures.json - Records: command frequency, hourly patterns, branch disasters - Tracks: total failures, current streak, longest streak - Learns: worst commands, worst hours, worst branches - Thread-safe with mutex protection DYNAMIC TEMPLATE GENERATION: - Infinite insult combinations from components - Achievement-style: prefix + core + suffix patterns - Variable substitution: {command}, {streak}, {branch}, etc. - Command-specific templates for git, docker, npm, cargo, kubectl - Exit-code-aware dynamic generation FAILURE STREAK ESCALATION: - Escalates brutality with consecutive failures - Tier 1 (2-3 failures): Mild sarcasm - Tier 2 (4-6 failures): Increasing brutality - Tier 3 (7-9 failures): Harsh mockery - Tier 4 (10+ failures): Maximum savagery HISTORICAL PATTERN RECOGNITION: - Detects repeated command failures - Identifies time-based patterns - Recognizes branch disaster zones - Generates context-aware historical insults SAVAGE MODE (320 NEW INSULTS): - ego_destruction (40): Devastating competence attacks - career_advice (40): Brutal quit suggestions - team_impact (40): How you're ruining everyone - existential (40): Deep philosophical cuts - comparison (40): Unfavorable comparisons - timeline (40): Past/present/future failure analysis - meta (40): Self-aware commentary - technical_savage (40): Brutal tech-specific burns ACTIVATION LOGIC: - Savage mode: 5+ streak OR 50+ total failures - Learning persists across sessions - Gets smarter and more brutal over time Total insult database: 2,570+ (2,250 + 320 savage) Intelligence layers: 20 (4 Tier 4 + 5 Tier 3 + 2 Tier 2 + 7 Tier 1 + 2 original) Fallback system now rivals local LLM with: - Machine learning-like pattern recognition - Persistent memory across sessions - Dynamic generation (millions of combinations) - Adaptive brutality escalation - Historical context awareness Co-authored-by: espadonne <espadonne@outlook.com> Co-authored-by: mfwolffe <wolffemf@dukes.jmu.edu>Claude Assistant committed -
Tier 3 Intelligence: LLM-like context awareness
Implemented advanced Tier 3 intelligence features: - Repeated failure tracking (escalating insults) - CI/CD environment detection (GitHub, GitLab, Jenkins, Circle, Travis) - Error pattern recognition (8 exit code patterns) - Build system awareness (Docker, Makefile) - Dependency complexity detection Expanded database with 240+ new insults across 7 categories: - api (REST, GraphQL, OAuth, rate limiting) - database_ops (queries, transactions, migrations) - frontend (React, Vue, Angular, Webpack, TypeScript) - backend (servers, middleware, load balancing) - infrastructure (CDN, clusters, storage, networking) - debugging (stack traces, breakpoints, exceptions) - versioning (git operations, merges, releases) Total: 2,250+ insults with 16-layer intelligence priority Fallback system now rivals local LLM contextual awareness Co-authored-by: espadonne <espadonne@outlook.com> Co-authored-by: mfwolffe <wolffemf@dukes.jmu.edu>
Claude Assistant committed -
Tier 2 Intelligence: Project & git context awareness
Implemented Tier 2 smart fallback features: - Git branch detection (.git/HEAD parsing) - Project type detection (16 file patterns) - Priority system favoring contextual insults Expanded database with 445+ new insults across 13 categories: - testing, deployment, configuration, security, network, performance - cloud (AWS/GCP/Azure), devops (CI/CD), filesystem - package_management, compilation, monitoring Total: 1,600+ insults with deep context awareness Co-authored-by: espadonne <espadonne@outlook.com> Co-authored-by: mfwolffe <wolffemf@dukes.jmu.edu>
Claude Assistant committed