vocabulary:
  - term: Agent
    definition: >
      The core PydanticAI class that orchestrates LLM interactions, tool registration,
      structured output validation, and dependency injection for building AI applications.
    type: class
    domain: pydantic-ai

  - term: Tool
    definition: >
      A Python function registered with an Agent that the LLM can call during execution,
      supporting synchronous and asynchronous functions with automatic schema generation.
    type: class
    domain: pydantic-ai

  - term: RunContext
    definition: >
      The context object passed to tools during agent execution, providing access to
      dependencies, usage statistics, prompt, and current message history.
    type: class
    domain: pydantic-ai

  - term: ModelProvider
    definition: >
      An integration adapter connecting PydanticAI to a specific LLM backend such as
      OpenAI, Anthropic, Google Gemini, Mistral, Groq, AWS Bedrock, or Cohere.
    type: class
    domain: pydantic-ai

  - term: StructuredOutput
    definition: >
      A Pydantic model or dataclass used to define and validate the typed return value
      of an agent run, ensuring LLM responses conform to an explicit schema.
    type: class
    domain: pydantic-ai

  - term: DependencyInjection
    definition: >
      A pattern used in PydanticAI to supply runtime dependencies (database connections,
      HTTP clients, configuration) to agents and tools via a typed dependency container.
    type: pattern
    domain: pydantic-ai

  - term: StreamingResponse
    definition: >
      An incremental LLM response delivered token-by-token, supported by PydanticAI
      with both text streaming and structured output streaming.
    type: class
    domain: pydantic-ai

  - term: PydanticGraph
    definition: >
      A subpackage of PydanticAI providing a graph-based execution engine for complex
      multi-step agentic workflows with explicit state management and cycle support.
    type: module
    domain: pydantic-ai

  - term: PydanticEvals
    definition: >
      A subpackage providing systematic evaluation tooling for PydanticAI agents,
      including dataset management, scoring, and LLM-as-judge capabilities.
    type: module
    domain: pydantic-ai

  - term: Logfire
    definition: >
      Pydantic's OpenTelemetry-based observability platform providing conversation
      panels, token tracking, cost monitoring, and SQL-queryable production traces
      for LLM applications built with PydanticAI.
    type: service
    domain: pydantic-logfire

  - term: AIGateway
    definition: >
      A Logfire feature that proxies and routes model provider API calls for
      centralized FinOps tracking, cost attribution, and provider switching.
    type: feature
    domain: pydantic-logfire

  - term: WriteToken
    definition: >
      A Logfire credential granting permission to ingest OpenTelemetry traces and
      spans into a specific Logfire project.
    type: credential
    domain: pydantic-logfire

  - term: ReadToken
    definition: >
      A Logfire credential granting read access to query traces and metrics from
      a specific Logfire project.
    type: credential
    domain: pydantic-logfire

  - term: APIKey
    definition: >
      A Logfire personal or service account credential used to authenticate
      management API requests for administering organizations, projects, and resources.
    type: credential
    domain: pydantic-logfire

  - term: Project
    definition: >
      A Logfire resource namespace that groups related observability data, dashboards,
      alerts, and tokens under a single organizational unit.
    type: resource
    domain: pydantic-logfire

  - term: Dashboard
    definition: >
      A Logfire visualization resource containing a grid layout of panels displaying
      time-series metrics, logs, and LLM traces.
    type: resource
    domain: pydantic-logfire

  - term: Alert
    definition: >
      A Logfire monitoring rule that triggers notifications via configured channels
      (email, webhook, PagerDuty) when metric thresholds are exceeded.
    type: resource
    domain: pydantic-logfire

  - term: AuditLog
    definition: >
      An enterprise Logfire feature providing immutable records of all administrative
      actions (INSERT, UPDATE, DELETE, LOGIN) performed within an organization.
    type: feature
    domain: pydantic-logfire

  - term: SCIM
    definition: >
      System for Cross-domain Identity Management support in Logfire enabling
      automated user and group provisioning from identity providers.
    type: protocol
    domain: pydantic-logfire

  - term: OAuthApp
    definition: >
      A registered Logfire OAuth 2.0 client application supporting Authorization Code
      with PKCE, Device Flow, and Dynamic Client Registration.
    type: resource
    domain: pydantic-logfire

  - term: Region
    definition: >
      A geographic deployment partition of the Logfire API, currently US
      (logfire-us.pydantic.dev) and EU (logfire-eu.pydantic.dev), providing
      data residency options.
    type: concept
    domain: pydantic-logfire