[Issue] embeddings model needs to be provided when using Azure API
axen2u opened this issue · comments
I'm getting following message when deployed zep image using docker-compose and failed the zep container creation.
embeddings model needs to be provided when using Azure API
Do we have to use an embedding model when using Azure? Can I just use Azure Open AI for the summarizing? I have tried using Azure Embedding Model as well. When using the embedding model my bot responses get very slow.
Zep embeds all chat conversations so that they may be recalled via semantic search. Are you using Azure's OpenAI embedding service or Zep's local embedding service? More about Zep's embedding configuration here: https://docs.getzep.com/deployment/embeddings/
I want to use long-term memory with summarization. Because my message window is about 5 messages since this a real-time rag I want to summarize the last few messages to handle the flow. So I don't need to use embeddings. However I want to use Azure Open AI models for summarization. It won't let me disable the embedding when using Azure Open AI endpoint and LLM, the following message is coming :
embeddings model needs to be provided when using Azure API
@axen2u Were you able to resolve this? I'm facing a similar issue.
@TheAdamGalloway I have used Azure OpenAI embedding models. You can provide deployment name there.
@axen2u Are you using environment variables? Would you mind sharing which ones I need to set? I can't seem to get past this error
@TheAdamGalloway You can do that in config.yaml
@axen2u I don't think I can modify config.yaml since I'm hosting a container of the zep image in Azure. I've tried setting the corresponding environment variables but it doesn't seem to be working.
@TheAdamGalloway In that case I think you should provide .env when starting the container or you should compose image from the ground and use it. I haven't much experience with containers But I'm happy to help. I will let you know If I find .env configuration for that.
We've merged a fix for this in #346 and released in v0.26.0