RAG & Retrieval

Versatility of RAG Applications

Three families of RAG applications, from reading comprehension to code generation to conversational data access.

Key takeaways

  • Building a RAG application means constructing a RAG pipeline, and those pipelines have evolved into different iterations tailored to specific use cases.
  • The post identifies three primary categories of RAG applications: RAG for Reading Comprehension, RAG for Code Generation, and RAG for Conversational Data Interaction.
  • RAG for Reading Comprehension helps people work faster — frontline service advisors, customer support agents, retail store employees, and account executives using natural-language search to find relevant content.
  • RAG for Code Generation augments developers across programming languages, with co-pilot examples like Microsoft Co-pilot, AWS/MongoDB Co-pilot, and Couchbase Capella IQ.
  • RAG for Conversational Data Interaction uses Natural Language Query (NLQ) against datastores — relational databases, data warehouses, and data lakes — returning tabular or natural-language answers with the corresponding SQL query.
On this pagesections

    Crafting a RAG application involves constructing a RAG pipeline, which has evolved into various iterations tailored to specific use cases.

    Presently, we observe three primary categories within the realm of RAG applications:

    1. RAG for Reading Comprehension:

    Revolves around empowering humans to streamline their tasks efficiently. Key scenarios include:

    • Frontline service advisors streamlining customer queries by swiftly accessing relevant information with RAG’s assistance.
    • Customer support agents navigating through intricate manuals, often resorting to multiple sources, to effectively address customer inquiries.
    • Retail store employees guiding customers through an array of service options, aiding in decision-making and customization.
    • Content discovery facilitated by account executives utilizing natural language search tools to access pertinent product collateral, spanning whitepapers, blogs, wikis, videos, and more.

    2. RAG for Code Generation:

    Focuses on augmenting developers’ coding endeavors across diverse programming languages. Notable examples include:

    • Microsoft’s Co-pilot, a widely recognized tool supporting developers in coding tasks.
    • AWS/MongoDB’s Co-pilot, proficient in generating code for manipulating MongoDB data.
    • Couchbase Capella IQ, another prominent Co-pilot, enhancing coding experiences within its ecosystem.

    3. RAG for Conversational Data Interaction:

    Showcases the utilization of Natural Language Query (NLQ) to interact with datastores, such as:

    • Relational databases like PostGres, MySQL, and SQL Server.
    • Data warehouses including Snowflake and BigQuery.
    • Data lakes such as S3, GCS, and Azure Blob Storage.

    In this scenario, the application interprets natural language queries from users and provides responses in either tabular data format or through natural language, accompanied by the corresponding SQL query.

    See what your engineers can ship in production

    Dataworkz is the agentic AI platform enterprises trust to put AI agents into production. Teams build production-ready agents in days, every decision stays traceable, and every agent has ROI attached.

    Explore the platform