vocabulary:
  domain: Amazon Bedrock
  description: >-
    Taxonomy and vocabulary for Amazon Bedrock covering generative AI,
    foundation models, RAG, agents, guardrails, and responsible AI concepts.

  concepts:
    - term: Foundation Model
      definition: >-
        A large AI model pre-trained on vast amounts of data that can be
        adapted for a wide range of tasks including text generation,
        summarization, classification, and code generation.

    - term: Inference
      definition: >-
        The process of using a trained foundation model to generate outputs
        (text, images, embeddings) from input prompts or data.

    - term: Prompt
      definition: >-
        The input text or instructions provided to a foundation model to
        guide its response generation.

    - term: Token
      definition: >-
        The basic unit of text processed by language models. A token is
        approximately 4 characters or 0.75 words in English.

    - term: Temperature
      definition: >-
        An inference parameter controlling randomness in model outputs.
        Lower values produce more deterministic outputs; higher values
        produce more creative and varied outputs.

    - term: Top-P (Nucleus Sampling)
      definition: >-
        An inference parameter that limits token selection to the smallest
        set of tokens whose cumulative probability exceeds the specified value.

    - term: Embedding
      definition: >-
        A dense numerical vector representation of text or other data
        that captures semantic meaning, used for similarity search in RAG.

    - term: RAG
      label: Retrieval-Augmented Generation
      definition: >-
        A technique combining retrieval from a knowledge base with generative
        AI to produce more accurate, grounded, and up-to-date responses.

    - term: Knowledge Base
      definition: >-
        A collection of documents stored as vector embeddings in a vector
        database, enabling semantic search for RAG applications.

    - term: Vector Store
      definition: >-
        A database optimized for storing and searching high-dimensional
        embedding vectors using approximate nearest neighbor algorithms.

    - term: Guardrail
      definition: >-
        Content filtering and safety policy that can block topics, filter
        harmful content, redact PII, and detect prompt injection in Bedrock.

    - term: Agent
      definition: >-
        An autonomous AI system that orchestrates multi-step tasks by
        reasoning over user requests, calling APIs (action groups), and
        querying knowledge bases.

    - term: Action Group
      definition: >-
        A set of API actions defined by an OpenAPI schema that a Bedrock
        Agent can call during orchestration.

    - term: Fine-Tuning
      definition: >-
        Training a pre-trained foundation model on domain-specific data
        to improve performance on targeted tasks.

    - term: Continued Pre-Training
      definition: >-
        Continuing foundation model training on unlabeled domain data to
        improve model knowledge of a specific domain without task labels.

    - term: Provisioned Throughput
      definition: >-
        Reserved inference capacity purchased in advance for consistent
        model performance and predictable latency at production scale.

    - term: Streaming Inference
      definition: >-
        A mode where model output tokens are returned progressively as
        they are generated, enabling faster time-to-first-token for UX.

    - term: Converse API
      definition: >-
        A unified Bedrock Runtime API that provides a consistent chat interface
        across multiple foundation model providers.

    - term: Cross-Region Inference
      definition: >-
        Automatic routing of inference requests across AWS regions for
        improved availability and throughput during peak demand.

  model-providers:
    - Anthropic (Claude)
    - Amazon (Titan, Nova)
    - Meta (Llama)
    - Mistral AI
    - Cohere
    - AI21 Labs
    - Stability AI
    - Writer
    - DeepSeek

  tags:
    - Generative AI
    - Foundation Models
    - AI
    - Machine Learning
    - NLP
    - Large Language Models
    - LLM
    - RAG
    - Agents
    - Responsible AI
    - AWS
    - Amazon Web Services
    - Serverless