Demo: RAG/docling/llama-index service with an Instructlab frontend

[nerdalert]

RAG Demo:

• Adds RAG APIs to docling-serve to create a RAG service.
• llama-index for categorical document collections.
• Instructlab quantized model is used for generations.
• Two collections are created, one via URL and one via file
  uploading to via the Instructlab UI.
• Collections are then queried and returned with the Answer
  to the query along with the sources and metadata from the
  vector DB.
• Do a negative test to ensure if the document does not
  contain information matching the query it does not
  hallucinate an answer.

** All components are running locally on a MAC M1.
The PDF ingestions are fast forwarded as they take about 60s each.**

Backend RAG service code will be posted to the docling repo and link here.

Demo video:

rag-demo-v2-oct21.mp4

[nerdalert]

The backend code is posted here DS4SD/docling-serve#9