Notifications
Clear all
Topic starter
Since yesterday, my previously functioning AI Agent workflows have ceased executing tools. The problem is now evident in all flows that utilize the AI Agent Node, including those I haven't modified. No tool invocations are occurring; only the OpenAI Chat Model Node is running, while configured tools such as the HTTP Request Nodes or Vector Store are not being triggered.
This behavior persists even though:
- The input clearly aligns with the conditions specified in the prompt.
- The same input previously worked without issues.
- No alterations have been made to the workflow logic.
The agent is simply returning an output without engaging any tools.
What is the error message (if any)?
There is no discernible error message. The AI Agent Node provides partial or empty output, and the section detailing “tools used in this run” remains entirely blank.
Within the execution data:
- The
tool_callsarray is empty. invalid_tool_callsis empty.- No
tool_call_idis being utilized.
Information on your callin.io setup
- callin.io version:
1.94.1 - Database: SQLite (default)
- callin.io EXECUTIONS_PROCESS setting:
own - Running callin.io via: cloud
- Operating system: Windows
Additional Notes
- All tool nodes are correctly registered and connected.
- This issue affects all flows, including those that were previously functional.
- No modifications have been made to the prompt or node configurations.
- The problem began abruptly, likely coinciding with the update to version 1.94.x.
Please provide the rewritten markdown content *it should be in the markdown format.
Posted : 28/05/2025 7:41 am
Hello!
Based on your description, it appears the tool update has caused an issue.
Perhaps try switching to version 1.95.X, even if it's in Beta, to see if the problem is resolved.
Posted : 28/05/2025 9:57 am
Topic starter
Thanks for getting back to me.
Could you clarify what you mean? Are you suggesting downgrading to an older version? If so, how would I do that?
I only recently upgraded because I was hoping it would resolve the issue, but unfortunately, it hasn't improved.
Could you please provide instructions on how to proceed?
Posted : 28/05/2025 2:24 pm
My suggestion is to upgrade to the next beta version to see if it can resolve your issue.
Posted : 28/05/2025 3:45 pm
Could you attempt to rebuild the workflow, or perhaps a test version, within a new workflow?
Posted : 28/05/2025 3:51 pm
Topic starter
I've done the same. As I mentioned, another workflow of mine, which I haven't modified, is experiencing the same problem...
Posted : 28/05/2025 4:11 pm
My apologies. I was considering if generating a new AI agent from the node list might resolve the problem. I've encountered some peculiar issues with the AI agent myself.
I don't believe updating it will resolve the problem, as the changelog doesn't mention anything related to your issue.
However, downgrading might be a viable solution.
Posted : 28/05/2025 4:16 pm
Topic starter
I upgraded due to the issue, as mentioned.
How can I downgrade in the cloud?
Posted : 28/05/2025 4:31 pm
What version were you running yesterday when this problem began? I also just realized that downgrading isn't possible on the cloud version. So, that's the situation...
However, your issue is quite interesting. I haven't encountered anything similar on the forum, nor have I found any open issues on GitHub.
Could you try switching the AI model to a different provider, such as mistral, to see if that makes any difference?
Posted : 28/05/2025 4:46 pm
I've consistently encountered this problem (it began when I started self-hosting on version 1.94.1), and it results in later executions disregarding instructions to utilize tools. I haven't discovered a reliable method to prompt around this behavior.
This issue is occurring with a Postgres node. Has anyone else experienced this?
Posted : 01/06/2025 4:13 pm