dotai / .ai

dotai is a self-improving AI system designed to track errors and learnings, creating better context-aware interactions in your project.

Features

• Codex (codex.md): Logs mistakes and learnings.
• Learn Protocol (learn.md): Updates the Codex after corrections.
• Split-Codex: Categorizes Codex into manageable sections.
• Session Management: Keeps project context across sessions.
• Blueprints: Guides for setting up various tech stacks.
• Snippets: Code templates to streamline AI's code generation.

Installation

1. Clone this repository or use it as a template.
2. Create a .ai folder in your existing project:

   mkdir .ai

3. Move the contents of this repo into the .ai folder.

Usage Workflow

1. Run Prompt: Interact with the AI.
2. Error Correction: If AI makes a mistake, correct it by invoking Learn.
3. Invoke Learn: Enter @learn.md @codex.md to update Codex.
4. Manage Codex: Use @split-codex to categorize entries when needed.
5. Persistent Memory Setup:
   • Create a Composer/Notepad project named 'Codex'.
   • Add codex.md to the project context for persistent project memory without needing explicit learn.md mentions.

Important Note: Adding Codex to Chat Context

codex.md should be added to the context of every chat:

• For regular chats: Use the plus button at the top of the chat to add the file.
• For Composers: Add the file to a Project Composer so all Composers created in that project will automatically have the file.

This ensures that the AI has access to all logged errors and learnings for each interaction.

Session Management

Session files create a "memory layer" for the AI across multiple interactions, enabling contextual awareness and adaptive assistance.

Key Benefits:

• Maintains project context between sessions
• Reduces repetition of project details
• Provides consistent guidance aligned with project direction

Usage:

1. Start a session: Use @start-session command

   • start-session.md: Initiates a new AI session

2. End a session: Use @end-session command

   • end-session.md: Concludes the current AI session

The AI will generate and read status files in the status directory to maintain project continuity.

Codex Structure

The Codex is divided into two main sections:

1. Errors: Mistakes made and how to prevent them.
2. Learnings: Insights gained and their applications.

Each entry includes context, description, correction/application, and related entries.

Snippets

Snippets include code templates that AI can use to generate or refactor code:

• Use create-snippet.md to create new snippets.

Blueprints

Comprehensive guides for implementing specific technical architectures:

• supabase-drizzle-actions.md: Backend with Supabase, Drizzle ORM, and Server Actions
• flux-with-replicate.md: Image generation using Flux and Replicate

Additional Components

• Libraries: lib contains documentation examples for library usage.
• Plugins: v0 provides a guide for using v0.dev to generate component ideas and prompts.
• Rules: Rules contains rules for default AI behavior and interaction.
  • code-tutor.md: To provide a supportive learning environment for beginners learning to code.
  • mckay.md: For efficient, high-level interactions without unnecessary explanations.

Contributing

This is an open-source template. Contributions are welcome! Please add a changelog entry with your contribution.

License

[Include license information here]