aid: unstructured-vocabulary
name: Unstructured API Vocabulary
description: Domain terms, concepts, and definitions used across the Unstructured
  Platform API and Partition API.
url: https://raw.githubusercontent.com/api-evangelist/unstructured/refs/heads/main/vocabulary/unstructured-vocabulary.yml
tags:
- document-processing
- RAG
- LLM
- ETL
- chunking
- embeddings
- OCR
- PDF
terms:
- term: source_connector
  label: Source Connector
  definition: A configured connection to an input data system (e.g. S3, SharePoint,
    Dropbox, Confluence) from which documents are ingested for processing.
  tags:
  - connector
  - ingestion
  - ETL
- term: destination_connector
  label: Destination Connector
  definition: A configured connection to an output data system (e.g. vector databases,
    S3, Databricks) where processed document chunks are written.
  tags:
  - connector
  - output
  - ETL
  - vector-database
- term: workflow
  label: Workflow
  definition: A named pipeline configuration that links a source connector to a destination
    connector with processing nodes (partitioning, chunking, embedding) and scheduling
    options.
  tags:
  - pipeline
  - orchestration
- term: job
  label: Job
  definition: A single execution run of a workflow. Each job tracks status, progress,
    and results including any failed files.
  tags:
  - execution
  - run
  - monitoring
- term: partitioning
  label: Partitioning
  definition: The process of splitting a document into its constituent elements (titles,
    narrative text, tables, images) with associated metadata.
  tags:
  - document-processing
  - parsing
- term: chunking
  label: Chunking
  definition: The process of grouping document elements into fixed-size or semantically
    coherent text chunks suitable for LLM context windows and RAG pipelines.
  tags:
  - RAG
  - LLM
  - text-processing
- term: embedding
  label: Embedding
  definition: The conversion of text chunks into vector representations using an embedding
    model, enabling semantic search and retrieval.
  tags:
  - RAG
  - vector
  - semantic-search
- term: hi_res_strategy
  label: Hi-Res Processing Strategy
  definition: A document processing mode that uses OCR and layout detection models
    for higher accuracy on complex PDFs and scanned documents, at increased compute
    cost.
  tags:
  - OCR
  - PDF
  - processing-strategy
- term: fast_strategy
  label: Fast Processing Strategy
  definition: A document processing mode optimized for speed using text extraction
    only, suitable for digitally-native PDFs and text-based files.
  tags:
  - PDF
  - processing-strategy
  - performance
- term: element
  label: Document Element
  definition: A discrete unit extracted from a document, such as Title, NarrativeText,
    Table, Image, or ListItem, along with metadata like page number and coordinates.
  tags:
  - document-processing
  - output
- term: template
  label: Workflow Template
  definition: A reusable workflow configuration blueprint that can be instantiated
    to create multiple similar workflows.
  tags:
  - workflow
  - configuration
- term: notification_channel
  label: Notification Channel
  definition: A configured endpoint (email or webhook) for receiving job completion,
    failure, and status notifications from workflows.
  tags:
  - monitoring
  - alerting
  - webhooks
- term: unstructured_api_key
  label: Unstructured API Key
  definition: The authentication credential passed in the 'unstructured-api-key' HTTP
    header to authenticate requests to both the Platform API and Partition API.
  tags:
  - authentication
  - security
- term: page_based_billing
  label: Page-Based Billing
  definition: Unstructured charges per page processed. The free tier includes 15,000
    pages; pay-as-you-go is $0.03/page.
  tags:
  - pricing
  - billing
  - finops
- term: rag_pipeline
  label: RAG Pipeline
  definition: A Retrieval-Augmented Generation pipeline that uses Unstructured to
    convert raw documents into vector-indexed chunks for LLM question-answering systems.
  tags:
  - RAG
  - LLM
  - AI