FlowiseAI / Flowise

Drag & drop UI to build your customized LLM flow
https://flowiseai.com
Apache License 2.0
27.4k stars 14.17k forks source link

[BUG] - [inputText] is not permitted when upgrading to Claude 3 and cohere models #2060

Open haimco50 opened 3 months ago

haimco50 commented 3 months ago

Upgrading bedrock model from claude 2 to claude 3 and embedding model from amazon.titan-embed-text-v1 to cohere.embed-multilingual-v3 and started getting the bellow errror when tryting to upsert embeddings.

image

issue seeems to be related to this issue : https://github.com/langchain-ai/langchain/pull/18630

and may need upgradiung the conversation retrieval QA chain node to support this but not sure...

image

To Reproduce trying to upserts / chat after upgrading will cause a validation error to apear.

  1. Go to '...'
  2. Click on '....'
  3. Scroll down to '....'
  4. See error

Expected behavior A clear and concise description of what you expected to happen.

Screenshots If applicable, add screenshots to help explain your problem.

Flow If applicable, add exported flow in order to help replicating the problem.

Setup

Additional context Add any other context about the problem here.

haimco50 commented 3 months ago

I changed the embedding model to amazon.titan-embed-g1-text-02 (as it's shown in my flow) so it can work - the issue is happenning with cohere.embed-multilingual-v3 model

HenryHengZJ commented 3 months ago

I changed the embedding model to amazon.titan-embed-g1-text-02 (as it's shown in my flow) so it can work - the issue is happenning with cohere.embed-multilingual-v3 model

thanks for pinning down the issue, can u try to see if upgrading langchain community package solves the issue

haimco50 commented 3 months ago

Already did so, this is the first thing I did - it didn't work - IMHO the issue seems to be with the Flowise AWS BedrockChat componnent (something with the prompt setup) - see this thread issue - https://github.com/langchain-ai/langchain/issues/18514

HenryHengZJ commented 3 months ago

found out the issue, because we are using body format that is only compatible with amazon titan embeddings - https://github.com/FlowiseAI/Flowise/blob/main/packages/components/nodes/embeddings/AWSBedrockEmbedding/AWSBedrockEmbedding.ts#L150

cohere embeddings require another type of body format