Grounding layer

As explained in Generative AI in Action

A grounding layer is an application-architecture layer that acquires, prepares, represents, stores, and delivers external data for use by generative models.

The grounding layer acquires, stores, processes, represents, and delivers enterprise data for generative-AI applications. It connects source systems to LLM-compatible data, especially for retrieval-augmented generation (RAG). section 10.4 section 10.7

A grounding layer typically combines:

• Connectors, APIs, and plugins that handle different record systems, formats, protocols, authentication requirements, and data-refresh settings. section 10.2.3 section 10.4
• Batch and streaming data pipelines designed for high throughput and low delay. section 10.2.3
• Preprocessing, embedding models, and vector databases that make source data usable by LLMs. section 10.4
• Distributed processing frameworks that improve scalability and fault tolerance. section 10.2.3

RAG data path

Connect a system of record through a native API or plugin, ingest data through a batch or streaming pipeline, preprocess it for LLM use, create embeddings, and store the resulting vectors in a database such as Redis, Cosmos DB, Pinecone, Milvus, or Qdrant. section 10.2.3 section 10.4

Figure 10.4 GenAI app stack

Use the GenAI app stack diagram to locate the grounding layer alongside the model, orchestration, and response-filtering layers. section 10.2.3

How Generative AI in Action covers it

How deep it goes

Explains connectors, pipelines, preprocessing, embeddings, vector databases, and distributed processing at an architecture level.

Examples

Uses APIs, plugins, Redis, Cosmos DB, Pinecone, Milvus, Qdrant, Apache Spark, and Azure Fabric as implementation examples.

What you can do after

Build it from scratch by designing source connectors, data pipelines, preprocessing, embedding, vector storage, and scalable processing.

How it connects

Connects RAG, LLMs, embeddings, vector databases, orchestration, model, and response-filtering layers.

liveBookmanning.com

Read Generative AI in Action on liveBook Get the book at manning.com

Related concepts

• generative ai application stackpart of · A generative AI application stack is a layered architecture of models, services, infrastructure, interfaces, and operational components that supports generative AI applications.
• model architecturepart of · The structural organization of a computational system’s components, relationships, and operations that determines how it processes information or behaves.
• embeddinguses · A numerical vector representation of an item or input that encodes features or relationships useful for computational tasks.
• contractuses · An API contract is a specification of the operations, data, and behavioral expectations that govern interactions between software components.
• vectorizationuses · Vectorization is the representation of data or computations as vectors or arrays so they can be processed numerically or collectively.

This explainer was generated by Manning's knowledge graph from the text of Generative AI in Action. The section links point to the parts of the book it draws on.