Custom Agent doesnt work in Agentic RAG Pipeline

[AnasShahid]

Bug Description

When 'Custom' Language Model is selected and connected to a 'Google Generative AI' Language model, the tool calling (and the agent) does not work.
Note that when the LLM is changed to something like 'OpenAI' it works fine.

Reproduction

Attached is the flow that is giving error. It goes into infinite loop.
This error can also be reproduced if we use 'Agentic RAG' flow from the store and change the LLM to Custom

Expected behavior

Custom LLMs and other language models that are not listed in Agent's dropdown should work.

Who can help?

No response

Operating System

Mac

Langflow Version

1.1.1

Python Version

3.12

Screenshot

Flow File

MatchMinds.AI.json

[dosubot]

Hey @AnasShahid! I'm here to help you with any bugs, questions, or contributions. Let's work together to solve the issue you're facing.

The issue you're experiencing with the 'Custom' Language Model in the 'Agentic RAG' pipeline, where it goes into an infinite loop when connected to 'Google Generative AI', might be related to specific configurations or conditions that are not compatible with the 'Custom' model. While the exact cause isn't detailed in the available information, here are a few steps you can consider to troubleshoot and potentially resolve the issue:

1. Check for Cycles: Ensure that there are no unintended cycles in your workflow that could lead to infinite loops. The Langflow system has mechanisms to detect and handle cycles, as seen in the test cases for cyclicity ^[1].

2. Configuration Parameters: Review the configuration parameters for both the 'Custom' Language Model and 'Google Generative AI'. Ensure that all required inputs, such as API keys and model names, are correctly set up ^[2].

3. Error Handling: Implement error handling to catch any exceptions that might occur during the execution of the 'Custom' model. This can help identify specific points of failure.

4. Compatibility: Verify that the 'Custom' model is compatible with the tools and agents used in your workflow. The 'Custom' model is designed to be flexible, but there might be specific requirements or limitations when integrating with certain models like 'Google Generative AI' ^[3].

5. Logs and Debugging: Check the logs for any error messages or warnings that could provide more insight into why the infinite loop occurs. This might help pinpoint the exact step or component causing the issue.

If these steps do not resolve the issue, you might need to further investigate the specific interactions between the 'Custom' model and the 'Google Generative AI' within your workflow.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}