Deno Fresh Starter

This is an opinionated starter project for Deno Fresh with Tailwind CSS. The goal is to set the groundwork for developing fullstack apps using Cursor/Windsurf IDEs relying heavily on AI code generation. AI works best when given examples, so this is what the project is providing.

Initially started as a basic Deno Fresh template, this project has evolved to include a complete AI agent ecosystem implementation. This ecosystem demonstrates how to build and manage AI agents with a feedback-driven improvement system.

AI Agent Ecosystem

The project includes a complete implementation of an AI agent system in the /routes/agents folder. Key features include:

Core Concepts

  1. AI Agents: Each agent is defined by a system prompt that fully specifies its behavior and capabilities. Agents can be created, versioned, and improved over time.

  2. Feedback System: The ecosystem includes a sophisticated feedback mechanism:

    • Users can provide feedback on agent responses
    • A specialized "Coach" agent analyzes the feedback
    • The Coach suggests improvements to the agent's instructions
    • New versions are created with improved instructions
  3. Version Control:

    • Each agent maintains a linear version history
    • Only the most recent version is active
    • Each version stores:
      • Name
      • System prompt
      • Changelog
      • Previous version reference
  4. Task Management:

    • Each agent interaction is tracked as a task
    • Tasks store the prompt, response, and any feedback
    • Responses are streamed in real-time
    • Task history is maintained for analysis

Using Without Agents

If you want to use this template without the AI agent ecosystem - just remove the /routes/agents folder.

Development Guidelines

  • See .windsurfrules for development guidelines for Windsurf editor.
  • See docs for functionality description and design decisions.

Website: https://deno-fresh-starter.dev/

GitHub: https://github.com/ai-1st/deno-fresh-starter