Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

AI agent tool randomly (but consistently) launching subflow with prior/duplicate $fromAI data which DOESN'T MATCH the current $fromAI INPUT #13453

Open
thethrowdown opened this issue Feb 24, 2025 · 2 comments
Labels
in linear Issue or PR has been created in Linear for internal review

Comments

@thethrowdown
Copy link

thethrowdown commented Feb 24, 2025

Describe the problem/error/question

Re-writing this to further help clarify behavior seen.

AI Tool to store variables is being used.

I send in a chat: store variable $one as ok.
Agent detects variable_name = $one and value = ok
Subflow executes correctly with the variables as listed above.

I send in chat: store variable $two as hi.
Agent detects variable_name =$two and value = hi. The ai agent bot detects the request correctly and these variable values are clearly listed as the inputs detected and fed into $fromAI.
Subflow executes 2x, both times ignoring the inputs listed and sending in variable_name = $one and value = ok (from the original request).

This is a major problem with AI chatbots that will effect anytime you call a tool multiple times. This isn't an AI problem as the INPUTs coming in from the chat agent are showing correct. I have now re-created this dozens of times across various different workflows.

What is the error message (if any)?

N/A

Please share your workflow/screenshots/recording

https://trustguard.app.n8n.cloud/workflow/ghnSz4Oyw9yqmzqk/executions/929
Image

(Select the nodes on your canvas and use the keyboard shortcuts CMD+C/CTRL+C and CMD+V/CTRL+V to copy and paste the workflow.)
⚠️ WARNING ⚠️ If you have sensitive data in your workflow (like API keys), please remove it before sharing.

Share the output returned by the last node

Debug info

core

  • n8nVersion: 1.79.3
  • platform: docker (cloud)
  • nodeJsVersion: 20.18.2
  • database: sqlite
  • executionMode: regular
  • concurrency: 5
  • license: community

storage

  • success: all
  • error: all
  • progress: false
  • manual: true
  • binaryMode: filesystem

pruning

  • enabled: true
  • maxAge: 168 hours
  • maxCount: 2500 executions

client

  • userAgent: mozilla/5.0 (macintosh; intel mac os x 10_15_7) applewebkit/537.36 (khtml, like gecko) chrome/133.0.0.0 safari/537.36
  • isTouchDevice: false

Generated at: 2025-02-23T23:36:59.260Z}

@Joffcom
Copy link
Member

Joffcom commented Feb 24, 2025

Hey @thethrowdown,

We have created an internal ticket to look into this which we will be tracking as "GHC-937"

@Joffcom Joffcom added the in linear Issue or PR has been created in Linear for internal review label Feb 24, 2025
@thethrowdown thethrowdown changed the title AI agent tool processing 2nd, 3rd, and 4th requests as the same data as the 1st AI agent tool randomly (but consistently) launching subflow with prior/duplicate $fromAI data which DOESN'T MATCH the current $fromAI INPUT Feb 25, 2025
@thethrowdown
Copy link
Author

thethrowdown commented Feb 25, 2025

Adding note to further clarify.

AI Tool to store variables is being used.

I send in a chat: store variable $one as ok.
Agent detects variable_name = $one and value = ok
Subflow executes correctly with the variables as listed above.

I send in chat: store variable $two as hi.
Agent detects variable_name =$two and value = hi. The ai agent bot detects the request correctly and these variable values are clearly listed as the inputs detected and fed into $fromAI.
Subflow executes 2x, both times ignoring the inputs listed and sending in variable_name = $one and value = ok (from the original request request).

This is a major problem with AI chatbots that will effect anytime you call a tool multiple times. This isn't an AI problem as the INPUTs coming in from the chat agent are showing correct. I have now re-created this dozens of times across various different workflows.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
in linear Issue or PR has been created in Linear for internal review
Projects
None yet
Development

No branches or pull requests

2 participants