Skip to content

architecture.md

Latest commit

 

History

History
91 lines (71 loc) · 2.99 KB

architecture.md

File metadata and controls

91 lines (71 loc) · 2.99 KB

Architecture Evolution

Current Setup (Version 1)

Message-based code generation with single-file output

sequenceDiagram
    participant U as User
    participant AI as AI

    U->>AI: Send message
    AI->>U: Generate code (single file)
    U->>AI: Send message with code (versioned)
    AI->>U: Extract code from all messages containing code strings and process
Loading
  • Workflow: User request -> AI creates entire app.tsx -> Extract single code file from AI response
  • User interactions via messages
  • Code stored/extracted from message history
  • Versioning through message chains

Version 2: Vercel AI SDK Integration

Vercel AI SDK with tool streaming and enhanced reasoning

sequenceDiagram
    participant U as User/Browser
    participant V as Vercel AI SDK
    participant T as AI Tools

    U->>V: Send request with file context
    V->>T: Stream tools for reasoning & file operations
    T->>V: Execute tool calls (edit, search, run commands)
    V->>U: Stream responses with tool outputs
    U->>V: Continue conversation with tool results
Loading
  • Workflow: User request -> Vercel AI SDK streams tools -> Tools execute operations -> Stream results back
  • Migrate from Together AI to Vercel AI SDK for better tool streaming
  • Enhanced reasoning capabilities with tool calling
  • Real-time streaming of AI thinking and tool execution
  • Support for complex multi-step operations

Version 3: Multi-File Project Support

Client-side multi-file storage and editing

sequenceDiagram
    participant U as User/Browser
    participant AI as AI

    U->>AI: Send message
    AI->>U: Generate multiple files in browser storage
    U->>AI: Send message with file context
    AI->>U: Regenerate smaller files as needed
Loading
  • Workflow: User request -> AI creates multiple files -> Store in browser storage -> Regenerate smaller files as needed
  • Migrate away from storing code in messages
  • Project-based approach with client-side file storage
  • Faster editing due to smaller, targeted file regeneration

Version 4: Autonomous AI Agent

Fully autonomous development with comprehensive tool access

sequenceDiagram
    participant U as User/Browser
    participant A as AI Agent
    participant T as Tools
    participant FS as File System
    participant WS as Web Services

    U->>A: Send request with project context
    A->>T: Execute complex operations autonomously
    T->>FS: Modify multiple files, create directories
    T->>WS: Web searches, API calls, install dependencies
    FS->>A: Return file system state
    WS->>A: Return external data
    A->>U: Present completed project with reasoning
Loading
  • Workflow: User request -> AI agent autonomously executes complex operations -> Full file system manipulation -> External integrations -> Deliver complete solution
  • Fully autonomous AI agent with comprehensive tool access
  • Complete project lifecycle management
  • Advanced integrations: web APIs, package management, deployment
  • Reasoning and planning across entire development workflow