private-gpt
private-gpt copied to clipboard
zylon-ai
Faster document listing by directly loading ref docs
Improves the speed of launching PrivateGPT's UI when lots of documents are ingested. Improves the speed of the "list" API.
Testing using Postgres 14 docstore with 2,421,062 rows
With PR 1763 elapsed time 12s
22:58:37.941 [INFO ] llama_index.core.indices.loading - Loading all indices.
22:58:49.454 [INFO ] private_gpt.ui.ui - Mounting the gradio UI, at path=/
Without change - 4m 50s
23:00:13.423 [INFO ] llama_index.core.indices.loading - Loading all indices.
23:05:03.283 [INFO ] private_gpt.ui.ui - Mounting the gradio UI, at path=/
This work. Good job. Ship it !
Testing using Postgres 14 docstore with 2,421,062 rows
With PR 1763 elapsed time 12s
22:58:37.941 [INFO ] llama_index.core.indices.loading - Loading all indices. 22:58:49.454 [INFO ] private_gpt.ui.ui - Mounting the gradio UI, at path=/Without change - 4m 50s
23:00:13.423 [INFO ] llama_index.core.indices.loading - Loading all indices. 23:05:03.283 [INFO ] private_gpt.ui.ui - Mounting the gradio UI, at path=/This work. Good job. Ship it !
Thanks for the test and the stats!