openai
Automatic use API key once chatGPT account limits exceeded
What feature would you like to see?
Hello Could you please add configuration to continue a workflow smoothly using API key once chatGPT account limits exceeded? Very convenient and do not break action in the middle of code change.
Thanks in advance
Are you interested in implementing this feature?
No response
Additional information
No response
Probably closely related: https://github.com/openai/codex/issues/2733
Not really. Using IDE extension you do not see an option to have both login and API. You have to log off before you see an option to connect using API KEY. If you login and reach your limits, you have manually logoff and force to use the KEY. Result - broken workflow and lost context. My expectation is that the agent continue workflow after login limit exhausted using the KEY.
Result - broken workflow and lost context.
This is so true. Losing context also means higher token usage, which leads to hitting limits faster, spending more, and getting worse results.
I’ve lost context before due to an authentication error right in the middle of a feature, and I had to log out and log back in.
Lack of history is another known issue, and it should definitely be taken into consideration when switching from subscription to API key.
Probably closely related: https://github.com/openai/codex/issues/2733
Not really. Using IDE extension you do not see an option to have both login and API. You have to log off before you see an option to connect using API KEY. If you login and reach your limits, you have manually logoff and force to use the KEY. Result - broken workflow and lost context. My expectation is that the agent continue workflow after login limit exhausted using the KEY.
I don't even see that. Where do you see you can even put in any API credentials in the first place? It's only a ChatGPT login no? I'm not talking about the CLI only. I'm talking about the VS code extension. Like in the extensions side bar. And it looks like a ChatGPT UI.
A small initial remedy, even if it is not the ideal solution sought here:
Resuming old sessions
Resuming old conversations with codex resume (https://github.com/openai/codex/pull/3537, https://github.com/openai/codex/pull/3625).
Edit: not working - https://github.com/openai/codex/issues/4065