跳轉至

V14.0.0 - The Cognitive Revolution

Release Date: 2026-01-15 Codename: "Silent Sentinel" & "System 2"

Boring V14.0.0 marks the transition from an "AI Coding Tool" to a Cognitive Autonomous Agent. It introduces the 4-Layer Cognitive Architecture, enabling the agent to "Think before it Acts" using System 2 reasoning.

👑 Sovereign Edition (Phase VII)

The final iteration of V14.0 achieves 100-Point Perfection by introducing: - Project Biography (boring bio): Knowledge continuity for teams. - Policy-as-Code (boring policy): Enterprise-grade tool permissions. - Self-Healing Doctor (boring doctor --fix): Automated environment repair. - Project Sovereign Audit (boring perfection): Final certification of readiness. - Migration Engine (boring migrate): Forward-compatible project state transition. - V14.1 Scale & Optimization: - Incremental Reconciler: Milestone-based integrity checks. - Ledger Rotation: Hash-chained archiving for infinite scalability. - State Compression: Gzip-compressed snapshots for large projects. - Predictive Pre-warming: Background context fetching for near-zero latency.

🎯 Perfection Validation Roadmap

Stage Milestone Business Goal Gate
Stage 1 Beginner UX Zero-config setup verification. Newbie can finish flow with README.
Stage 2 Power User Stress test & fault tolerance. Clear fix guidance under pressure.
Stage 3 Team Reality Shadow Adoption metrics. Stickiness & behavior tracking.
Stage 4 Operation Readiness Logging & version stability. Observability for non-devs.
Stage 5 Governance Fit Policy-as-Code & audit trail. Behavior is restricted & auditable.
Stage 6 Vendor Risk Zero Lock-in verification. Project maintainable without Boring.
Stage 7 Executive Readiness Business KPI & ROI validation. Value understood by non-tech leaders.

🌟 Major Features

🧠 Cognitive Architecture (System 2)

  • Sequential Thinking: The agent now uses a structured thinking process (sequentialthinking) to analyze problems, revise assumptions, and plan strategies before writing a single line of code.
  • Semantic Tool Router: A new context-aware router (tool_router.py) that filters 60+ tools based on the current Flow Stage (Design, Implement, Polish), reducing token usage by 80%.

🐉 One Dragon Flow

  • End-to-End Autonomy: A unified flow that takes a single user intent and autonomously traverses the entire SDLC:
    1. Design: Architecture planning & Impact prediction.
    2. Implement: Code generation & test writing.
    3. Polish: Security scanning & Code review.
    4. Verify: Automated testing & validation.
  • Command: boring flow "Your intent here"

🔔 Notification System 2.0

  • Unified Manager: Centralized NotificationManager supporting multiple channels simultaneously.
  • Channels:
    • Desktop: Native OS toast notifications with sound.
    • Slack/Discord: Webhook integration for team updates.
    • Email: Critical alerts via SMTP.
  • UI: New "Glassmorphism" design for Web Dashboard and Toast notifications.

💻 Developer Experience (DevX)

  • VSCode Extension: A dedicated sidebar extension (vscode-boring) to control the agent, view status, and launch the dashboard.
  • Interactive Tutorial 2.0: A gamified CLI tutorial (boring tutorial) to onboard new users with missions and badges.
  • Visual Dashboard: Real-time WebSocket-powered dashboard for monitoring agent thoughts and detailed logs.

🔌 Offline-First Mode

  • Zero-Network Dependency: Core functionalities (RAG, Logic, Editing) now work 100% offline.
  • Local LLM Support: Native integration with local_llm via llama-cpp-python.
  • GGUF Model Manager: CLI to download and manage quantized models (Llama 3, Qwen 2.5).

🚀 CLI Improvements

  • New Commands:
    • boring flow: Start the One Dragon autonomous flow.
    • boring tutorial: Launch the interactive gamified tutorial.
    • boring wizard: Now supports 15+ clients (Cursor, Claude, VSCode, Trae, etc.).
    • boring doctor: Comprehensive check (with --fix and --optimize support).
    • boring bisect: AI-driven autonomous bug hunting.
    • boring bio: Generate project biography for knowledge continuity.
    • boring policy: Enforce tool governance policies.
    • boring perfection: Run the 100-point perfection audit.
  • Enhanced boring start: Supports profile and flow arguments.

🛠️ Infrastructure

  • Flow Stage Metadata: Tools are now semantically tagged with stages (Design, Implement, Polish) for smarter routing.
  • Unified State Directory: Consolidated all config/state into .boring/ directory.
  • FastMCP Async: Optimized asynchronous tool registration.

⚠️ Breaking Changes

  • Legacy Removal: Old directories (.boring_memory, .agent) are deprecated. Use boring clean to remove.
  • Config: Main config moved to .boring.toml.
  • Tool Names: Some tools renamed for consistency (e.g., boring_skills_browse -> boring_skills_list).

👥 Contributors

Thanks to the Boring Team and the Open Source community for enabling this Cognitive Revolution!