Skip to content

Latest commit

 

History

History
73 lines (55 loc) · 3.43 KB

modules.md

File metadata and controls

73 lines (55 loc) · 3.43 KB

Modules

Contents

  1. Module Documentation
  2. Module Symbol Order
  3. Disabling a Module
  4. Components of a Module

Module Documentation

Module Symbol Order

In the settings.yaml in the "Linguflex/lingu" folder there is a section "modules".
The icons displayed in the top right of Linguflex will be shown in the order defined under this section

Disable a module

Every module whose directory folder starts with an underscore ("_") will be ignored at the start of Linguflex.

You can disable any linguflex module by preceding a underscore before the module directory name:

  • navigate to the folder "Linguflex/lingu/modules"
  • select a module (for example, say "weather" module)
  • disable the module by renaming the directory (rename "weather" to "_weather") and add an underscore at the beginning

Components of a Module

Linguflex modules are designed to provide a clear framework for development for and interaction with LLM and to clearly separate business logic from system-specific code. The module structure ensures that each critical aspect of the module's inference process is encapsulated in its respective file, providing a comprehensive and efficient framework for development within the Linguflex ecosystem.

  1. Handlers Folder:

    • Contains business logic that is independent of Linguflex, allowing for isolated testing.
    • This is where you implement everything that does not directly involve Linguflex.

  2. Inference.py:

    • Defines the functions to interact with the LLM.
    • Every class derived from Populatable defines such a function.
    • Includes Pydantic Fields as data structures to be populated by AI.
    • The on_populated method is executed after the fields have been populated and may call business logic functions.
    • May also call business logic functions from logic.py.

  3. Inference.XY.json:

    • Tailors functions for specific languages, denoted by the language shortcut 'XY'.
    • Outlines keywords correlating to the functions defined in inference.py.
    • A function is available for the LLM to call only if a keyword match is found (wildcard * allowed)
    • Limits the functions offered to the LLM, ideally restricting it to 3-5 functions, which prevents overwhelming it and helps avoiding confusion.
    • Optional prompts: init_prompt when a function is offered, success_prompt after successful function execution, and fail_prompt if the function fails.

  4. Logic.py:

    • Contains business logic specifically related to Linguflex.
    • Calls business logic from the Handlers folder.
    • Performs additional Linguflex-related actions like state handling or raising triggers.
    • Methods can be called from the UI or inference.py

  5. State.py:

    • Acts as a data class maintaining the module’s state.
    • This state is accessible from both logic and UI.
    • Stores module-related data (like the symbol).
    • Includes a save() method for serialization to file.
    • All data structures used must be serializable as they might be stored on hard disk.

  6. UI.py:

    • Defines the user interface (window that opens when the user clicks the module symbol)
    • Offers on-the-fly module configurations and runtime adjustments.