Introduction to NVIDIA Generative AI Examples
The NVIDIA Generative AI Examples repository is designed as a resourceful starting point for developers aiming to incorporate NVIDIA's software ecosystem into their generative AI projects. Whether developers are crafting Retrieval-Augmented Generation (RAG) pipelines, agentic workflows, or fine-tuning AI models, this repository equips them with the necessary tools to seamlessly integrate NVIDIA's robust technology into their development processes.
What's New?
Knowledge Graph RAG
The repository introduces a powerful example focused on creating and querying knowledge graphs through RAG. This implementation uses GPU-acceleration, NIM microservices, and the RAPIDS ecosystem, ensuring efficient processing of large-scale datasets. Developers can explore using NVIDIA AI Foundation Models and Endpoints to enhance their knowledge graph capabilities.
Agentic Workflows with Llama 3.1
Discover how to build an Agentic RAG Pipeline utilizing Llama 3.1 alongside NVIDIA NeMo Retriever NIM microservices. This section includes a comprehensive blog and a hands-on notebook for developers eager to delve into agent-based RAG pipelines. The integration with NVIDIA Morpheus and NIM microservices showcases how to create LLM-based agent workflows.
RAG with Local NIM Deployment and LangChain
For those interested in deploying RAG pipelines, this section provides insightful tips employing NVIDIA AI LangChain AI Endpoints. A detailed blog and a supportive notebook guide developers through the setup and operation of a RAG pipeline with local NIM deployment.
Try it Now!
Getting started with NVIDIA RAG Pipelines is straightforward. Here’s a quick guide:
- Obtain an NVIDIA API Key: Visit the NVIDIA API Catalog to select a model and retrieve your API key.
- Clone the Repository: Use Git to clone the repository to your local machine.
- Build and Run RAG Pipeline: Navigate to the specified directory and use Docker to build and run the basic RAG pipeline.
- Interact with RAG Playground: Access the sample RAG Playground via a browser and experiment with queries.
- Stop the Services: Once finished, properly shut down the running containers.
RAG (Retrieval-Augmented Generation)
RAG Notebooks
NVIDIA provides robust support for popular generative AI frameworks like LangChain, LlamaIndex, and Haystack. The repository includes end-to-end notebooks demonstrating the integration of NIM microservices, accessible through your preferred AI development framework.
LangChain Notebooks
These notebooks cover basic and advanced RAG implementations using LangChain, such as integrating local NIM microservices and interacting with financial reports.
LlamaIndex Notebooks
Explore the integration of RAG with LlamaIndex, which provides a structured approach to implementing basic RAG models.
RAG Examples
The repository offers a variety of examples ranging from basic to advanced RAG applications. Developers can experiment with different data types and configurations, leveraging NVIDIA’s preview NIM endpoints or deploying the examples on local premises.
RAG Tools
Augment your LLM development process with tools tailored to improve productivity and evaluation in NVIDIA RAG pipelines.
Documentation
Resources are available to assist developers at every step of the way, from getting started to customization and runtime configuration. The documentation provides detailed guides, including configuring LLM parameters, supporting multi-turn conversations, and leveraging NVIDIA Riva for speech interaction.
Community
NVIDIA encourages contributions and feedback from the LLM community to enhance the repository’s reach and efficacy. They invite developers to engage through GitHub issues, pull requests, and community examples.
Overall, the NVIDIA Generative AI Examples repository offers a comprehensive suite of tools, examples, and documentation, empowering developers to integrate NVIDIA technologies into their generative AI projects with ease and efficiency.