Amazon Bedrock Knowledge Bases with Retrieval Augmented Generation (RAG) is the appropriate architecture because it separates frequently changing enterprise information from the foundation model ' s learned parameters. Instead of retraining or fine-tuning the FM every time documentation changes, the application retrieves relevant information from the current knowledge source at query time and supplies that information as context to the model.
AWS specifically states that adding a knowledge base improves cost-effectiveness by “removing the need to continually train your model to be able to use your private data.” This directly satisfies the requirement to avoid recurring model-training costs.
With Amazon Bedrock Knowledge Bases, the organization connects supported data sources, ingests and indexes the documents, and retrieves relevant chunks when a user asks a question. RAG then combines that retrieved content with the user ' s query so that the FM can generate an answer grounded in company information. AWS describes RAG as using information from data sources to improve the relevance and accuracy of generated responses.
When source documents change, the knowledge base can be synchronized or re-ingested so the retrieval layer reflects the updated content. This is substantially more appropriate than fine-tuning the model after every documentation change.
Option B would unnecessarily introduce repeated customization jobs, cost, operational overhead, and model-version management. Option C would become stale because static examples would continue referencing earlier documentation. Option D is especially unsuitable because cached answers can become outdated and would not verify them against newly updated source material.
RAG is specifically designed for scenarios where models require access to current, proprietary, or rapidly changing information without embedding that knowledge permanently into model weights.
Therefore, A. Amazon Bedrock Knowledge Bases with RAG is the correct solution.
===========
