Skip to content

CardSorting/codemarie

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

156 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

CodeMarie: The Architectural Guardian

CodeMarie Banner

CodeMarie is an industrial-grade, model-agnostic agentic coding assistant designed to maintain architectural integrity in complex software ecosystems. Beyond simple code generation, CodeMarie acts as an Architectural Guardian, enforcing strict layering, managing structural entropy, and ensuring transactional stability across your workspace.

Important

Oracle Grade Release (v3.87.0): This release marks the transition to the Observe-Act-Adjust execution model, prioritizing forward progress and reliability by removing high-latency grounding loops while maintaining strict architectural enforcement via the Spider Engine.


๐Ÿ—๏ธ Core Pillars of Intelligence

๐Ÿงฌ Joy-Zoning Framework

CodeMarie enforces a rigorous architectural pattern known as Joy-Zoning. It automatically categorizes every file into distinct layers and enforces "Outside-In" dependency rules.

  • Layer 1 (Shared/Core): Foundational types and utilities.
  • Layer 2 (Services): Business logic and data providers.
  • Layer 3 (UI/Handlers): Components and interaction logic.
  • Layer 4 (Controllers): Extension lifecycle and event management.

Tip

The Fluid Policy Engine monitors every file operation via the Universal Guard to prevent layer leaks and circular dependencies in real-time.

๐Ÿ•ท๏ธ Spider Structural Intelligence Engine

The Spider Engine represents a fundamental shift from static code analysis to Structural Intelligence:

  • Dependency Web analysis: Treats the codebase as a living web, sensing "vibrations" (structural shifts) on every write.
  • Incremental $O(1)$ Audits: Uses a CAS-hash based model to provide near-instantaneous architectural validation without scanning the entire repo.
  • Blast Radius Intelligence: Predicts the multi-hop impact of a change before execution, identifying potential "Structural Entropy" growth.

๐Ÿ”ฎ Oracle Grade Suggestion Engine

The Oracle Suggestion Engine transforms prompt suggestions into an architectural and diagnostic compass:

  • Spider-Powered Symbol Resolution: Automatically identifies and resolves the definitions of types causing workspace errors.
  • Project-Wide Consistency: Extracts and enforces your project's dominant idioms, error handling conventions, and naming styles.
  • Parallelized Source Discovery: Simultaneously gathers intelligence from Diagnostics, Git, BroccoliDB, and Tree-Sitter with zero detectable latency.

๐Ÿ“ System Architecture

1. Functional Layout

CodeMarie is built on a decoupled, high-throughput architecture designed for stability and observability.

graph TD
    classDef primary fill:#1e293b,stroke:#3b82f6,stroke-width:2px,color:#fff;
    classDef secondary fill:#1e293b,stroke:#a855f7,stroke-width:2px,color:#fff;
    classDef database fill:#1e293b,stroke:#10b981,stroke-width:2px,color:#fff;
    classDef warning fill:#1e293b,stroke:#ef4444,stroke-width:2px,color:#fff;

    User((User Workspace)):::primary --> Extension[VS Code Gateway]:::primary
    Extension --> Controller[Core Controller / Main Event Loop]:::primary
    
    subgraph "Structural Intelligence"
        Controller --> Spider[Spider Engine]:::warning
        Spider --> JZ[Joy-Zoning Guard]:::warning
    end

    subgraph "Reasoning & Context"
        Controller --> KGS[Knowledge Graph Service]:::primary
        Controller --> SugS[Oracle Suggestion Service]:::secondary
    end
    
    subgraph "Transactional Persistence"
        Controller --> DB[(BroccoliDB / SQLite)]:::database
        DB --> Shadows[Shadow Workspaces]:::database
        DB --> Nodes[Knowledge Nodes]:::database
    end
    
    Controller --> MCPHub["Model Context Protocol (MCP) Hub"]:::primary
Loading

2. Execution Flow: Observe-Act-Adjust

CodeMarie utilizes a high-reliability autonomous loop that prioritizes forward progress over recursive validation.

graph LR
    Start([User Intent]) --> Observe[1. Observe]
    Observe -- "Context & Diagnostics" --> Act[2. Act]
    Act -- "Execute Tools" --> Audit{3. Policy Audit}
    Audit -- "Violation" --> Feedback[Correction Feedback]
    Feedback --> Act
    Audit -- "Success" --> Result[4. Observation]
    Result -- "Status Update" --> Observe
Loading

๐Ÿ› ๏ธ Industrial Infrastructure

๐Ÿ”— Advanced MCP Hub

Full integration with the Model Context Protocol (MCP):

  • SSE & Stdio Transports: Multi-protocol support for local and remote tool servers.
  • Native OAuth: Integrated authentication for enterprise-grade tool integrations.

๐Ÿ“Š OpenTelemetry Observability

High-fidelity telemetry for audit trails and performance tuning:

  • Token Economics: Precise cost tracking per task and turn.
  • Stability Metrics: Monitoring "Architectural Entropy" and policy violation trends.

โšก Quick Start

  1. Install: Search for "CodeMarie" in the VS Code Marketplace.
  2. Configure: Add your API keys for OpenRouter, Anthropic, or Google.
  3. Activate: Click the CodeMarie icon in the sidebar and start your first "Architectural Intent" resolution task.

๐Ÿ•ฐ๏ธ History & Origins

CodeMarie is a completely transformed, industrial-grade evolution of the original Cline repository. While it shares foundational DNA, the architecture, orchestration, and policy safeguarding have been reconstructed from the ground up to support enterprise-scale agentic coding.


Built with โค๏ธ by the CodeMarie Team. Architectural Integrity is the core.

About

CodeMarie is your autonomous coding partner, capable of creating/editing files, running commands, and more with your permission.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors