You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
ragKnowledge embeds entire documents and then retrieves them at runtime, blowing up prompts.
To Reproduce
Put a large (>1MB) text file in rag knowledge, and then message your agent with a large portion of that document so that it gets retrieved via ragKnowledge during runtime.
Expected behavior
Either don't embed entire documents or filter them out in runtime.
Additional context
There is some code duplication in ragKnowledge.ts between processFile and createKnowledge. I think those can be cleaned up and we must consdier 1. Do we need entire document embeddings? How to properly filter them out at runtime retrieval especially if they are too big?
I have a temporary fix in sqlite adapter doing this:
let sql = `SELECT * FROM knowledge WHERE (isMain = 0) AND (agentId = ? OR isShared = 1)`;
Hello @LinuxIsCool! Welcome to the elizaOS community. Thank you for opening your first issue; we appreciate your contribution. You are now an elizaOS contributor!
Describe the bug
ragKnowledge embeds entire documents and then retrieves them at runtime, blowing up prompts.
To Reproduce
Put a large (>1MB) text file in rag knowledge, and then message your agent with a large portion of that document so that it gets retrieved via ragKnowledge during runtime.
Expected behavior
Either don't embed entire documents or filter them out in runtime.
Additional context
There is some code duplication in ragKnowledge.ts between processFile and createKnowledge. I think those can be cleaned up and we must consdier 1. Do we need entire document embeddings? How to properly filter them out at runtime retrieval especially if they are too big?
I have a temporary fix in sqlite adapter doing this:
I originally commented on this here: #1620 (comment)
The text was updated successfully, but these errors were encountered: