Suggestion: Streamline Focus on Code Completion

[dmanresa-saes]

Hello,

I've observed that recent additions like AI chat and commit messages are complicating the codebase. Repositories such as RooVetGit/Roo-Code offer these features and have a large user base, but they lack code completion.

By focusing on enhancing the code completion feature, this extension can attract Roo-Code’s users who are looking for this capability, thereby increasing the project's appeal.

Thank you for your consideration!

[igardev]

@dmanresa-saes, thank you for the suggestion.
On one hand makes sense to focus the time and ressources on one thing and make it as good as possible. On the other hand the future is unclear. It is possible that in this way I could endup in a situation similar to someone, who makes the best compact disks in the world (but disks are not used anymore).
For me llama.vscode should focus on making AI benefits (in IDE) available for free. By free I mean:

• Free from paying money
• Free from giving the source code to external parties
• Free from sharing secrets with external parties
  The approach to achieve this is by providing optimized local solutions with llama.cpp and open source LLMs.

Therefore I think that it makes sense to provide new feature and optimize them for local usage even if they are not at the level of some other extensions.

Maybe some day the code completion part could be split as a separate extension for simplicity. We will see.

[dmanresa-saes]

Hello, igardev.

I truly appreciate your efforts with llma.vscode. It's refreshing to see an extension without overwhelming code, making it much easier to understand the internal workings. I was particularly pleased to grasp the strategy of the chunks sent as context for code completion, something that becomes quite complex in larger and more intricate extensions.

However, one challenge I've encountered is the time taken to index a new project. With thousands of files, it can sometimes become impractical. I hope that future updates will address this issue, optimizing the indexing process while maintaining the qualities that make llma.vscode a great choice for many users.

Initially, when I started using Roo-Code, I noticed the lack of code completion and sought out extensions to fill this gap. The first one I tried was continue.dev, which impressed me with its performance on local projects. It offers additional features like chat and commits and ensures privacy as it doesn't send code to external servers, aligning with your emphasis on free and secure development.

Despite initially favoring continue.dev, I ultimately chose llma.vscode for its lightweight and simple structure. It would be ideal if llma.vscode could adopt a modular approach, allowing users to enable or disable features based on their needs. This would enhance the specialization and attract users seeking cleanliness and technical finesse.

I'm excited to see how the extension evolves and how you might integrate feature modularity in future versions. It could take llma.vscode to great heights!

[igardev]

Hi @dmanresa-saes,

Thank you for the kind words!
You are right - it is important that the users could disable/enable features. In the next version (0.0.12) I will add a menu Item "Disable RAG". It will disable Chat with AI with project context and indexing. This menu item will toggle the new setting "rag_enabled".

Just for information - for showing the menu you could use the shortcut Ctrl+Shift+M or just click llama.vscode item in the statusbar.

Thanks for helping us improve llama.vscode.

[igardev]

Version 0.0.12 is now released.