[pydantic ValidationError:]

[Kree-max]

Hi! Thanks for developing this tool.
I've been trying to run the demo and encountered some problems with "pydantic validaton" when using the srx-info agent , find-dataset agent and tissue-ontology agent.

error:_pydantic_core.pydantic_core.ValidationError: 1 validation error for UBERON_IDS
Invalid JSON: expected value at line 3 column 1 [type=json_invalid, input_value='\n\n\boxed{UBERON:00021...\boxed{UBERON:0002113}', input_type=str]
For further information visit https://errors.pydantic.dev/2.11/v/json_invalid
During task with name 'generate_structured_response' and id 'fb6a2cd7-7dfe-73ea-9e8d-0bf1ea601fab'

I will be very grateful if you can offer me some advice on solving this problem, thanks a lot!

[nick-youngblut]

What model(s) are you using? Can you share your settings file?

[Kree-max]

Thanks for reply! Since gpt is not available in our region, I changed the model into deepseek. The entrez agent and sra-gent work out fine, the problem occurs when using the srx-info agent , find-dataset agent and tissue-ontology agent, with the "type=json_ invalid"_ issue.(since the .yml file is not supported to upload, I put it in the .txt file)

BTW:If it doesn't work, can I use sra-gent to replace the functions of srx-info agent and find-dataset agent ?
settings.txt

[nick-youngblut]

I'm probably hard-coding a couple of models accidentally. I'll have a look

[nick-youngblut]

I don't have access to deepseek, but I've added retries of structured output calls to this new branch: https://github.com/ArcInstitute/SRAgent/tree/issue9

Hopefully, that fixes your issue. Let me know.

[Kree-max]

Thanks！I‘ll have a try

[Kree-max]

problems solved, the key solution is to replace langchain ChatOpenai with langchain-deepseek ChatDeepSeek, thanks!

[r-sayar]

@Kree-max issue can be closed right?