Closed golfinq closed 6 months ago
I didn't know if I should make the *_MODEL
accessors functions or properties. I figured making them functions was safer as to not leak state accidentally.
oh wow, it'll take me a bit to work through this one. I'll get to it later
Looks good, merged. I fixed one edge case
As discussed in https://github.com/tkellogg/fossil/issues/15#issuecomment-1881460619 the configuration variables were not propagating into the code. I fixed this by creating a configuration manager object which performs a series of look ups when asked for a specific variable from the configuration state: first from the sqllite database, then from the
.env
file, and finally from the host environment.I also had to change the way the reduction step works to use the
session_id
and to account for models not found intiktoken
. Additionally, I sorted the imports using isort, changed "embedding model" to "summarizing model" in the settings page, and added atqdm
progress bar to the summarization step.