Aimed to provide solution architects, data engineers, developers, and 'hands on' technical folks with practical advice on implementing and optimizing RAG solutions; This collection of documentation, diagrams, and assets will help you get started with understanding, building, demoing, and depoloying RAG applications.
For those new to RAG, the IBM RAG Reference Architecture contains a conceptual introduction to RAG, IBM's POV on the pattern, relevant IBM products, and common use cases.
Although RAG and Agentic AI have become tightly coupled it has been decided to seperate the two for this cookbook. For more information on Agentic AI and IBM's POV please visit our AI Agent Development Portal.
Guidance on options and choices solution developers must make when creating RAG solutions.
Tips and techniques to accelerate and optimize the processing and ingestion of documents to make them searchable in a RAG solution.
Lessons learned and practical advice for organizing ingested data to optimize the search performance and results quality of the solution.
Advantages and drawbacks of different embedding models and techniques.
Tips and techniques to accelerate and optimize the retrieval of information to support RAG solutions.
Tips and techniques for writing (and re-writing) large language model queries to optimize the speed, relevance, and overall quality of solution results.
Metrics and measurement schemes to quantify, and thus compare and improve, the accuracy of RAG solution results.
Lessons learned on designing, developing, and operating orchestration layers for RAG solutions.
Guidance on choosing and/or developing user interfaces for your RAG solutions.