This guide
This guide
-
The guide includes directions and links to the Dell Technologies Generative AI repository on GitHub, along with instructions for building the POC with Jupyter Notebooks, Conda, Miniconda, Python, and other tools. It also contains information about the value of building Generative AI use cases with NVIDIA AI Enterprise.
What is covered:
- RAG process diagram
- Chatbot software stack diagram
- Hardware and software requirements
- Detailed step-by-step instructions to build your POC
- Guidance on running and using the POC
- Prompt engineering tips
Other resources
Table 1. Other resources in the RAG POC kit
RAG POC prompting emails
Topical use case solution briefs
Topical use case “cheat sheets”
RAG POC persona guide
These emails are pre-written, to help you execute a training program using your new POC. They include pre-launch informative emails, launch details, and post-launch exercises for your organization. Email text marked in red [lorem ipsum] will need to be completed by your team.
The solution briefs outline potential RAG POC use cases and how to cluster them for efficiency in your Generative AI practice.
The solution briefs outline potential RAG POC use cases and how to cluster them for efficiency in your Generative AI practice.
This one-page guide details the importance of the different leaders and personas within your organization for achieving success with your RAG POC. It takes a village to find success with Generative AI; use this guide to help guide your onboarding.
Retrieval Augmented Generation (RAG)
Figure 1. Reference diagram of RAG
In figure 1 there are several parts. 1) Knowledge base content is split and encoded into vector database. 2) User prompts the application. 3) LangChain library tools ingest the prompt, analyze it and gather similar chunks of content. 4) Chunks of related content are ranked and then attached to the prompt. 5) Prompt and eligible chunks are fed to the LLM for response.
Chatbot software stack
Figure 2. Representation of a chatbot software stack