Better response parsing

[Parth-Jain-2002]

1. Currently, in response parsing, it expects from the LLM to send the message with proper JSON format without any other content. This can sometimes lead to many response failures and multiple API calls to the LLM.

Some response with this structure would fail the parsing

2. In this PR, have modified that logic on how the response is parsed

[Parth-Jain-2002]

@ARajgor Can you check this PR and merge it?

[ARajgor]

make one file called utils or services on /agent folder and have a common function for all the responses. We don't want to write same function to all files