[BUG] Agents not invoked?

[codenamics]

Describe the bug I have setup agnets like on screenshot. Also the bug is no response form agents like in screenshot

To Reproduce Refer to screenshot

Expected behavior Anser from agents

Screenshots

Flow

Setup

• Installation npx
• Flowise Version latates
• OS: Windows
• Browser Chrome

Additional context

[dentroai]

I get a similar error as can be seen here: https://smith.langchain.com/public/59c442a4-854d-4e5c-9643-46abe64551ca/r

The error occurs when a worker with an attached tool is called. If no tools are attached to any worker the chat runs through.

OS: Ubuntu Deployement: docker compose Flowise: 1.8.1 Browser: Firefox

This bug makes tool usage impossible in Agent workflows.

[toi500]

@codenamics you are connecting the ChatMistralAI node to the Supervisor and the Worker at the same time.

In theory, the Worker node would inherit whatever model the Supervisor has attached, no need to add it again.

Workers have their own model connection in case you want to use a different model than the Supervisor has in use.

Ex:

Supervisor →Mistral Worker→Gemini 1.5

[codenamics]

@toi500 nope still not working

[codenamics]

btw i was watching this tutoral https://www.youtube.com/watch?v=284Z8k7yJRE&ab_channel=LeonvanZyl

and he attachted tools and it worked

@HenryHengZJ can you take a look what is going on?

[bigredcoder]

@HenryHengZJ I am having the same issue. Hosting on Render if that matters.

It seems like the supervisor commands are not being passed to the worker when there are tools.

[toi500]

I am having the same issue. Hosting on Render if that matters.

It seems like the supervisor commands are not being passed to the worker when there are tools.

This issue was submitted a while ago https://github.com/FlowiseAI/Flowise/issues/2472 and was reported as fixed, but it still persists in the latest version 1.8.1

[HenryHengZJ]

strange.. can't seem to replicate the issue. Have FW deployed on Render and Railway, both works fine:

On 1.8.1

Flow used: test.Agents.json

[toi500]

@HenryHengZJ just tested and it works now. I tried your flow and other a bit more complex using the web browser tool and worked too.

For some reason this last evening did not work.

Working right now, using both Workers as expected:

--

[dentroai]

Still doesn't work for me as soon as I add any tool

[dentroai]

What should I do to troubleshoot?

[toi500]

@dentro-innovation Please share ur flow, so I can test it.

[bigredcoder]

I too am still having issues. I have recreated the screenshot @HenryHengZJ shared and I am still getting the same. Attached is what I have tried. I am going to check some logs and see if that can gleen anything.

Here is some information that may be helpful: I am using render to host. I am using a custom domain. I tried the direct render url and got the same result. Going to try to ping via API and embed on a page and see if anything different happens.

[bigredcoder]

I tested by embedding on my server and I do not get any response from the bot. I thought maybe it was a browser issue. I have also tried restarting the service and did a new deploy. Tools test Gitbub example Agents.json

[toi500]

I tested by embedding on my server and I do not get any response from the bot. I thought maybe it was a browser issue. I have also tried restarting the service and did a new deploy. Tools test Gitbub example Agents.json

Working locally for me. Flowise is up to date (today).

[bigredcoder]

#2472

I wonder if its something with the render configuration? Or maybe with OpenAI as they are currently down. I did have the same issues yesterday. Not sure how to troubleshoot this.

[bigredcoder]

When I test with an old GPT model gpt-3.5-turbo-16k-0613 it works... hmm. Still testing more

These are the models ist works on:

gpt-3.5-turbo-16k-0613 gpt-4-turbo-preview gpt-3.5-turbo-16k

This is so strange.

[toi500]

@bigredcoder

On Render the 4o worked too.

Is your instance up to date?

[bigredcoder]

@bigredcoder

On Render the 4o worked too.

Is your instance up to date?

I did a fresh pull and deploy this morning.

[toi500]

@bigredcoder it is so strange. Yesterday I had the same issue you reported and today it is working just fine.

[bigredcoder]

@bigredcoder On Render the 4o worked too. Is your instance up to date?

I did a fresh pull and deploy this morning.

[bigredcoder]

@bigredcoder it is so strange. Yesterday I had the same issue you reported and today it is working just fine.

What I am finding is that some models are working for me, but most are not. This is very bizarre.

[bigredcoder]

When I test with an old GPT model gpt-3.5-turbo-16k-0613 it works... hmm. Still testing more

These are the models ist works on:

gpt-3.5-turbo-16k-0613 gpt-4-turbo-preview gpt-3.5-turbo-16k

This is so strange.

Now those models do not work. I am sure its something with my set up. I appreciate all of the help.

[dentroai]

I am hosting it on an ubuntu server. Deployed using my own tutorial https://www.youtube.com/watch?v=Y-YsPVblUK0

I just created a new Agent Flow and tried out a Mistral Model instead of OpenAI, also doesn't work.

This is my Flow: Debug Agent Flow Agents.json

[avst650]

Hey guys, yesterday I got this error: message stopped. today I got this error. I followed Leon's agent tutorial Any idea what's going on?

[bigredcoder]

for what its worth I see this message when it does not work

[dentroai]

for what its worth I see this message when it does not work

yes me too

[bigredcoder]

Hey guys, yesterday I got this error: message stopped. today I got this error. I followed Leon's agent tutorial Any idea what's going on?

Its strange, when I try different models it sometimes works and then breaks again. I can try 1+1 and it works and then try 2+2 it does not. I think the only reason its working is because its not using the tool to get a response and is using the LLM.

I am on render and followed Leon's tutorials as well. I am stuck.

[avst650]

So frustrating, I got a webinar on LLM agents on sat and now this breaks!

[toi500]

I am hosting it on an ubuntu server. Deployed using my own tutorial https://www.youtube.com/watch?v=Y-YsPVblUK0

I just created a new Agent Flow and tried out a Mistral Model instead of OpenAI, also doesn't work.

This is my Flow: Debug Agent Flow Agents.json

@dentro-innovation just tested your flow using OpenAI (i dont have Mistral) and it worked if you change the name of the worker and supervisor to lowercase, like this:

FLOW: TEST2 Agents.json

[bigredcoder]

I am hosting it on an ubuntu server. Deployed using my own tutorial https://www.youtube.com/watch?v=Y-YsPVblUK0 I just created a new Agent Flow and tried out a Mistral Model instead of OpenAI, also doesn't work. This is my Flow: Debug Agent Flow Agents.json

@dentro-innovation just tested your flow using OpenAI (i dont have Mistral) and it worked if you change the name of the worker and supervisor to lowercase, like this:

FLOW: [TEST2 Agents.json](https://github.com/user-attachments/files/15568635/TEST2.Agents.json)

@toi500 did you get the error when you had a capital letter?