Adding post-generation guardrails: RARR

[opertifelipe]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

Addinf a guardrails class RARR (Researching and Revising What Language Models Say, Using Language Models) integrated with the other LangChain components that can be used for virtual assistants and RAG solutions to detect and reduce hallucinations.

Motivation

I believe it is important to start adding post-generation guardrails inside LangChain. Always more, it is requested an high reliability of the answers virtual assistants and invest in developing tools helping to reduce hallucination or detect hallucination could be a valuable feature.
RARR (Researching and Revising What Language Models Say, Using Language Models) from https://arxiv.org/pdf/2210.08726 could be one of the valuable option to include.

Proposal (If applicable)

No response