In Memory Vector Store not working [BUG]

[mikkelkrogsholm]

Seems like the new docker image is messing up the vector store.

[toi500]

I have noticed this too, the In Memory Vector Store not working when you try to upsert from the Plain Text node.

[SoundMindsAI]

I have seen this too. I upserted a simple PDF and the In-Memory Vector store. The error I am getting is: Received tool input did not match expected schema

Error: Received tool input did not match expected schema
    at DynamicStructuredTool.call (/usr/local/lib/node_modules/flowise/node_modules/@langchain/core/dist/tools.cjs:70:19)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async /usr/local/lib/node_modules/flowise/node_modules/flowise-components/dist/src/agents.js:315:39
    at async Promise.all (index 0)
    at async AgentExecutor._call (/usr/local/lib/node_modules/flowise/node_modules/flowise-components/dist/src/agents.js:302:30)
    at async AgentExecutor.invoke (/usr/local/lib/node_modules/flowise/node_modules/langchain/dist/chains/base.cjs:58:28)
    at async ConversationalAgent_Agents.run (/usr/local/lib/node_modules/flowise/node_modules/flowise-components/dist/nodes/agents/ConversationalAgent/ConversationalAgent.js:119:19)
    at async utilBuildChatflow (/usr/local/lib/node_modules/flowise/dist/utils/buildChatflow.js:251:15)
    at async createInternalPrediction (/usr/local/lib/node_modules/flowise/dist/controllers/internal-predictions/index.js:7:29)

[SoundMindsAI]

The issue might have something to do with my attempt to use a ConversationAgent. What I thought would be a better Agent appears to be deprecated "Conversational Retrieval Agent". I thought that perhaps the Newer "Tool Agent" might be what I need; however, I am unsure about what to use for the AgentExecutor

[SoundMindsAI]

Actually... The AgentExecutor is not required. This is working for me. The problem was that I was using the ConversationalAgent but the ToolAgent works better.

[SoundMindsAI]

Actually... The AgentExecutor is not required. This is working for me. The problem was that I was using the ConversationalAgent but the ToolAgent works better.

I forgot to mention... I had previously switched over to a Pinecone DB rather than the in-memory Vector Store and was still getting this error message until I switched out the ConversationalAgent for the ToolAgent. I have not tested this with the In-Memory Vector Store.

[toi500]

I think the team, following the LangChain initiative, is trying to standardize the tools. Some of these tools may become outdated since they were built just for OpenAI.

Now, in theory, the tools should work for Gemini, Mistral, Clause, and OpenAI, but this is not always the case.

[SoundMindsAI]

@mikkelkrogsholm and @toi500 Can you confirm that you are using the ToolAgent and not using the ConversationalAgent? I am curious if this might be the fix for the issue that you are seeing.

[mikkelkrogsholm]

@SoundMindsAI I was using the conversational qa chain

[drndos]

Same issue here, seems like reverting to 1.7.2 helped

[AlehandroL]

Same issue here. I use: Tool Agent -> Retriever Tool -> In Memory Vector Store

In the chat window and in Langsmith I can see that output is returning empty. The issue started before updating from 1.7.2 to 1.8.0

[trunghuynh228]

I have the same issue when using OpenAI Assistant. The chatbot cannot get the information from in-memory vector store at all. 1.8.0 is not working and I had to revert to 1.7.2

[automaton82]

Hi @HenryHengZJ can you have this one (and https://github.com/FlowiseAI/Flowise/discussions/2494 which I think is the same) looked at? I think something broke in 1.8.0 with RAG.

[SoundMindsAI]

Where can we find information on the RoadMap? Would it be appropriate to include a Roadmap.md file?

Being new to Flowise, I’m not sure if this bug is something that should be fixed or if there is a RoadMap direction that would mean that this should never be fixed and that the preferred approach is X.

Do you have any documentation describing what should eventually be consolidated based upon the direction of LaingChain? I’m not even sure what consolidation Langchain is working on anyway. If someone wanted to explain that would be great. If someone does explain this might be a great GitHub Discussion.

Thanks for your help! Eric

On Tue, May 28, 2024 at 11:56 PM automaton82 @.***> wrote:

Hi @HenryHengZJ https://github.com/HenryHengZJ can you have this one (and #2494 https://github.com/FlowiseAI/Flowise/discussions/2494 which I think is the same) looked at? I think something broke in 1.8.0 with RAG.

— Reply to this email directly, view it on GitHub https://github.com/FlowiseAI/Flowise/issues/2501#issuecomment-2136466776, or unsubscribe https://github.com/notifications/unsubscribe-auth/BIBCBWA32KN3NKQTCGNLORLZEVGYDAVCNFSM6AAAAABIMS45GCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMZWGQ3DMNZXGY . You are receiving this because you were mentioned.Message ID: @.***>

[tiago-falves]

The In-Memory vector store is also not working for me! I Have this workflow, I saved, upserted and then started a new chat. The model does not recognize any information I gave it in the pdf file. What can I do to fix this? Thank you very much!

[HenryHengZJ]

Will be pushing a new release with this fix soon!

[HenryHengZJ]

Pushed to 1.8.1!

[automaton82]

Thanks!!