vocabulary:
  - term: parse
    label: Parse
    definition: >
      The core LlamaParse operation that converts documents (PDF, DOCX, PPTX, and 130+ formats)
      into structured LLM-ready output such as markdown, text, or JSON. Supports tiered modes
      including fast, cost-effective, agentic, and agentic-plus.
    related:
      - parsing_job
      - result_type
      - parsing_instruction

  - term: parsing_job
    label: Parsing Job
    definition: >
      An asynchronous processing unit created when a file is submitted for parsing. A job
      has a unique ID used to poll for status and retrieve results once processing completes.
    related:
      - parse
      - job_status

  - term: job_status
    label: Job Status
    definition: >
      The current state of an asynchronous job. Common values are PENDING (queued or in progress),
      SUCCESS (completed successfully), ERROR (failed), PARTIAL_SUCCESS (some pages failed),
      and CANCELLED.
    related:
      - parsing_job

  - term: result_type
    label: Result Type
    definition: >
      The output format requested for a parse job. Options include 'markdown', 'text', 'json',
      'structured', and 'pdf'. Determines how the parsed content is returned via the result
      retrieval endpoints.
    related:
      - parse

  - term: parsing_instruction
    label: Parsing Instruction
    definition: >
      A natural language prompt provided by the developer to guide how the document should be
      parsed. Used to focus extraction on specific elements, tables, figures, or layout regions.
    related:
      - parse

  - term: extract
    label: Extract
    definition: >
      The LlamaParse structured data extraction operation that pulls typed JSON output from
      documents using a developer-defined output schema. Composable with Parse to reduce costs
      by re-extracting from previously parsed documents.
    related:
      - extract_job
      - extraction_agent
      - output_schema

  - term: extract_job
    label: Extract Job
    definition: >
      An asynchronous unit of work for structured data extraction from a document.
      Created by submitting a file and an extraction agent; results retrieved via job ID.
    related:
      - extract
      - extraction_agent

  - term: extraction_agent
    label: Extraction Agent
    definition: >
      A configured schema-based extraction definition that specifies what structured data to
      extract from documents. Agents are reusable and can be applied to multiple files.
    related:
      - extract
      - output_schema

  - term: output_schema
    label: Output Schema
    definition: >
      A JSON Schema document that defines the expected structure and types of extracted data
      fields for an Extract or Classify operation.
    related:
      - extract
      - extraction_agent

  - term: classify
    label: Classify
    definition: >
      The LlamaParse document classification operation that categorizes uploaded files by
      document type or content class. Available in fast (1 credit/page) and multimodal
      (2 credits/page) modes.
    related:
      - classify_job
      - classification_schema

  - term: classify_job
    label: Classify Job
    definition: >
      An asynchronous unit of work for document classification. Returns a document class
      label and confidence score once completed.
    related:
      - classify

  - term: split
    label: Split
    definition: >
      The LlamaParse operation that divides a multi-section document into logical segments
      such as chapters, sections, or pages for downstream pipeline routing.
    related:
      - split_job

  - term: split_job
    label: Split Job
    definition: >
      An asynchronous unit of work for document splitting. Returns a list of segment
      boundaries and associated metadata.
    related:
      - split

  - term: pipeline
    label: Pipeline
    definition: >
      A LlamaCloud managed ingestion pipeline that automates the flow of documents through
      parse, extract, embed, and index steps. Pipelines have data sources, data sinks, and
      embedding model configurations.
    related:
      - data_source
      - data_sink
      - pipeline_document

  - term: data_source
    label: Data Source
    definition: >
      A configured connector for a document ingestion pipeline, pointing to an upstream
      storage system such as S3, Google Drive, SharePoint, or a custom webhook.
    related:
      - pipeline

  - term: data_sink
    label: Data Sink
    definition: >
      A configured connector for a pipeline's output, directing indexed or extracted
      results to a downstream vector store or storage system.
    related:
      - pipeline

  - term: index
    label: Index
    definition: >
      A LlamaCloud managed vector index that stores parsed and embedded document chunks
      for semantic retrieval. Supports standard, spreadsheet, and multi-modal indexing modes.
    related:
      - pipeline
      - retriever

  - term: retriever
    label: Retriever
    definition: >
      A LlamaCloud query interface that performs semantic search over an index to return
      relevant document chunks. Supports direct and composite retrieval strategies.
    related:
      - index

  - term: credits
    label: Credits
    definition: >
      The billing unit for LlamaParse operations. Each page processed consumes a fixed number
      of credits depending on the operation tier (e.g., fast parse = 1 credit/page,
      agentic parse = 10 credits/page).
    related:
      - parse
      - classify
      - extract

  - term: premium_mode
    label: Premium Mode
    definition: >
      A parsing configuration flag that activates advanced LLM-assisted parsing for
      higher-fidelity output on complex documents. Consumes more credits than standard mode.
    related:
      - parse
      - parsing_job

  - term: invalidate_cache
    label: Invalidate Cache
    definition: >
      A flag on parse upload requests that forces re-processing of a previously cached
      document, bypassing the stored result and computing a fresh parse.
    related:
      - parse

  - term: batch_job
    label: Batch Job
    definition: >
      A grouped set of parse, extract, or classify operations submitted together for
      asynchronous bulk processing. Supports status monitoring and bulk result retrieval.
    related:
      - parsing_job
      - extract_job
      - classify_job