Fixed a major bug where context from the AI's responses would not be saved, leading partial context being saved only for the user's messages -- causing the AI to not remember what it said itself.
Before
Now
Additonally, streams are now displayed in actual real time, not simulated through sleep(), this should run as fast as your requests can be completed while sending messages in chunks.
callbacked
commented
1 week ago
Fixed a major bug where context from the AI's responses would not be saved, leading partial context being saved only for the user's messages -- causing the AI to not remember what it said itself.
Before
Now
Additonally, streams are now displayed in actual real time, not simulated through sleep(), this should run as fast as your requests can be completed while sending messages in chunks.