Skip to main content

Vibe Coding / IDE AI Guide

Give Cursor, Codex, Claude Code, and similar assistants the right context to generate accurate XSpoonAi code. Pick one of the four methods below; they are independent, ordered from most guided to most manual.

Method 1: Use MCP / online retrieval (no local clone)​

  • Configure MCP connectors (e.g., deepwiki/context7) to fetch files from GitHub or the web.
  • Target upstream repos directly: XSpoonAi/spoon-core and XSpoonAi/toolkit (source + examples) plus this Cookbook site for docs.
  • Instruct the assistant to: list relevant files, fetch the exact file contents it will mimic, then cite paths/lines. No need to mirror or clone locally.

Method 2: Supply the bundled docs file (cookbook/llm.txt)​

  • The repo ships an auto-generated bundle of every Markdown doc at cookbook/llm.txt (CI keeps it fresh).
  • Share that single file with your LLM to give it the full Cookbook context for Vibe Coding.
  • GitHub copy (fallback): llm.txt.

Method 3: Point to installed package paths​

  • If you installed editable/local packages, expose their locations instead of the repo: 
    python -c "import importlib.util, os; spec = importlib.util.find_spec('spoon_ai'); print('spoon_ai:', os.path.dirname(spec.origin) if spec and spec.origin else 'not installed')"
    python -c "import importlib.util, os; spec = importlib.util.find_spec('spoon_toolkits'); print('spoon_toolkits:', os.path.dirname(spec.origin) if spec and spec.origin else 'not installed')"
  • Provide those paths to the assistant as read-only references so it can scan the actual shipped code and examples inside the installed packages.
  • If a package isn’t installed (output shows “not installed”), just share the corresponding repo directory (e.g., core/spoon_ai or toolkit/spoon_toolkits) instead.

Method 4: Share the workspace directly​

  • Allow the assistant to read this repo (especially core/, toolkit/, cookbook/docs/). No extra cloning required.
  • Need a local checkout first? From your working directory run: 
    git clone https://github.com/XSpoonAi/spoon-core.git
    git clone https://github.com/XSpoonAi/spoon-toolkit.git
  • Keep the repo up to date (git pull) so the assistant always sees current code and docs.
  • Ask it to open real source first, e.g. core/spoon_ai/** and runnable samples in core/examples/ and toolkit/**/examples.
  • Tell the AI to derive function signatures and configs from code files or examples, not invented abstractions.

Hallucination-reduction tips (tell the AI)​

  • "Read the source or examples you will copy from; cite file path and line range before coding."
  • "Confirm tool/agent/LLM signatures from code, not from guesswork; show the imports you plan to use."
  • "If unsure, fetch the specific file via MCP and restate the interface before implementing."