We have a large part of our code documentation on Confluence in our project. Is it possible to give Cursor AI access to our protected Confluence so that Cursor, for example, knows the background of our architecture decisions?
Thanks in advance!
2 Likes
The docs indexing feature doesn’t handle any authentication yet but it is on our to-do.
If possible you could download the documentation, add it to your project folder, and run it through the codebase indexing feature. Then when you send messages with codebase context, it will also retrieve the most relevant snippets from your downloaded documentation.
3 Likes
I have now exported the Confluence pages as HTML and placed them in a folder within my repository, then waiting for Cursor AI to complete the indexing process of these files.
However, even with the “Chat with Codebase” feature, he was unable to provide me with a satisfactory answer to relatively simple questions. The answers though are clearly described in the documentation. Instead, he often responds like this:
Unfortunately, there are no specific information in the provided code snippets
…
but without further information or access to the full code or documentation, I can’t give an accurate answer.
Surprisingly, however, you can see that he finds the right files as context in which the information should be. But he is not able to extract the sought-after information satisfactorily.
Could it be that he is not so clear with HTML here or that one would have to explicitly tell him that these are documentation files and not source code? Or are the files simply too large?
What can I do to better extract the right information from the documentation?
HTML can have a lot of unnecessary stuff in it. I would just “Ctrl + A => Copy” every page.
1 Like
So I did some text extraction for the .html files und now Cursor AI extract and summarizes some of the relevant context. 
But: Some of the files are quite large and I have found that Cursor in the normal “Chat with Context” function only uses a few small chunks for information extraction. So I played with the advanced context settings and found out that the results get significantly better when I activate Reasoning Steps. Honestly, I don’t know why this is the case.
Is there any documentation on what exactly these Advanced Context Settings do? What is a Reranker? What are Reasoning Steps? And under what circumstances should I activate this?
2 Likes