Originally created by @ghost on GitHub (Feb 2, 2024).
Describe the solution you'd like
Enable unique system prompt per-chat and/or per-model. This will allow for flexibility when working with multiple models.
Originally created by @ghost on GitHub (Feb 2, 2024).
**Describe the solution you'd like**
Enable unique system prompt per-chat and/or per-model. This will allow for flexibility when working with multiple models.
Coding models need a really small temperature and a repeat penalty of just a fraction over 1.0 to work properly.
Story generation models need a much higher temperature and repeat penalty.
General purpose models need something in between.
@jukofyork commented on GitHub (Feb 3, 2024):
Yeah, per-model settings would be good too:
- Coding models need a really small temperature and a repeat penalty of just a fraction over 1.0 to work properly.
- Story generation models need a much higher temperature and repeat penalty.
- General purpose models need something in between.
I vote for per chat system prompt, either loaded from library or setup in the chat window when creating the new message to have the ability to select the role - system.
@OneAtDrt commented on GitHub (Feb 4, 2024):
I vote for per chat system prompt, either loaded from library or setup in the chat window when creating the new message to have the ability to select the role - system.
I was wondering if it could be like the Modelfiles but the prompt version of it, i.e., one could import/export/create/discover all sorts of pre-set prompts and/or parametes setting. In the meantime, just like the llm model selection for each chat, there could be an entry where prompts could be selected for different chats as well.
@jcdiv47 commented on GitHub (Mar 19, 2024):
I was wondering if it could be like the Modelfiles but the prompt version of it, i.e., one could import/export/create/discover all sorts of pre-set prompts and/or parametes setting. In the meantime, just like the llm model selection for each chat, there could be an entry where prompts could be selected for different chats as well.
Per chat config feature has been added to our dev branch as "Chat Controls"!
@tjbck commented on GitHub (Jul 9, 2024):

Per chat config feature has been added to our dev branch as "Chat Controls"!
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @ghost on GitHub (Feb 2, 2024).
Describe the solution you'd like
Enable unique system prompt per-chat and/or per-model. This will allow for flexibility when working with multiple models.
@jukofyork commented on GitHub (Feb 3, 2024):
Yeah, per-model settings would be good too:
@OneAtDrt commented on GitHub (Feb 4, 2024):
I vote for per chat system prompt, either loaded from library or setup in the chat window when creating the new message to have the ability to select the role - system.
@jcdiv47 commented on GitHub (Mar 19, 2024):
I was wondering if it could be like the Modelfiles but the prompt version of it, i.e., one could import/export/create/discover all sorts of pre-set prompts and/or parametes setting. In the meantime, just like the llm model selection for each chat, there could be an entry where prompts could be selected for different chats as well.
@minhhungit commented on GitHub (May 10, 2024):
ChatTemplate or ChatProfile is good name I guess
@tjbck commented on GitHub (May 25, 2024):
Per model config has been added to our dev branch!
@tjbck commented on GitHub (Jul 9, 2024):
Per chat config feature has been added to our dev branch as "Chat Controls"!