EAISCRUM-1524-design-llm-for-image-feature-explicit-content-and-logo-detection

[Daggx]

No description provided.

[KyrianC]

I think it would be cool to use Langchain for this, with structred_outputs as we already have all the pydantic models ready, we would just add description to each fields.

only problem is maybe there would be dependency conflicts between pydantic v1 & v2.
What do you think?

[Daggx]

I think it would be cool to use Langchain for this, with structred_outputs as we already have all the pydantic models ready, we would just add description to each fields.

only problem is maybe there would be dependency conflicts between pydantic v1 & v2. What do you think?

That's what i was going to use for the document parsing part since the financial/invoice/receipt dataclass are pretty huge, here since they are pretty simple l(ogo & explicit content) i didn't think it would add anything, but we could do it for them also.

[KyrianC]

I think it would be cool to use Langchain for this, with structred_outputs as we already have all the pydantic models ready, we would just add description to each fields.
only problem is maybe there would be dependency conflicts between pydantic v1 & v2. What do you think?

That's what i was going to use for the document parsing part since the financial/invoice/receipt dataclass are pretty huge, here since they are pretty simple l(ogo & explicit content) i didn't think it would add anything, but we could do it for them also.

Okay! we can let it like that, and if ever there is errors, like the model not returning the right output, we can try with langchain if it performs better