login (via github oauth) doesnt get detected after link is successfully opened. Issue 1st started in freebuff about last week, so i would have to log into codebuff (successfully), then running freebuff after this worked/would share the successful login , Now the same issue has started in codebuff as of yesterday, making both unable to use and no way to bypass even after manually reopening the links or trying to rerun the login in multiple tabs of both codebuff and freebuff at the same time with multiple newly generated login links, refreshes, or retrying after codebuff.com and gh logout and relogin. (Windows,, npm)
While I'm here there's also an older preexisting issue with tool calls such spawning agents intermittently failing(?), displaying as raw text/json instead of rendering like usual and then chat stops, needing to be reprompted constantly. or at least seemingly failing though often the LLM calls seem to be going through bc the LLM often continues to return subsequent reflections and further raw text tool calls and editing files which dont actually apply to file and just show as raw text tool calls yet model thinks it finished the task, in this case when the chat responses dont stop immediately it always goes incredibly fast through all these false edits and response steps in a second or 2. likely a waste of real server inference
login (via github oauth) doesnt get detected after link is successfully opened. Issue 1st started in freebuff about last week, so i would have to log into codebuff (successfully), then running freebuff after this worked/would share the successful login , Now the same issue has started in codebuff as of yesterday, making both unable to use and no way to bypass even after manually reopening the links or trying to rerun the login in multiple tabs of both codebuff and freebuff at the same time with multiple newly generated login links, refreshes, or retrying after codebuff.com and gh logout and relogin. (Windows,, npm)
While I'm here there's also an older preexisting issue with tool calls such spawning agents intermittently failing(?), displaying as raw text/json instead of rendering like usual and then chat stops, needing to be reprompted constantly. or at least seemingly failing though often the LLM calls seem to be going through bc the LLM often continues to return subsequent reflections and further raw text tool calls and editing files which dont actually apply to file and just show as raw text tool calls yet model thinks it finished the task, in this case when the chat responses dont stop immediately it always goes incredibly fast through all these false edits and response steps in a second or 2. likely a waste of real server inference