Introduce "none" LLM provider for execution control

[pelikhan]

Add a new "none" model to prevent LLM execution, enhancing control over script execution with error logging capabilities. This model is useful for scenarios where inline prompts are used exclusively.

---

Summary of Changes

• ✨ Introduced the none model provider:

  • Prevents LLM execution, useful for top-level scripts relying solely on inline prompts.
  • Returns an error indicating no LLM execution is allowed in this context.

• 🛠️ Enhanced parseTokenFromEnv logic:

  • Added support for the none provider alongside echo.

• 📚 Documentation Updates:

  • Updated the "Echo" provider section with additional debugging details.
  • Added a new section for the none provider, explaining its purpose and usage.

• 🗂️ New Model Implementation:

  • Added NoneModel with a custom completer to handle scenarios where LLM execution is disabled.

• 🚀 Extended Model Resolution:

  • Integrated NoneModel into the language model resolution logic.

• 🛡️ Updated llms.json Configuration:

  • Registered the none provider with relevant metadata, marking it as hidden and tokenless.

These changes enhance debugging capabilities and provide a mechanism to disable LLM execution when necessary.

AI-generated content by prd may be incorrect