Similar to how Open ai have their memory with ChatGPT, where you can store data that the AI can remember across different chats, and also it can reinforce memory of a current chat.
It already more or less works like that: unlike you have hundreds of messages, the whole conversation is sent with each request and is taken into consideration. It’s just that the LLM is not so good at identifying which part of the previous messages in long conversation should be taken into consideration for his current response…
Red Bucket
•
Dec 12, 2025
I think you just explained why this is necessary: It’s just that the LLM is not so good at identifying which part of the previous messages in long conversation should be taken into consideration for his current response…
In my experience, feeding it very large data and then having it keep track of the details has been a sub-optimal experience.
I’m working on a TTRPG, which has significant worldbuilding, mechanics, and rules that I’m trying to keep interwoven. The more the conversation grows, the more the AI tends to forget details and start making stuff up. I have to keep reminding it of details from much earlier in the conversation, and eventually it just gets borked.
I’ve been chopping up the work across multiple conversations, but this adds more load to my own cognition, as I now have to track that I’m keeping all the aspects in mind when working in the different chats. I’d LOVE if I could point multiple chats at a single dataset and have one chew through probability matrices, another checking game balance based on experience spend across multiple classes, and another verifying that each rule and mechanic fits back into the worldbuilding as much as possible.
Aqua Toast
•
May 7, 2025
Kind of duplicate (notice that the whole history is always passed to the LLM), please see:
Absolutely what I asked for. Hope they provide itm
Sapphire Frog
•
Jul 1, 2025
This is becoming a huge issue. I don’t want to renew my pro subscription until the models have better context into the conversation.
Silver River
•
Feb 9, 2025
Seems like you could do that with a system prompt in the settings panel but I could be wrong.
Yellow Oak
•
Jul 23, 2025
I'm not expert but read a bit about ControlNet integrations with SD, for things like Character constant pose control and more. Such features would be nice and useful for NFT parts creation and more. Is Venice using such tech right now? Thank you… 🙏
Can we get customizable accounts. By that I mean the assistant learns the users style and rules for writing and implements them as it produces content.
Jack
Team•
Jul 14, 2025
A memory system is in progress, for now you can use system prompts to tailor responses from the LLMs
Log in to comment and vote
Comments39
Jack
Jan 22
PinnedMemory is live for Pro users.
Jack
Jul 1, 2025
We’re working on this.
Aqua Toast
May 7, 2025
It already more or less works like that: unlike you have hundreds of messages, the whole conversation is sent with each request and is taken into consideration. It’s just that the LLM is not so good at identifying which part of the previous messages in long conversation should be taken into consideration for his current response…
Red Bucket
Dec 12, 2025
I think you just explained why this is necessary: It’s just that the LLM is not so good at identifying which part of the previous messages in long conversation should be taken into consideration for his current response…
In my experience, feeding it very large data and then having it keep track of the details has been a sub-optimal experience.
I’m working on a TTRPG, which has significant worldbuilding, mechanics, and rules that I’m trying to keep interwoven. The more the conversation grows, the more the AI tends to forget details and start making stuff up. I have to keep reminding it of details from much earlier in the conversation, and eventually it just gets borked.
I’ve been chopping up the work across multiple conversations, but this adds more load to my own cognition, as I now have to track that I’m keeping all the aspects in mind when working in the different chats. I’d LOVE if I could point multiple chats at a single dataset and have one chew through probability matrices, another checking game balance based on experience spend across multiple classes, and another verifying that each rule and mechanic fits back into the worldbuilding as much as possible.
Aqua Toast
May 7, 2025
Kind of duplicate (notice that the whole history is always passed to the LLM), please see:
https://featurebase.venice.ai/fr/p/characterstory-state-file
https://featurebase.venice.ai/fr/p/save-to-character-context-button
Coffee Falcon
Dec 21, 2025
Absolutely what I asked for. Hope they provide itm
Sapphire Frog
Jul 1, 2025
This is becoming a huge issue. I don’t want to renew my pro subscription until the models have better context into the conversation.
Silver River
Feb 9, 2025
Seems like you could do that with a system prompt in the settings panel but I could be wrong.
Yellow Oak
Jul 23, 2025
I'm not expert but read a bit about ControlNet integrations with SD, for things like Character constant pose control and more. Such features would be nice and useful for NFT parts creation and more. Is Venice using such tech right now? Thank you… 🙏
An Anonymous User
Aug 8, 2025
•Merged request
•1 vote
Customized Accounts
Can we get customizable accounts. By that I mean the assistant learns the users style and rules for writing and implements them as it produces content.
Jack
Jul 14, 2025
A memory system is in progress, for now you can use system prompts to tailor responses from the LLMs
https://venice.ai/blog/god-mode-the-power-of-ai-system-prompts
An Anonymous User
Aug 19, 2025
•Merged request
•1 vote
Saved Memories
Add memory feature where you can store preferences for longer-term.