By skyforbes However, ChatGPT Pro cannot read any file contents and can only list and search (which confirms it's successfully connecting to my repo).
I tried waiting, tried multiple different sessions, same results.
Specifically, it is unable to do fetch_file or fetch as they fail and result in an AttributeError: 'str' object has no attribute 'get'. Apparently the the file‑content codepath is broken because when the connector gets a string body (HTML page or raw bytes), it treats it like a dict and calls .get(…), causing the AttributeError.
It's quite frustrating because it seems like 1) the Host allow‑list is too strict (fetch(url=…) rejects api.github.com and raw.githubusercontent.com), 2) there's some kind of type assumption bug (the implementation assumes JSON/dicts and calls .get(…) even when the body is text/HTML (string) or raw bytes), and 3) fetch_file looks like it likely builds a github.com/…/blob/… URL under the hood (HTML), not the API URL, which combined with 2), makes it crashes?
I need the Pro model to analyze the code inside the files in my repo which I thought was the entire point of this Github connector in ChatGPT, so how can I fix this?