You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Good afternoon!
I had a great idea when playing with ModelFiles in the webui.
What if it were possible to attach RAG documents (or a group of RAG documents) to a modelfile? I know a lot of work would have to be put into this because ollama doesn't support this directly, but I think it could be done. It would be nice to have a model file that would just always have a rag document attached to it when conversing with it, rather than including all my data in the system prompt or manually attaching a document every time I ask it a question.
For example, I could make a modelfile that's based on mixtral or something but has a RAG document attached to it with a big essay about my life or whatever, and anyone who asks the model a question about me would get an answer from it utilizing RAG. This would also be nice if this worked through the API as well.
The text was updated successfully, but these errors were encountered:
Good afternoon!
I had a great idea when playing with ModelFiles in the webui.
What if it were possible to attach RAG documents (or a group of RAG documents) to a modelfile? I know a lot of work would have to be put into this because ollama doesn't support this directly, but I think it could be done. It would be nice to have a model file that would just always have a rag document attached to it when conversing with it, rather than including all my data in the system prompt or manually attaching a document every time I ask it a question.
For example, I could make a modelfile that's based on mixtral or something but has a RAG document attached to it with a big essay about my life or whatever, and anyone who asks the model a question about me would get an answer from it utilizing RAG. This would also be nice if this worked through the API as well.
The text was updated successfully, but these errors were encountered: