[FEATURE] New Component For Remote Custom Model with Inference Endpoint, Schema (input & Output) and Target Field

[DebajitKumarPhukan]

Describe the feature you'd like Let's say I have my custom model hosted at some remote server having an inference API endpoint. The inference API endpoint takes an input in a particular JSON format and output back in a particular JSON format. I should have a "Generic/Universal" component which I can drag & drop and specify the below:

1. Inference endpoint of the custom model (hosted in some remote server) to trigger.
2. Specify the schema of the input payload to the inference point and have the ability to map the input from the user to the field(s) in the schema.
3. Specify headers (in some cases for API keys).
4. Specify the output schema and have the ability to map field to choose for the output to the user.

Additional context The idea here is to have a generic component having the ability to get inference from any server hosting custom model instead of limiting to Open AI, Hugging Face, Azure, Replicate, etc.

[michele-fadda]

and of course being able to use whatever model, local or remote.

[the-gates-of-Zion]

Yes I would like this feature too. Thank you