alexrozanski / LlamaChat

Chat with your favourite LLaMA models in a native macOS app
https://llamachat.app
MIT License
1.43k stars 53 forks source link

Sources don't persist #1

Closed luke3butler closed 1 year ago

luke3butler commented 1 year ago

After closing the app and reopening it, the chat sources are cleared out and need to be re-added.

My steps:

  1. Opened app, and added Alpaca in GGML format.
  2. Used the app for a bit.
  3. Closed the app.
  4. Opened app, and got prompted to add a source again. Previous addition disappeared.
alexrozanski commented 1 year ago

Thanks for the report, @luke3butler. Couple of things here:

fakechris commented 1 year ago

Same problems here, ~/Library/Application Support/com.alexrozanski.LlamaChat/sources.json is exists, with contents

{"sources":[{"id":"60E16EE0-B03C-48E4-9005-7E518ADBBBF0","modelDirectoryId":null,"modelURL":"file:\/\/\/Users\/user\/source\/ai\/llama.cpp\/models\/7B\/ggml-model-q4_0.bin.1","avatarImageName":null,"modelSize":"size7B","type":"llama","name":"LLaMA"}]}

alexrozanski commented 1 year ago

ah got it, there was a bug in the chat source deserialisation logic. fixed in 1fafb1f, will release an update shortly. for now, a workaround is to make sure to select an avatar when adding a source.

thanks for the reports!

alexrozanski commented 1 year ago

Fixed in v1.0.1: https://github.com/alexrozanski/LlamaChat/releases/tag/1.0.1