vector dimensions issue

[uannyao]

I have tried the meta-llama/Llama-3-8b-chat-hf chatmodel and togethercomputer/m2-bert-80M-8k-retrieval embedding model with embeddingDimension: 768, but I'm having the following error many times when agents are having conversations:

9/30/2024, 5:59:09 PM [CONVEX A(aiTown/agentOperations:agentGenerateMessage)] Uncaught Error: Expected a vector with dimensions 768, received 1024.
at async searchMemories (../../convex/agent/memory.ts:164:12)
at async startConversationMessage (../../convex/agent/conversation.ts:34:4)
at async handler (../../convex/aiTown/agentOperations.ts:74:6)

[ianmacartney]

You might want to clear the `embeddingsCache` table - this command will do it: `npx convex import --table embeddingsCache --replace --format jsonLines /dev/null -y` Or drop console logs along the code path to figure out where the larger embedding is coming from

…

On Mon, Sep 30, 2024 at 3:03 AM UAnn Yao ***@***.***> wrote: I have tried the meta-llama/Llama-3-8b-chat-hf chatmodel and togethercomputer/m2-bert-80M-8k-retrieval embedding model with embeddingDimension: 768, but I'm having the following error many times when agents are having conversations: 9/30/2024, 5:59:09 PM [CONVEX A(aiTown/agentOperations:agentGenerateMessage)] Uncaught Error: Expected a vector with dimensions 768, received 1024. at async searchMemories (../../convex/agent/memory.ts:164:12) at async startConversationMessage (../../convex/agent/conversation.ts:34:4) at async handler (../../convex/aiTown/agentOperations.ts:74:6) — Reply to this email directly, view it on GitHub <#255>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACZQW3VNDH5KLBEKTSSWRLZZEOXHAVCNFSM6AAAAABPC5JHR2VHI2DSMVQWIX3LMV43ASLTON2WKOZSGU2TMMJWGYYDENA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>