Repository navigation
Adding files / folders to a context/query #69
Description
Activity
- addeddocumentationImprovements or additions to documentationImprovements or additions to documentation
on Apr 8, 2025 The SDK APIs for uploading individual files will be:
prepare_fileinstead ofprepare_imagefiles=[...]parameter instead ofimages=[...]
There isn't an API design for adding URL references or entire folders at this stage.
Edit: while the file preparation APIs in the SDK have been implemented, they currently only support the same file types as the image-specific APIs. My original comment incorrectly indicated that other file types were already expected to work.
- addedenhancementNew feature or requestNew feature or requestdocumentationImprovements or additions to documentationImprovements or additions to documentationand removeddocumentationImprovements or additions to documentationImprovements or additions to documentation
on Apr 10, 2025 Hello,
I used lmstudio packages for python, I faced a question, that I want to send a .txt or .docx file to LM Studio local server. But system shows no such function. Is the function of upload these files was not implemented? But LM Studio have file upload function for .txt or .docx.
Reacted by ChilledSlim and Callum ScottHello,
is there any progress of attaching a file (txt, pdf, and others)?
The LM Studio shows this ability, but the lmstudio python-api does not?Looked into the source and saw "# Not yet implemented (server API only supports the same file types as prepare_image)".
Thank you.
I don't use python but csharp to connect to lm-studio at the moment (because I have to integrate with visual studio). But my problem is related to this issue. So I use lm-studio to help with coding and want to add files to my prompts. Like 'here is class x . Please add method y to this class' . Does not work for me yet. When I add input_file to my prompts, the model cannot process the json request. Pure text works, though. It seems this version is not implemented yet?
I have found an example, where somebody added a citation with file content to the prompt? Is there a better way? Maybe the model has to pull the file via MCP, or so?
Thanks in advance for any response
Hey there,
Is it possible to use not image, but a file (like a code file) or a code project dir which you would be running prompt against ?
Idea would be to summarize contents of repository for example
Thank you.