code blocks in md vs sql / py / json / other in the vault?

[JroachAu]

Hi,
Does Copilot index /understand code blocks within markdown?
This would be a deal maker / breaker for me.

And can I ask it about code issues etc?
I know that there are other plugins that enable the execution of code, using glaredb for example, but I wonder how well copilot works in these situations.

Right now, I usually use Cline in VS Code payg, when vs code pilot doesn't work out quick enough.

Noting that my SQL and py files are in their native file type, but to make them work in obsidian I run a python script over them that creates a md from the original and adds the content to the middle of a template.

A workaround would be to convert non-MD files to MD and code block the code out.

Its probably too much of a hacky use case?

My use for it right now, is to create data dictionaries in md which leverage the content of SQL scripts in its creation.

[logancyang]

Any code block is just content inside your markdown notes. Whether Copilot understands them depends on the embedding model, which you can choose yourself. Copilot is highly customizable.

To pick a good embedding model that understands code and works well with RAG is a different ball game. I don't recommend that route personally but free feel to take a stab at it if you know what you are doing.

Overall it's better to just feed your project entirely into the context window of a good LLM. We are introducing "Project" mode in the next few releases, it will enable this approach.