Open andaag opened 8 months ago
thanks for kudos @andaag, checked, all good.
create a new assistant with "code interpreter", add it to tutors.yaml
, select one (/tutor
) before questions.
note: can't send files in chat yet. if code interpreter makes a file, you won't see it in chat
from logs: tools=[ToolCodeInterpreter(type='code_interpreter'), ToolRetrieval(type='retrieval')]
I had a few conversations going with it, and I realize what happened. If the assistant split the messages in two I think the responses from gptbot is mixed. Notice how answer comes before description on how you arrive at the conclusion.
I tested the example case:
The output ends with "lets do that now", and it seems the next part of it is not executed/fetched.
(Also cool project 👍, haven't seen any other telegram capable api that allows use of this beta api yet).