feat: Allow users to config LLM Inference Parameters as per bot from UI. Close #166.

[jessieweiyi]

Issue #, if available:

166

Description of changes:

This PR is to support users to config LLM Inference Parameters for each bot from UI.

During CI build time, the script located at frontend/scripts/copy-config.py copies the value of GENERATION_CONFIG/MISTRAL_GENERATION_CONFIG from backend/app/config.py based on the VITE_APP_ENABLE_MISTRAL environment variable to file frontend/src/constants/defaultGenerationConfig.json to be used as default generation config. This is to ensure backward compatibility, so that the value in backend/app/config.py is still the source of truth for configuring default value.

• New Bot

  • Create: When users click Create, the generation config (default values if users do not update) is saved to DDB to be used during inference
  • Inference: The generation config from DDB record is used

• Existing Bot (without GenerationConfig value in DDB)

  • Inference: The default generation config (Default/Mistral) from the config.py file is used as before
  • Edit: When users click Edit, generation config (default values if users do not update) is saved to DDB to be used to be used during inference in the future

Questions: For the i18n resource strings, I have added related texts for en/zh-hans/zh-hant. Do we have any processed to build up these? Otherwise, I am happy to use Transcribe to build them for other languages.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[statefb]

Memo: Screen shot of Generation configuration

[statefb]

Questions: For the i18n resource strings, I have added related texts for en/zh-hans/zh-hant. Do we have any processed to build up these? Otherwise, I am happy to use Transcribe to build them for other languages.

Actually we don't have any automated mechanism to do that. Thank you for translating into Chinese languages!

[statefb]

@wadabee When you have time, could you please review the frontend scripts 🙇 @jessieweiyi The review for frontend may be delayed because he is busy now, thank you for your understanding. (Personally, your implementation is LGTM!)

[statefb]

Memo: screenshot