Embed File Type Filter

[jdspugh]

Is your feature request related to a problem? Please describe.
Some binary files are being embedded which I don't want embedded.

Describe the solution you'd like
A filter for accepting only certain file types (e.g. .mjs, .js. .ts, .html).

Describe alternatives you've considered
A filter for rejecting certain file types.

[rjmacarthy]

Hey, thanks for the interest. Is the .gitignore file not usable for your use case, should there be an additional option?

[jdspugh]

I have some media files in my projects (in this specific case, .wof font files). And it seems they were being processed/added into the embeddings. I want to see the files in my VSC project, but I don't want them processed/added to the embeddings. That's why I thought a file extension filter would be useful. What do you think?

[fishshi]

Hello, this issue should be resolved. Now, Twinny not only uses .gitignore to ignore files during embedding, but you can also add custom ignore items in the global settings.

[mmorys]

I had the same question and found this issue. The above solution works and this issue can probably be closed.

[psyray]

Hello, this issue should be resolved. Now, Twinny not only uses .gitignore to ignore files during embedding, but you can also add custom ignore items in the global settings.

Hi, thanks for the modification. It works, but why not using a specific file for twinny ? Like .twinnyignore
For example, with Cursor.ai we could set a .cursorignore file to prevent specific files to be indexed.

Because when doing embedings using

• .gitignore is project specific but we cannot add files that we want to track using git, or we could add it but we need to remove them later when doing a commit if those ignored files have changed
• global settings is workspace specific, and we could have, in the same workspace, more than one project, git tracked or not, with different techs, and the ignore could be to wide.

Using a specific file for twinny to ignore files could be a more granular approach and provide a lot of flexibility to construct a pertinent embeding context.
We could place those files in each base folder of our workspace and ignore files according to the techs used.
Or set it at the root of the workspace and define accurate rules for each sub folder

So, that file could be placed in the .gitignore file to not be tracked.
And you can let the user the choice to use .gitignore file, or .twinnyignore file.
And the global settings remains for files we are sur we never want to index.

What do you think about this ?