Feature: Consolidate Configuration into a Unified Settings File

[MeetCoding]

To streamline user configuration and enhance flexibility, the current prompts.json system should be deprecated in favor of a single, unified configuration file: settings.ghostdev. This new file will serve as the central location for managing custom prompts, specifying LLM providers, and securing API keys, simplifying the setup process for users.

Proposed Functionality

• New Configuration File: ​Introduce a new configuration file, settings.ghostdev, to be located in the project's root directory.
  ​This file will replace the existing .ghostdev/prompts.json. The system should prioritize settings.ghostdev if both files are present and provide a clear deprecation warning.

• Configuration Schema: The settings.ghostdev file will support the following configurations:
  ​LLM Provider and Model Selection: Allow users to specify which Large Language Model they wish to use (e.g., OpenAI, Google, Anthropic) and the specific model (e.g., GPT-4, Gemini Pro).

• ​API Key Management: Provide a field for users to securely input their personal LLM API key, enabling them to use their own accounts.

• ​Custom Prompts: Migrate the custom prompt definition functionality from prompts.json into the new settings file.

[jeloskaisabel]

@MeetCoding could you assign it to me?

[MeetCoding]

Hey @jeloskaisabel
Thank you for your interest, can you please outline your approach to resolve this?

[jeloskaisabel]

I'd start by creating the new format that brings everything together, LLM provider choice, API keys, and the custom prompts. The main work would be updating load configuration to check for the new file first, while still supporting the old prompts.json for backward compatibility (with a heads-up that it's deprecated).
The trickier part is making the LLM integration flexible enough to work with different providers. Right now everything's hardcoded for Gemini, so I'd need to abstract that out to handle OpenAI, Anthropic, etc. based on what's in the config.

[D3vanshC]

Hey @jeloskaisabel ,
The approach fits well with our vision of this project, the issue has now been assigned to you, happy coding :)