Suppport for RAG with Hugging Face LLM Models using Connector. Post and Pre Process Function Support.

[ramda1234786]

Is your feature request related to a problem?
Need to do RAG with Hugging Face Models. All the Hugging Face models gives response in the below format and post_process_function does make it into JSON object. Like we have out of the box support for Cohere, Amazon and OpenAI. Need support for Hugging Face Models (LLM Models and not Embedding).

{
    "inference_results": [
        {
            "output": [
                {
                    "name": "response",
                    "dataAsMap": {
                        "response": [
                            {
                                "generated_text": "Generated text here...................."
                            }
                        ]
                    }
                }
            ],
            "status_code": 200
        }
    ]
}

What solution would you like?
Like we have out of the box support for Cohere, Amazon and OpenAI. Need support for Hugging Face Models (LLM Models and not Embedding).
Should be able to perform post_process_function and pre_process_function

What alternatives have you considered?
Tried doing post_process_function but it never works

Do you have any additional context?
Tried below series of steps

"post_process_function": "\n def json = \"{\" +\n \"\\\"completion\\\":\\\"\" + params['response'][0].generated_text + \"\\\" \"}\";\n return json;\n "

Tried with painless script as well

POST /_scripts/painless/_execute

{
    "script": {
        "source": "\n      def json = \"{\" +\n                 \"\\\"completion\\\":\\\"\" + params['response'][0].generated_text + \"\\\" }\";\n      return json;\n    ",
        "params": {
            "response": [
                {
                    "generated_text": "Your text here"
                }
            ]
        }
    }
}

I got this

{
    "result": "{\"completion\":\"Your text here\" }"
}

After Connector model, it did not worked i still get this error :
"reason": "Cannot invoke \"Object.getClass()\" because \"callArgs[0]\" is null"

Expected result is this

{
    "result": {"completion":"Your text here" }
}

Much better if we can have a API to test, the post_process_function other than Painless script test/execution

[ramda1234786]

@b4sjoo , this is working now and we can close this request, by using post_process_function we can achieve this.
One working way is here

We can close this request