Hello there! Finally decided to give VectHare a try after getting a little frustrated with another memory extension in my 1000+ messages chat. I don't really know what I'm doing most of the time when it comes to SillyTavern, so please bear with me here.
I followed the instructions in the readme and installed the extension, then I picked "Standard" as the vector backend, and Koboldcpp as the embedding provider (running the following embedding model locally: bge-m3-Q4_K_M-GGUF). As for the model I'm using for the actual chat, I'm running a local 12b model, in case that makes any difference.
Then, I proceeded to vectorize my entire chat, which took about 9 minutes or so. After that was done, I disabled the previous memory extension I had been using, unlinked any lorebooks from my chat, and tested prompting the AI to generate a message. After the message was generated, I inspected the prompt via the Prompt Itemization menu, and seemingly no chunks were injected. I then checked the Database Browser, and even though the collection is there, it for some reason has 3259 total chunks, even though my chat has 1083 messages and I'm using the "Per Message" chunking strategy. Also, when I click on the "Open Folder" and "View Chunks" buttons, it returns the error message "Failed to load chunks/open folder: Failed to list chunks/open folder: Not Found."
It feels like there's something very wrong going on here, but I'm not sure how to check if the extension is working as intended or not. Seems strange that not a single chunk is being inserted when my chat has more than 1000 messages. Any help would be much appreciated, and I'm happy to provide more info if necessary. Here's the diagnostics report.
Hello there! Finally decided to give VectHare a try after getting a little frustrated with another memory extension in my 1000+ messages chat. I don't really know what I'm doing most of the time when it comes to SillyTavern, so please bear with me here.
I followed the instructions in the readme and installed the extension, then I picked "Standard" as the vector backend, and Koboldcpp as the embedding provider (running the following embedding model locally: bge-m3-Q4_K_M-GGUF). As for the model I'm using for the actual chat, I'm running a local 12b model, in case that makes any difference.
Then, I proceeded to vectorize my entire chat, which took about 9 minutes or so. After that was done, I disabled the previous memory extension I had been using, unlinked any lorebooks from my chat, and tested prompting the AI to generate a message. After the message was generated, I inspected the prompt via the Prompt Itemization menu, and seemingly no chunks were injected. I then checked the Database Browser, and even though the collection is there, it for some reason has 3259 total chunks, even though my chat has 1083 messages and I'm using the "Per Message" chunking strategy. Also, when I click on the "Open Folder" and "View Chunks" buttons, it returns the error message "Failed to load chunks/open folder: Failed to list chunks/open folder: Not Found."
It feels like there's something very wrong going on here, but I'm not sure how to check if the extension is working as intended or not. Seems strange that not a single chunk is being inserted when my chat has more than 1000 messages. Any help would be much appreciated, and I'm happy to provide more info if necessary. Here's the diagnostics report.