> ## Documentation Index > Fetch the complete documentation index at: https://docs.unstructured.io/llms.txt > Use this file to discover all available pages before exploring further. # Workflows To use the [Unstructured API's workflow operations](/api-reference/workflow/overview) to manage workflows, do the following: * To get a list of available workflows, use the `UnstructuredClient` object's `workflows.list_workflows` function (for the Python SDK) or the `GET` method to call the `/workflows` endpoint (for `curl` or Postman). [Learn more](/api-reference/workflow/overview#list-workflows). * To get information about a workflow, use the `UnstructuredClient` object's `workflows.get_workflow` function (for the Python SDK) or the `GET` method to call the `/workflows/` endpoint (for `curl` or Postman)use the `GET` method to call the `/workflows/` endpoint. [Learn more](/api-reference/workflow/overview#get-a-workflow). * To create a workflow, use the `UnstructuredClient` object's `workflows.create_workflow` function (for the Python SDK) or the `POST` method to call the `/workflows` endpoint (for `curl` or Postman). [Learn more](#create-a-workflow). * To run a workflow manually, use the `UnstructuredClient` object's `workflows.run_workflow` function (for the Python SDK) or the `POST` method to call the `/workflows//run` endpoint (for `curl` or Postman). [Learn more](/api-reference/workflow/overview#run-a-workflow). * To update a workflow, use the `UnstructuredClient` object's `workflows.update_workflow` function (for the Python SDK) or the `PUT` method to call the `/workflows/` endpoint (for `curl` or Postman). [Learn more](#update-a-workflow). * To delete a workflow, use the `UnstructuredClient` object's `workflows.delete_workflow` function (for the Python SDK) or the `DELETE` method to call the `/workflows/` endpoint (for `curl` or Postman). [Learn more](/api-reference/workflow/overview#delete-a-workflow). * To get a list of available workflow templates, use the `GET` method to call the `/templates` endpoint (for `curl` or Postman). [Learn more](#list-templates). * To get information about a workflow template, use the `GET` method to call the `/templates/` endpoint (for `curl` or Postman). [Learn more](#get-a-template). The following examples assume that you have already met the [requirements](/api-reference/workflow/overview#requirements) and understand the [basics](/api-reference/workflow/overview#basics) of working with the Unstructured API's workflow operations. ## Create a workflow To create a workflow, use the `UnstructuredClient` object's `workflows.create_workflow` function (for the Python SDK) or the `POST` method to call the `/workflows` endpoint (for `curl` or Postman). The following instructions create a workflow that exists until it is explicitly deleted (also known as a *long-lived workflow*). To create a workflow that exists only for the duration of that workflow's associated job run, and that job run's temporary workflow takes one or more local files only as input, see [Run an on-demand job](/api-reference/workflow/overview#run-an-on-demand-job). In the `CreateWorkflow` object (for the Python SDK) or the request body (for `curl` or Postman), specify the settings for the workflow, as follows: ```python theme={null} import os from unstructured_client import UnstructuredClient from unstructured_client.models.shared import ( WorkflowNode, CreateWorkflow, WorkflowType, Schedule ) from unstructured_client.models.operations import CreateWorkflowRequest client = UnstructuredClient( api_key_auth=os.getenv("UNSTRUCTURED_API_KEY") ) workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) another_workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) # And so on for any additional nodes. workflow = CreateWorkflow( name="", source_id="", destination_id="", workflow_type=WorkflowType., reprocess_all=True|False, template_id="", workflow_nodes=[ workflow_node, another_workflow_node # And so on for any additional nodes. ], schedule=Schedule("") ) response = client.workflows.create_workflow( request=CreateWorkflowRequest( create_workflow=workflow ) ) info = response.workflow_information print(f"name: {info.name}") print(f"id: {info.id}") print(f"status: {info.status}") print(f"type: {info.workflow_type}") print("source(s):") for source in info.sources: print(f" {source}") print("destination(s):") for destination in info.destinations: print(f" {destination}") print("schedule(s):") for crontab_entry in info.schedule.crontab_entries: print(f" {crontab_entry.cron_expression}") ``` To use a local source and a local destination, do not specify a `source_id` or `destination_id` value. Also, the `workflow_type` must be set to `CUSTOM`. A workflow with a local source has the following limitations: * The workflow cannot be set to run on a repeating schedule. * The workflow cannot be run from the Unstructured user interface (UI), even though the workflow is visible in the UI. However, you can run the workflow with the Unstructured Python SDK, or REST API clients such as `curl` or Postman. [Learn how](/api-reference/workflow/overview#run-a-workflow). ```python theme={null} import os from unstructured_client import UnstructuredClient from unstructured_client.models.shared import ( WorkflowNode, CreateWorkflow, WorkflowType ) from unstructured_client.models.operations import CreateWorkflowRequest workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) another_workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) # And so on for any additional nodes. workflow=CreateWorkflow( name="", workflow_type=WorkflowType.CUSTOM, workflow_nodes=[ workflow_node, another_workflow_node # And so on for any additional nodes. ] ) with UnstructuredClient(api_key_auth=os.getenv("UNSTRUCTURED_API_KEY")) as client: response = client.workflows.create_workflow( request=CreateWorkflowRequest( create_workflow=workflow ) ) info = response.workflow_information print(f"name: {info.name}") print(f"id: {info.id}") print(f"status: {info.status}") print(f"type: {info.workflow_type}") print("source(s):") for source in info.sources: print(f" {source}") print("destination(s):") for destination in info.destinations: print(f" {destination}") print("schedule(s):") for crontab_entry in info.schedule.crontab_entries: print(f" {crontab_entry.cron_expression}") ``` To use a local source and a remote destination, specify a `destination_id` value, and do not specify a `source_id` value. Also, the `workflow_type` must be set to `CUSTOM`. A workflow with a local source has the following limitations: * The workflow cannot be set to run on a repeating schedule. * The workflow cannot be run from the Unstructured user interface (UI), even though the workflow is visible in the UI. However, you can run the workflow with the Unstructured Python SDK, or REST API clients such as `curl` or Postman. [Learn how](/api-reference/workflow/overview#run-a-workflow). ```python theme={null} import os from unstructured_client import UnstructuredClient from unstructured_client.models.shared import ( WorkflowNode, CreateWorkflow, WorkflowType ) from unstructured_client.models.operations import CreateWorkflowRequest workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) another_workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) # And so on for any additional nodes. workflow=CreateWorkflow( name="", destination_id="", workflow_type=WorkflowType.CUSTOM, workflow_nodes=[ workflow_node, another_workflow_node # And so on for any additional nodes. ] ) with UnstructuredClient(api_key_auth=os.getenv("UNSTRUCTURED_API_KEY")) as client: response = client.workflows.create_workflow( request=CreateWorkflowRequest( create_workflow=workflow ) ) info = response.workflow_information print(f"name: {info.name}") print(f"id: {info.id}") print(f"status: {info.status}") print(f"type: {info.workflow_type}") print("source(s):") for source in info.sources: print(f" {source}") print("destination(s):") for destination in info.destinations: print(f" {destination}") print("schedule(s):") for crontab_entry in info.schedule.crontab_entries: print(f" {crontab_entry.cron_expression}") ``` ```python theme={null} import os import asyncio from unstructured_client import UnstructuredClient from unstructured_client.models.shared import ( WorkflowNode, CreateWorkflow, WorkflowType, Schedule ) from unstructured_client.models.operations import CreateWorkflowRequest async def create_workflow(): client = UnstructuredClient( api_key_auth=os.getenv("UNSTRUCTURED_API_KEY") ) workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) another_workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) # And so on for any additional nodes. workflow = CreateWorkflow( name="", source_id="", destination_id="", workflow_type=WorkflowType., reprocess_all=True|False, template_id="", workflow_nodes=[ workflow_node, another_workflow_node # And so on for any additional nodes. ], schedule=Schedule("") ) response = await client.workflows.create_workflow_async( request=CreateWorkflowRequest( create_workflow=workflow ) ) info = response.workflow_information print(f"name: {info.name}") print(f"id: {info.id}") print(f"status: {info.status}") print(f"type: {info.workflow_type}") print("source(s):") for source in info.sources: print(f" {source}") print("destination(s):") for destination in info.destinations: print(f" {destination}") print("schedule(s):") for crontab_entry in info.schedule.crontab_entries: print(f" {crontab_entry.cron_expression}") asyncio.run(create_workflow()) ``` To use a local source and a local destination, do not specify a `source_id` or `destination_id` value. Also, the `workflow_type` must be set to `CUSTOM`. A workflow with a local source has the following limitations: * The workflow cannot be set to run on a repeating schedule. * The workflow cannot be run from the Unstructured user interface (UI), even though the workflow is visible in the UI. However, you can run the workflow with the Unstructured Python SDK, or REST API clients such as `curl` or Postman. [Learn how](/api-reference/workflow/overview#run-a-workflow). ```python theme={null} import os import asyncio from unstructured_client import UnstructuredClient from unstructured_client.models.shared import ( WorkflowNode, CreateWorkflow, WorkflowType ) from unstructured_client.models.operations import CreateWorkflowRequest async def create_workflow(): workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) another_workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) # And so on for any additional nodes. workflow = CreateWorkflow( name="", workflow_type=WorkflowType.CUSTOM, workflow_nodes=[ workflow_node, another_workflow_node # And so on for any additional nodes. ] ) with UnstructuredClient(api_key_auth=os.getenv("UNSTRUCTURED_API_KEY")) as client: response = await client.workflows.create_workflow_async( request=CreateWorkflowRequest( create_workflow=workflow ) ) info = response.workflow_information print(f"name: {info.name}") print(f"id: {info.id}") print(f"status: {info.status}") print(f"type: {info.workflow_type}") print("source(s):") for source in info.sources: print(f" {source}") print("destination(s):") for destination in info.destinations: print(f" {destination}") print("schedule(s):") for crontab_entry in info.schedule.crontab_entries: print(f" {crontab_entry.cron_expression}") asyncio.run(create_workflow()) ``` To use a local source and a remote destination, specify a `destination_id` value, and do not specify a `source_id` value. Also, the `workflow_type` must be set to `CUSTOM`. A workflow with a local source has the following limitations: * The workflow cannot be set to run on a repeating schedule. * The workflow cannot be run from the Unstructured user interface (UI), even though the workflow is visible in the UI. However, you can run the workflow with the Unstructured Python SDK, or REST API clients such as `curl` or Postman. [Learn how](/api-reference/workflow/overview#run-a-workflow). ```python theme={null} import os import asyncio from unstructured_client import UnstructuredClient from unstructured_client.models.shared import ( WorkflowNode, CreateWorkflow, WorkflowType ) from unstructured_client.models.operations import CreateWorkflowRequest async def create_workflow(): workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) another_workflow_node = WorkflowNode( name="", type="", subtype="", settings={ "...": "..." } ) # And so on for any additional nodes. workflow = CreateWorkflow( name="", destination_id="", workflow_type=WorkflowType.CUSTOM, workflow_nodes=[ workflow_node, another_workflow_node # And so on for any additional nodes. ] ) with UnstructuredClient(api_key_auth=os.getenv("UNSTRUCTURED_API_KEY")) as client: response = await client.workflows.create_workflow_async( request=CreateWorkflowRequest( create_workflow=workflow ) ) info = response.workflow_information print(f"name: {info.name}") print(f"id: {info.id}") print(f"status: {info.status}") print(f"type: {info.workflow_type}") print("source(s):") for source in info.sources: print(f" {source}") print("destination(s):") for destination in info.destinations: print(f" {destination}") print("schedule(s):") for crontab_entry in info.schedule.crontab_entries: print(f" {crontab_entry.cron_expression}") asyncio.run(create_workflow()) ``` ```bash theme={null} curl --request 'POST' --location \ "$UNSTRUCTURED_API_URL/workflows" \ --header "unstructured-api-key: $UNSTRUCTURED_API_KEY" \ --header 'accept: application/json' \ --data \ '{ "name": "", "source_id": "", "destination_id": "", "workflow_type": "", "reprocess_all": "true"|"false", "template_id": "", "workflow_nodes": [ { "name": "", "type": "", "subtype": "", "settings": { "...": "..." } }, { "...": "..." } ], "schedule": "" }' ``` To use a local source and a local destination, do not specify a `source_id` or `destination_id` value. Also, the `workflow_type` must be set to `custom`. A workflow with a local source has the following limitations: * The workflow cannot be set to run on a repeating schedule. * The workflow cannot be run from the Unstructured user interface (UI), even though the workflow is visible in the UI. However, you can run the workflow with the Unstructured Python SDK, or REST API clients such as `curl` (or Postman). [Learn how](/api-reference/workflow/overview#run-a-workflow). ```bash theme={null} curl --request 'POST' --location \ "$UNSTRUCTURED_API_URL/workflows" \ --header "unstructured-api-key: $UNSTRUCTURED_API_KEY" \ --header 'accept: application/json' \ --data \ '{ "name": "", "workflow_type": "custom", "template_id": "", "workflow_nodes": [ { "name": "", "type": "", "subtype": "", "settings": { "...": "..." } }, { "...": "..." } ] }' ``` To use a local source and a remote destination, specify a `destination_id` value, and do not specify a `source_id` value. Also, the `workflow_type` must be set to `custom`. A workflow with a local source has the following limitations: * The workflow cannot be set to run on a repeating schedule. * The workflow cannot be run from the Unstructured user interface (UI), even though the workflow is visible in the UI. However, you can run the workflow with the Unstructured Python SDK, or REST API clients such as `curl` (or Postman). [Learn how](/api-reference/workflow/overview#run-a-workflow). ```bash theme={null} curl --request 'POST' --location \ "$UNSTRUCTURED_API_URL/workflows" \ --header "unstructured-api-key: $UNSTRUCTURED_API_KEY" \ --header 'accept: application/json' \ --data \ '{ "name": "", "destination_id": "", "workflow_type": "custom", "template_id": "", "workflow_nodes": [ { "name": "", "type": "", "subtype": "", "settings": { "...": "..." } }, { "...": "..." } ] }' ``` 1. In the method drop-down list, select **POST**. 2. In the address box, enter the following URL: ```text theme={null} {{UNSTRUCTURED_API_URL}}/workflows ``` 3. On the **Headers** tab, enter the following headers: * **Key**: `unstructured-api-key`, **Value**: `{{UNSTRUCTURED_API_KEY}}` * **Key**: `accept`, **Value**: `application/json` 4. On the **Body** tab, select **raw** and **JSON**, and specify the settings for the workflow: ```json theme={null} { "name": "", "source_id": "", "destination_id": "", "workflow_type": "", "reprocess_all": "true"|"false", "template_id": "", "workflow_nodes": [ { "name": "", "type": "", "subtype": "", "settings": { "...": "..." } }, { "...": "..." } ], "schedule": "" } ``` 5. Click **Send**. To use a local source and a local destination do not specify a `source_id` or `destination_id` value. Also, the `workflow_type` must be set to `custom`. A workflow with a local source has the following limitations: * The workflow cannot be set to run on a repeating schedule. * The workflow cannot be run from the Unstructured user interface (UI), even though the workflow is visible in the UI. However, you can run the workflow with the Unstructured Python SDK, or REST API clients such as Postman (or `curl`). [Learn how](/api-reference/workflow/overview#run-a-workflow). 1. In the method drop-down list, select **POST**. 2. In the address box, enter the following URL: ```text theme={null} {{UNSTRUCTURED_API_URL}}/workflows ``` 3. On the **Headers** tab, enter the following headers: * **Key**: `unstructured-api-key`, **Value**: `{{UNSTRUCTURED_API_KEY}}` * **Key**: `accept`, **Value**: `application/json` 4. On the **Body** tab, select **raw** and **JSON**, and specify the settings for the workflow: ```json theme={null} { "name": "", "workflow_type": "custom", "template_id": "", "workflow_nodes": [ { "name": "", "type": "", "subtype": "", "settings": { "...": "..." } }, { "...": "..." } ] } ``` 5. Click **Send**. To use a local source and a remote destination, specify a `destination_id` value, and do not specify a `source_id` value. Also, the `workflow_type` must be set to `custom`. A workflow with a local source has the following limitations: * The workflow cannot be set to run on a repeating schedule. * The workflow cannot be run from the Unstructured user interface (UI), even though the workflow is visible in the UI. However, you can run the workflow with the Unstructured Python SDK, or REST API clients such as Postman (or `curl`). [Learn how](/api-reference/workflow/overview#run-a-workflow). 1. In the method drop-down list, select **POST**. 2. In the address box, enter the following URL: ```text theme={null} {{UNSTRUCTURED_API_URL}}/workflows ``` 3. On the **Headers** tab, enter the following headers: * **Key**: `unstructured-api-key`, **Value**: `{{UNSTRUCTURED_API_KEY}}` * **Key**: `accept`, **Value**: `application/json` 4. On the **Body** tab, select **raw** and **JSON**, and specify the settings for the workflow: ```json theme={null} { "name": "", "workflow_type": "custom", "template_id": "", "workflow_nodes": [ { "name": "", "destination_id": "", "type": "", "subtype": "", "settings": { "...": "..." } }, { "...": "..." } ] } ``` 5. Click **Send**. Replace the preceding placeholders as follows: * `` (*required*) - A unique name for this workflow. * `` (*required*) - The ID of the target source connector. To get the ID, use the `UnstructuredClient` object's `sources.list_sources` function (for the Python SDK) or the `GET` method to call the `/sources` endpoint (for `curl` or Postman). [Learn more](/api-reference/workflow/overview#list-source-connectors). * `` (*required*) - The ID of the target destination connector. To get the ID, use the `UnstructuredClient` object's `destinations.list_destinations` function (for the Python SDK) or the `GET` method to call the `/destinations` endpoint (for `curl` or Postman). [Learn more](/api-reference/workflow/overview#list-destination-connectors). * `` (*required*) - The workflow type. Available values include: * `CUSTOM` (for the Python SDK) and `custom` (for `curl` or Postman) - This workflow's nodes are manually specified in the `workflow_nodes` array. For instructions, see [Custom workflow DAG nodes](#custom-workflow-dag-nodes). * `TEMPLATE` (for the Python SDK) and `template` (for `curl` or Postman) - This workflow's nodes are specified by a workflow template. In this case, you would not add a `workflow_nodes` array. Instead, you must specify the workflow template's unique ID in the `template_id` field. For instructions, see [List workflow templates](#get-a-template). The previously-available workflow optimization types `ADVANCED`, `BASIC`, and `PLATINUM` (for the Python SDK) and `advanced`, `basic`, and `platinum` (for `curl` or Postman) are non-operational and planned to be fully removed in a future release. The ability to create an [automatic workflow](/ui/workflows#create-an-automatic-workflow) type is currently not available but is planned to be added in a future release. * `reprocess_all`: *Optional*. If set to true (the default), the workflow will reprocess all documents in the source location on every workflow run. If set to `false`, the workflow will only process new documents that are added to the source location, or existing documents that are updated in the source location (as determined by checking whether the file's version has changed), since the last workflow run. This setting applies only to blob storage connectors such as the Amazon S3, Azure Blob Storage, and Google Cloud Storage source connectors. * `` - The repeating automatic run schedule, specified as a predefined phrase. The available predefined phrases are: * `every 15 minutes` (for `curl` or Postman): Every 15 minutes (cron expression: `*/15 * * * *`). * `every hour`: At the first minute of every hour (cron expression: `0 * * * *`). * `every 2 hours`: At the first minute of every second hour (cron expression: `0 */2 * * *`). * `every 4 hours`: At the first minute of every fourth hour (cron expression: `0 */4 * * *`). * `every 6 hours`: At the first minute of every sixth hour (cron expression: `0 */6 * * *`). * `every 8 hours`: At the first minute of every eighth hour (cron expression: `0 */8 * * *`). * `every 10 hours`: At the first minute of every tenth hour (cron expression: `0 */10 * * *`). * `every 12 hours`: At the first minute of every twelfth hour (cron expression: `0 */12 * * *`). * `daily`: At the first minute of every day (cron expression: `0 0 * * *`). * `weekly`: At the first minute of every Sunday (cron expression: `0 0 * * 0`). * `monthly`: At the first minute of the first day of every month (cron expression: `0 0 1 * *`). If `schedule` is not specified, the workflow does not automatically run on a repeating schedule. Workflows with a local source cannot be set to run on a repeating schedule. ## Update a workflow To update information about a workflow, use the `UnstructuredClient` object's `workflows.update_workflow` function (for the Python SDK) or the `PUT` method to call the `/workflows/` endpoint (for `curl` or Postman), replacing `` with the workflow's unique ID. To get this ID, see [List workflows](#list-workflows). The following instructions can be used only to update a long-lived workflow (a workflow that exists until it is explicitly deleted). A workflow that exists only for the duration of a running [on-demand job](/api-reference/workflow/overview#run-an-on-demand-job) cannot be updated. In the request body, specify the settings for the workflow. For the specific settings to include, see [Create a workflow](/api-reference/workflow/workflows#create-a-workflow). ```python theme={null} import os from unstructured_client import UnstructuredClient from unstructured_client.models.shared import ( WorkflowNode, UpdateWorkflow, WorkflowType, Schedule, ), from unstructured_client.models.operations import UpdateWorkflowRequest client = UnstructuredClient( api_key_auth=os.getenv("UNSTRUCTURED_API_KEY") ) workflow_node = WorkflowNode( # Specify the settings for a workflow node here. ) another_workflow_node = WorkflowNode( # Specify the settings for another workflow node here. ) # And so on for any additional nodes. workflow = UpdateWorkflow( # Specify the settings for the workflow here. ) response = client.workflows.update_workflow( request=UpdateWorkflowRequest( workflow_id="", update_workflow=workflow ) ) info = response.workflow_information print(f"name: {info.name}") print(f"id: {info.id}") print(f"status: {info.status}") print(f"type: {info.workflow_type}") print("source(s):") for source in info.sources: print(f" {source}") print("destination(s):") for destination in info.destinations: print(f" {destination}") print("schedule(s):") for crontab_entry in info.schedule.crontab_entries: print(f" {crontab_entry.cron_expression}") ``` ```python theme={null} import os import asyncio from unstructured_client import UnstructuredClient from unstructured_client.models.shared import ( WorkflowNode, UpdateWorkflow, WorkflowType, Schedule, ), from unstructured_client.models.operations import UpdateWorkflowRequest async def update_workflow(): client = UnstructuredClient( api_key_auth=os.getenv("UNSTRUCTURED_API_KEY") ) workflow_node = WorkflowNode( # Specify the settings for a workflow node here. ) another_workflow_node = WorkflowNode( # Specify the settings for another workflow node here. ) # And so on for any additional nodes. workflow = UpdateWorkflow( # Specify the settings for the workflow here. ) response = await client.workflows.update_workflow_async( request=UpdateWorkflowRequest( workflow_id="", update_workflow=workflow ) ) info = response.workflow_information print(f"name: {info.name}") print(f"id: {info.id}") print(f"status: {info.status}") print(f"type: {info.workflow_type}") print("source(s):") for source in info.sources: print(f" {source}") print("destination(s):") for destination in info.destinations: print(f" {destination}") print("schedule(s):") for crontab_entry in info.schedule.crontab_entries: print(f" {crontab_entry.cron_expression}") asyncio.run(update_workflow()) ``` ```bash theme={null} curl --request 'PUT' --location \ "$UNSTRUCTURED_API_URL/workflows/" \ --header "unstructured-api-key: $UNSTRUCTURED_API_KEY" \ --header 'accept: application/json' \ --data \ '{ # Specify the settings for the workflow here. }' ``` 1. In the method drop-down list, select **PUT**. 2. In the address box, enter the following URL: ```text theme={null} {{UNSTRUCTURED_API_URL}}/workflows/ ``` 3. On the **Headers** tab, enter the following headers: * **Key**: `unstructured-api-key`, **Value**: `{{UNSTRUCTURED_API_KEY}}` * **Key**: `accept`, **Value**: `application/json` 4. On the **Body** tab, select **raw** and **JSON**, and specify the settings for the workflow. 5. Click **Send**. ## Custom workflow DAG nodes If `WorkflowType` is set to `CUSTOM` (for the Python SDK), or if `workflow_type` is set to `custom` (for `curl` or Postman), you must also specify the settings for the workflow's directed acyclic graph (DAG) nodes. These nodes' settings are specified in the `workflow_nodes` array. * A **Source** node is automatically created when you specify the `source_id` value outside of the\ `workflow_nodes` array. * A **Destination** node is automatically created when you specify the `destination_id` value outside of the `workflow_nodes` array. * You can specify [Partitioner](#partitioner-node), [Enrichment](#enrichment-node), [Chunker](#chunker-node), and [Embedder](#embedder-node) nodes. Unstructured can potentially generate image summary descriptions, table summary descriptions, table-to-HTML output, and generative OCR optimizations, only for workflows that are configured as follows: * With a **Partitioner** node set to use the **Auto** or **High Res** partitioning strategy, and an image summary description node, table summary description node, table-to-HTML output node, or generative OCR optimization node is added. * With a **Partitioner** node set to use the **VLM** partitioning strategy. No image summary description node, table summary description node, table-to-HTML output node, or generative OCR optimization node is needed (or allowed). Even with these configurations, Unstructured actually generates image summary descriptions, table summary descriptions, and table-to-HTML output only for files that contain images or tables and are also eligible for processing with the following partitioning strategies: * **High Res**, when the workflow's **Partitioner** node is set to use **Auto** or **High Res**. * **VLM** or **High Res**, when the workflow's **Partitioner** node is set to use **VLM**. Unstructured never generates image summary descriptions, table summary descriptions, or table-to-HTML output for workflows that are configured as follows: * With a **Partitioner** node set to use the **Fast** partitioning strategy. * With a **Partitioner** node set to use the **Auto**, **High Res**, or **VLM** partitioning strategy, for all files that Unstructured encounters that do not contain images or tables. Unstructured never produces generative OCR optimizations for workflows with a **Partitioner** node set to use the **Fast** partitioning strategy. * The order of the nodes in the `workflow_nodes` array will be the same order that these nodes appear in the DAG, with the first node in the array added directly after the **Source** node. The **Destination** node follows the last node in the array. * Be sure to specify nodes in the allowed order. The following DAG placements are all allowed: ```mermaid theme={null} flowchart LR Partitioner ``` ```mermaid theme={null} flowchart LR Partitioner-->Chunker ``` ```mermaid theme={null} flowchart LR Partitioner-->Chunker-->Embedder ``` ```mermaid theme={null} flowchart LR Partitioner-->Enrichment-->Chunker ``` ```mermaid theme={null} flowchart LR Partitioner-->Enrichment-->Chunker-->Embedder ``` For workflows that use **Chunker** and enrichment nodes together, the **Chunker** node should be placed after all enrichment nodes. Placing the **Chunker** node before any enrichment nodes could cause incomplete or no enrichment results to be generated. ### Partitioner node A **Partitioner** node has a `type` of `partition`. [Learn about the available partitioning strategies](/ui/partitioning). #### Auto strategy ```python theme={null} auto_partitioner_workflow_node = WorkflowNode( name="Partitioner", subtype="vlm", type="partition", settings={ "provider": "", "provider_api_key": None, "model": "", "output_format": "", "format_html": , "unique_element_ids": , "is_dynamic": True, "allow_fast": True } ) ``` ```json theme={null} { "name": "Partitioner", "type": "vlm", "subtype": "partition", "settings": { "provider": "", "provider_api_key": null, "model": "", "output_format": "", "format_html": , "unique_element_ids": , "is_dynamic": true, "allow_fast": true } } ``` Fields for `settings` include: * `provider`: *Optional*. If the Auto partitioning strategy needs to use the VLM partitioning strategy, then use the specified VLM provider. Allowed values include `anthropic`, `auto`, `bedrock`, `openai`, and `vertexai`. The default value is `anthropic`. * `provider_api_key`: *Optional*. If specified, use a non-default API key for calls to the specified VLM provider as needed. The default is none, which means to rely on using Unstructured's internal default API key for the VLM provider. * `model`: *Optional*. If the Auto partitioning strategy needs to use the VLM partitioning strategy, then use the specified VLM. The default value is `claude-sonnet-4-20250514`. The following models are no longer available as of the following dates: * For `anthropic`, `claude-3-5-sonnet-20241022`: October 22, 2025 * For `bedrock`, `us.anthropic.claude-3-5-sonnet-20241022-v2:0`: October 22, 2025 Unstructured recommends the following actions: * For new workflows, do not use any of these models. * For any workflow that uses any of these models, update that workflow as soon as possible to use a different model. Workflows that attempt to use any of these models on or after its associated date will return errors. The following list applies only to Unstructured **Let's Go** and **Pay-As-You-Go** accounts. For Unstructured **Business** accounts, to get your current list of available models, contact your Unstructured account administrator or Unstructured sales representative, or email Unstructured Support at [support@unstructured.io](mailto:support@unstructured.io). * For `anthropic`, available values for `model` are: * `claude-3-7-sonnet-20250219` * `claude-sonnet-4-20250514` * `claude-sonnet-4-5-20250929` * For `bedrock`, available values for `model` are: * `us.amazon.nova-lite-v1:0` * `us.amazon.nova-pro-v1:0` * `us.anthropic.claude-3-opus-20240229-v1:0` * `us.anthropic.claude-3-haiku-20240307-v1:0` * `us.anthropic.claude-3-sonnet-20240229-v1:0` * `us.anthropic.claude-3-7-sonnet-20250219-v1:0` * `us.anthropic.claude-sonnet-4-20250514-v1:0` * For `openai`, available values for `model` are: * `gpt-4o` * `gpt-5-mini-2025-08-07` * For `vertexai`, available values for `model` are: * `gemini-2.0-flash-001` * `output_format`: *Output*. The format of the response. Allowed values include `text/html` and `application/json`. The default is `text/html`. * `format_html`: *Optional*. If the Auto partitioning strategy needs to use the VLM partitioning strategy, true (the default) to apply Beautiful Soup's `prettify` method to the HTML that is generated by the VLM partitioner, which for example adds indentation for better readability. * `unique_element_ids`: *Optional*. True (the default) to assign UUIDs to element IDs, which guarantees their uniqueness. This is useful for example when using them as primary keys in a database. False to assign a SHA-256 of the element's text as its element ID. * `is_dynamic`: *Required*. True to enable dynamic routing of pages to Fast, High Res, or VLM as needed for better overall performance and cost savings. * `allow_fast`: *Required*. True to allow routing of pages to Fast as needed for better overall performance and cost savings. #### VLM strategy ```python theme={null} vlm_partitioner_workflow_node = WorkflowNode( name="Partitioner", subtype="vlm", type="partition", settings={ "provider": "", "provider_api_key": None, "model": "", "output_format": "", "format_html": , "unique_element_ids": , "is_dynamic": False, "allow_fast": True|False } ) ``` ```json theme={null} { "name": "Partitioner", "type": "partition", "subtype": "vlm", "settings": { "provider": "", "provider_api_key": null, "model": "", "output_format": "", "format_html": , "unique_element_ids": , "is_dynamic": false, "allow_fast": } } ``` Fields for `settings` include: * `provider`: *Optional*. Use the specified VLM provider. Allowed values include `anthropic`, `auto`, `bedrock`, `openai`, and `vertexai`. The default value is `anthropic`. * `provider_api_key`: *Optional*. If specified, use a non-default API key for calls to the specified VLM provider as needed. The default is none, which means to rely on using Unstructured's internal default API key for the VLM provider. * `model`: *Optional*. If the Auto partitioning strategy needs to use the VLM partitioning strategy, then use the specified VLM. The default value is `claude-sonnet-4-20250514`. The following models are no longer available as of the following dates: * For `anthropic`, `claude-3-5-sonnet-20241022`: October 22, 2025 * For `bedrock`, `us.anthropic.claude-3-5-sonnet-20241022-v2:0`: October 22, 2025 Unstructured recommends the following actions: * For new workflows, do not use any of these models. * For any workflow that uses any of these models, update that workflow as soon as possible to use a different model. Workflows that attempt to use any of these models on or after its associated date will return errors. The following list applies only to Unstructured **Let's Go** and **Pay-As-You-Go** accounts. For Unstructured **Business** accounts, to get your current list of available models, contact your Unstructured account administrator or Unstructured sales representative, or email Unstructured Support at [support@unstructured.io](mailto:support@unstructured.io). * For `anthropic`, available values for `model` are: * `claude-opus-4-5-20251101` * `claude-3-7-sonnet-20250219` * `claude-sonnet-4-20250514` * `claude-sonnet-4-5-20250929` * For `bedrock`, available values for `model` are: * `us.amazon.nova-lite-v1:0` * `us.amazon.nova-pro-v1:0` * `us.anthropic.claude-3-opus-20240229-v1:0` * `us.anthropic.claude-3-haiku-20240307-v1:0` * `us.anthropic.claude-3-sonnet-20240229-v1:0` * `us.anthropic.claude-3-7-sonnet-20250219-v1:0` * `us.anthropic.claude-sonnet-4-20250514-v1:0` * For `openai`, available values for `model` are: * `gpt-4o` * `gpt-5-mini-2025-08-07` * For `vertexai`, available values for `model` are: * `gemini-2.0-flash-001` * `output_format`: *Output*. The format of the response. Allowed values include `text/html` and `application/json`. The default is `text/html`. * `format_html`: *Optional*. True (the default) to apply Beautiful Soup's `prettify` method to the HTML that is generated by the VLM partitioner, which for example adds indentation for better readability. * `unique_element_ids`: *Optional*. True (the default) to assign UUIDs to element IDs, which guarantees their uniqueness. This is useful for example when using them as primary keys in a database. False to assign a SHA-256 of the element's text as its element ID. * `is_dynamic`: *Required*. False to use the VLM strategy. * `allow_fast`: *Optional*. True (the default) to allow routing of pages to Fast as needed for better overall performance and cost savings. #### High Res strategy ```python theme={null} high_res_paritioner_workflow_node = WorkflowNode( name="Partitioner", subtype="unstructured_api", type="partition", settings={ "strategy": "hi_res", "include_page_breaks": , "pdf_infer_table_structure": , "exclude_elements": [ "", "" ], "xml_keep_tags": , "encoding": "", "ocr_languages": [ "", "" ], "extract_image_block_types": [ "image", "table" ], "infer_table_structure": , "coordinates": } ) ``` ```json theme={null} { "name": "Partitioner", "type": "partition", "subtype": "unstructured_api", "settings": { "strategy": "hi_res", "include_page_breaks": , "pdf_infer_table_structure": , "exclude_elements": [ "", "" ], "xml_keep_tags": , "encoding": "", "ocr_languages": [ "", "" ], "extract_image_block_types": [ "image", "table" ], "infer_table_structure": , "coordinates": } } ``` * `strategy`: *Required*. The partitioning strategy to use. This field must be set to `hi_res`. * `include_page_breaks`: *Optional*. True to include page breaks in the output if supported by the file type. The default is false. * `pdf_infer_table_structure`: *Optional*. True for any `Table` elements extracted from a PDF to include an additional metadata field, `text_as_html`, where the value (string) is a just a transformation of the data into an HTML table. The default is false. * `exclude_elements`: *Optional*. A list of any Unstructured element types to exclude from the output. The default is none. Available values include: * `Address` * `EmailAddress` * `FigureCaption` * `Footer` * `Formula` * `Header` * `Image` * `ListItem` * `NarrativeText` * `PageBreak` * `Table` * `Title` * `UncategorizedText` * `xml_keep_tags`: *Optional*. True to retain any XML tags in the output. False (the default) to just extract the text from any XML tags instead. * `encoding`: *Optional*. The encoding method used to decode the text input. The default is `utf-8`. * `ocr_languages`: *Optional*. A list of languages present in the input, for use in partitioning, OCR, or both. Multiple languages indicate that the text could be in any of the specified languages. The default is `[ 'eng' ]`. [See the language codes list](https://github.com/Unstructured-IO/unstructured/blob/main/unstructured/partition/common/lang.py). * `extract_image_block_types`: *Optional*. A list of the Unstructured element types for use in extracting image blocks as Base64 encoded data stored in `metadata` fields. Available values include: * `Abstract` * `BulletedText` * `Caption` * `CodeSnippet` * `CompositeElement` * `Figure` * `FigureCaption` * `Form` * `FormKeysValues` * `Formula` * `Header` * `Image` * `List` * `List-item` * `ListItem` * `NarrativeText` * `Paragraph` * `Picture` * `Table` * `Text` * `Threading` * `Title` * `UncategorizedText` * `infer_table_structure`: *Optional*. True to have any table elements extracted from a PDF to include an additional `metadata` field named `text_as_html`, containing an HTML `` transformation. The default is false. * `coordinates`: *Optional*. True to have each element extracted from a PDF to include information about that element's position relative to its corresponding page. The default is false. #### Fast strategy ```python theme={null} fast_partitioner_workflow_node = WorkflowNode( name="Partitioner", subtype="unstructured_api", type="partition", settings={ "strategy": "fast", "include_page_breaks": , "pdf_infer_table_structure": , "exclude_elements": [ "", "" ], "xml_keep_tags": , "encoding": "", "ocr_languages": [ "", "" ], "extract_image_block_types": [ "image", "table" ], "infer_table_structure": , "coordinates": } ) ``` ```json theme={null} { "name": "Partitioner", "type": "partition", "subtype": "unstructured_api", "settings": { "strategy": "fast", "include_page_breaks": , "pdf_infer_table_structure": , "exclude_elements": [ "", "" ], "xml_keep_tags": , "encoding": "", "ocr_languages": [ "", "" ], "extract_image_block_types": [ "image", "table" ], "infer_table_structure": , "coordinates": } } ``` Fields for `settings` include: * `strategy`: *Required*. The partitioning strategy to use. This field must be set to `fast`. * \`include\_page\_breaks: *Optional*. True to include page breaks in the output if supported by the file type. The default is false. * `pdf_infer_table_structure`: *Optional*. Although this field is listed, it applies only to the `hi_res` strategy and will not work if set to true. The default is false. * `exclude_elements`: *Optional*. A list of any Unstructured element types to exclude from the output. The default is none. Available values include: * `Address` * `EmailAddress` * `FigureCaption` * `Footer` * `Formula` * `Header` * `Image` * `ListItem` * `NarrativeText` * `PageBreak` * `Table` * `Title` * `UncategorizedText` * `xml_keep_tags`: *Optional*. True to retain any XML tags in the output. False (the default) to just extract the text from any XML tags instead. * `encoding`: *Optional*. The encoding method used to decode the text input. The default is `utf-8`. * `ocr_languages`: *Optional*. A list of languages present in the input, for use in partitioning, OCR, or both. Multiple languages indicate that the text could be in any of the specified languages. The default is `[ 'eng' ]`. [See the language codes list](https://github.com/Unstructured-IO/unstructured/blob/main/unstructured/partition/common/lang.py). * `extract_image_block_types`: *Optional*. A list of the Unstructured element types for use in extracting image blocks as Base64 encoded data stored in `metadata` fields. Available values include: * `Abstract` * `BulletedText` * `Caption` * `CodeSnippet` * `CompositeElement` * `Figure` * `FigureCaption` * `Form` * `FormKeysValues` * `Formula` * `Header` * `Image` * `List` * `List-item` * `ListItem` * `NarrativeText` * `Paragraph` * `Picture` * `Table` * `Text` * `Threading` * `Title` * `UncategorizedText` * `infer_table_structure`: *Optional*. True to have any table elements extracted from a PDF to include an additional `metadata` field named `text_as_html`, containing an HTML `
` transformation. The default is false. * `coordinates`: *Optional*. True to have each element extracted from a PDF to include information about that element's position relative to its corresponding page. The default is false. ### Enrichment node An enrichment node includes a node such as **Image Description**, **Table Description**, **Table to HTML**, **NER**, or **Generative OCR**. An enrichment node has a `type` of `prompter`. [Learn about the available enrichments](/ui/enriching/overview). Unstructured can potentially generate image summary descriptions, table summary descriptions, table-to-HTML output, and generative OCR optimizations, only for workflows that are configured as follows: * With a **Partitioner** node set to use the **Auto** or **High Res** partitioning strategy, and an image summary description node, table summary description node, table-to-HTML output node, or generative OCR optimization node is added. * With a **Partitioner** node set to use the **VLM** partitioning strategy. No image summary description node, table summary description node, table-to-HTML output node, or generative OCR optimization node is needed (or allowed). Even with these configurations, Unstructured actually generates image summary descriptions, table summary descriptions, and table-to-HTML output only for files that contain images or tables and are also eligible for processing with the following partitioning strategies: * **High Res**, when the workflow's **Partitioner** node is set to use **Auto** or **High Res**. * **VLM** or **High Res**, when the workflow's **Partitioner** node is set to use **VLM**. Unstructured never generates image summary descriptions, table summary descriptions, or table-to-HTML output for workflows that are configured as follows: * With a **Partitioner** node set to use the **Fast** partitioning strategy. * With a **Partitioner** node set to use the **Auto**, **High Res**, or **VLM** partitioning strategy, for all files that Unstructured encounters that do not contain images or tables. Unstructured never produces generative OCR optimizations for workflows with a **Partitioner** node set to use the **Fast** partitioning strategy. #### Image Description task Unstructured can potentially generate image summary descriptions only for workflows that are configured as follows: * With a **Partitioner** node set to use the **Auto** or **High Res** partitioning strategy, and an image summary description node is added. * With a **Partitioner** node set to use the **VLM** partitioning strategy. No image summary description node is needed (or allowed). Even with these configurations, Unstructured actually generates image summary descriptions only for files that contain images and are also eligible for processing with the following partitioning strategies: * **High Res**, when the workflow's **Partitioner** node is set to use **Auto** or **High Res**. * **VLM** or **High Res**, when the workflow's **Partitioner** node is set to use **VLM**. Unstructured never generates image summary descriptions for workflows that are configured as follows: * With a **Partitioner** node set to use the **Fast** partitioning strategy. * With a **Partitioner** node set to use the **Auto**, **High Res**, or **VLM** partitioning strategy, for all files that Unstructured encounters that do not contain images. ```python theme={null} image_description_enrichment_workflow_node = WorkflowNode( name="Enrichment", subtype="", type="prompter", settings={ "provider_type": "", "model": "" } ) ``` ```json theme={null} { "name": "Enrichment", "type": "prompter", "subtype": "", "settings": { "provider_type": "", "model": "" } } ``` Allowed values for ``, ``, and `` include, respectively: * `anthropic_image_description`, `anthropic`, and: * `claude-opus-4-5-20251101` * `claude-3-7-sonnet-20250219` * `claude-sonnet-4-20250514` * `claude-sonnet-4-5-20250929` * `bedrock_image_description`, `bedrock`, and: * `us.amazon.nova-lite-v1:0` * `us.amazon.nova-pro-v1:0` * `us.anthropic.claude-3-haiku-20240307-v1:0` * `us.anthropic.claude-3-opus-20240229-v1:0` * `us.anthropic.claude-3-sonnet-20240229-v1:0` * `us.anthropic.claude-3-7-sonnet-20250219-v1:0` * `us.anthropic.claude-sonnet-4-20250514-v1:0` * `us.anthropic.claude-sonnet-4-5-20250929-v1:0` * `openai_image_description`, `openai`, and: * `gpt-4o` * `gpt-4o-mini` * `gpt-5-mini` * `vertexai_image_description`, `vertexai`, and: * `gemini-2.0-flash-001` The preceding list applies only to Unstructured **Let's Go** and **Pay-As-You-Go** accounts. For Unstructured **Business** accounts, to get your current list of available `` values, contact your Unstructured account administrator or Unstructured sales representative, or email Unstructured Support at [support@unstructured.io](mailto:support@unstructured.io). #### Table Description task Unstructured can potentially generate table summary descriptions only for workflows that are configured as follows: * With a **Partitioner** node set to use the **Auto** or **High Res** partitioning strategy, and a table summary description node is added. * With a **Partitioner** node set to use the **VLM** partitioning strategy. No table summary description node is needed (or allowed). Even with these configurations, Unstructured actually generates table summary descriptions only for files that contain tables and are also eligible for processing with the following partitioning strategies: * **High Res**, when the workflow's **Partitioner** node is set to use **Auto** or **High Res**. * **VLM** or **High Res**, when the workflow's **Partitioner** node is set to use **VLM**. Unstructured never generates table summary descriptions for workflows that are configured as follows: * With a **Partitioner** node set to use the **Fast** partitioning strategy. * With a **Partitioner** node set to use the **Auto**, **High Res**, or **VLM** partitioning strategy, for all files that Unstructured encounters that do not contain tables. ```python theme={null} table_description_enrichment_workflow_node = WorkflowNode( name="Enrichment", subtype="", type="prompter", settings={ "provider_type": "", "model": "" } ) ``` ```json theme={null} { "name": "Enrichment", "type": "prompter", "subtype": "", "settings": { "provider_type": "", "model": "" } } ``` Allowed values for ``, ``, and `` include, respectively: * `anthropic_table_description`, `anthropic`, and: * `claude-opus-4-5-20251101` * `claude-3-7-sonnet-20250219` * `claude-sonnet-4-20250514` * `claude-sonnet-4-5-20250929` * `bedrock_table_description`, `bedrock`, and: * `us.amazon.nova-lite-v1:0` * `us.amazon.nova-pro-v1:0` * `us.anthropic.claude-3-haiku-20240307-v1:0` * `us.anthropic.claude-3-opus-20240229-v1:0` * `us.anthropic.claude-3-sonnet-20240229-v1:0` * `us.anthropic.claude-3-7-sonnet-20250219-v1:0` * `us.anthropic.claude-sonnet-4-20250514-v1:0` * `us.anthropic.claude-sonnet-4-5-20250929-v1:0` * `openai_table_description`, `openai`, and: * `gpt-4o` * `gpt-4o-mini` * `gpt-5-mini` * `vertexai_table_description`, `vertexai`, and: * `gemini-2.0-flash-001` The preceding list applies only to Unstructured **Let's Go** and **Pay-As-You-Go** accounts. For Unstructured **Business** accounts, to get your current list of available `` values, contact your Unstructured account administrator or Unstructured sales representative, or email Unstructured Support at [support@unstructured.io](mailto:support@unstructured.io). #### Table to HTML task Unstructured can potentially generate table-to-HTML output only for workflows that are configured as follows: * With a **Partitioner** node set to use the **Auto** or **High Res** partitioning strategy, and a table-to-HTML output node is added. * With a **Partitioner** node set to use the **VLM** partitioning strategy. No table-to-HTML output node is needed (or allowed). Even with these configurations, Unstructured actually generates table-to-HTML output only for files that contain tables and are also eligible for processing with the following partitioning strategies: * **High Res**, when the workflow's **Partitioner** node is set to use **Auto** or **High Res**. * **VLM** or **High Res**, when the workflow's **Partitioner** node is set to use **VLM**. Unstructured never generates table-to-HTML output for workflows that are configured as follows: * With a **Partitioner** node set to use the **Fast** partitioning strategy. * With a **Partitioner** node set to use the **Auto**, **High Res**, or **VLM** partitioning strategy, for all files that Unstructured encounters that do not contain tables. ```python theme={null} table_to_html_enrichment_workflow_node = WorkflowNode( name="Enrichment", subtype="", type="prompter", settings={ "provider_type": "", "model": "" } ) ``` ```json theme={null} { "name": "Enrichment", "type": "prompter", "subtype": "", "settings": { "provider_type": "", "model": "" } } ``` Allowed values for ``, ``, and `` include, respectively: * `twopass_table2html` for agentic AI table-to-HTML output. Do not specify a `` or `` for this subtype. * `anthropic_table2html` for VLM table-to-HTML output by using Anthropic, `anthropic`, and: * `claude-opus-4-5-20251101` * `claude-3-7-sonnet-20250219` * `claude-sonnet-4-20250514` * `claude-sonnet-4-5-20250929` * `openai_table2html` for VLM table-to-HTML output by using OpenAI, `openai`, and: * `gpt-4o` * `gpt-4o-mini` * `gpt-5-mini` The preceding `subtype` value applies only to Unstructured **Let's Go** and **Pay-As-You-Go** accounts. For Unstructured **Business** accounts, to get your current list of available `subtype` values, contact your Unstructured account administrator or Unstructured sales representative, or email Unstructured Support at [support@unstructured.io](mailto:support@unstructured.io). #### Named Entity Recognition (NER) task ```python theme={null} ner_enrichment_workflow_node = WorkflowNode( name="Enrichment", subtype="", type="prompter", settings={ "provider_type": "", "model": "", "prompt_interface_overrides": { "prompt": { "user": "" } } } ) ``` ```json theme={null} { "name": "Enrichment", "type": "prompter", "subtype": "", "settings": { "provider_type": "", "model": "", "prompt_interface_overrides": { "prompt": { "user": "" } } } } ``` Fields for settings include: * Allowed values for ``, ``, and `` include, respectively: * `anthropic_ner`, `anthropic`, and: * `claude-opus-4-5-20251101` * `claude-3-7-sonnet-20250219` * `claude-sonnet-4-20250514` * `claude-sonnet-4-5-20250929` * `openai_ner`, `openai`, and: * `gpt-4o` * `gpt-4o-mini` * `gpt-5-mini` * `prompt_interface_overrides.prompt.user`: *Optional*. Any alternative prompt to use with the underlying NER model. The default is none, which means to rely on using Unstructured's internal default prompt when calling the NER model. The internal default prompt is as follows, which you can override by providing an alternative prompt: ```text theme={null} Extract named entities and their relationships from the following text. Provide the entities, their corresponding types and relationships as a structured JSON response. Entity types: - PERSON - ORGANIZATION - LOCATION - DATE - TIME - EVENT - MONEY - PERCENT - FACILITY - PRODUCT - ROLE - DOCUMENT - DATASET Relationship types: - PERSON - ORGANIZATION: works_for, affiliated_with, founded - PERSON - LOCATION: born_in, lives_in, traveled_to - ORGANIZATION - LOCATION: based_in, has_office_in - Entity - DATE: occurred_on, founded_on, died_on, published_in - PERSON - PERSON: married_to, parent_of, colleague_of - PRODUCT - ORGANIZATION: developed_by, owned_by - EVENT - LOCATION: held_in, occurred_in - Entity - ROLE: has_title, acts_as, has_role - DATASET - PERSON: mentions - DATASET - DOCUMENT: located_in - PERSON - DATASET: published - DOCUMENT - DOCUMENT: referenced_in, contains - DOCUMENT - DATE: dated - PERSON - DOCUMENT: published [START OF TEXT] {{text}} [END OF TEXT] Response format json schema: { "items": [ { "entity": "Entity name", "type": "Entity type" }, { "entity": "Entity name", "type": "Entity type" } ], "relationships": [ {"from": "Entity name", "relationship": "Relationship type", "to": "Entity name"}, {"from": "Entity name", "relationship": "Relationship type", "to": "Entity name"} ] } ``` If you provide an alternative prompt, you must provide the entire alternative prompt in the preceding format. For best results, Unstructured strongly recommends that you limit your changes only to certain portions of the internal default prompt, specifically: * Adding, renaming, or deleting items in the list of predefined types (such as `PERSON`, `ORGANIZATION`, `LOCATION`, and so on). * Adding, renaming, or deleting items in the list of predefined relationships (such as `works_for`, `based_in`, `has_role`, and so on). * As needed, adding any clarifying instructions only between these two lines: ```text theme={null} ... Provide the entities and their corresponding types as a structured JSON response. (Add any clarifying instructions here only.) [START OF TEXT] ... ``` * Changing any other portions of the internal default prompt could produce unexpected results. Allowed values for `` include: * `openai_ner` * `anthropic_ner` The preceding list applies only to Unstructured **Let's Go** and **Pay-As-You-Go** accounts. For Unstructured **Business** accounts, to get your current list of available `` values, contact your Unstructured account administrator or Unstructured sales representative, or email Unstructured Support at [support@unstructured.io](mailto:support@unstructured.io). #### Generative OCR task Unstructured can produce generative OCR optimizations for workflows that are configured as follows: * With a **Partitioner** node set to use the **Auto** or **High Res** partitioning strategy, and a generative OCR optimizations node is added. * With a **Partitioner** node set to use the **VLM** partitioning strategy. No generative OCR optimization node is needed (or allowed). Unstructured never produces generative OCR optimizations for workflows with a **Partitioner** node set to use the **Fast** partitioning strategy. Generative OCR does not process any text blocks by default. You must also explicitly specify which document element types containing text that you want generative OCR to process. To do this, in a workflow's [High Res](#high-res-strategy) partitioner node, add each [document element type](/ui/document-elements#element-type) that you want generative OCR to process to the `extract_image_block_types` list within the partitioner node's `settings` field definition. Generative OCR does not process the text of any `Image` or `Table` elements if they have already been processed by [image description](#image-description-task) or [table description](#table-description-task) enrichments, respectively. ```python theme={null} generative_ocr_enrichment_workflow_node = WorkflowNode( name="Enrichment", subtype="", type="prompter", settings={ "provider_type": "", "model": "" } ) ``` ```json theme={null} { "name": "Enrichment", "type": "prompter", "subtype": "", "settings": { "provider_type": "", "model": "" } } ``` Allowed values for ``, ``, and `` include, respectively: * `anthropic_ocr`, `anthropic`, and: * `claude-opus-4-5-20251101` * `claude-3-7-sonnet-20250219` * `claude-sonnet-4-20250514` * `claude-sonnet-4-5-20250929` * `bedrock_ocr`, `bedrock`, and: * `us.amazon.nova-lite-v1:0` * `us.amazon.nova-pro-v1:0` * `us.anthropic.claude-3-haiku-20240307-v1:0` * `us.anthropic.claude-3-opus-20240229-v1:0` * `us.anthropic.claude-3-sonnet-20240229-v1:0` * `us.anthropic.claude-3-7-sonnet-20250219-v1:0` * `us.anthropic.claude-sonnet-4-20250514-v1:0` * `us.anthropic.claude-sonnet-4-5-20250929-v1:0` * `openai_ocr`, `openai`, and: * `gpt-4o` * `gpt-4o-mini` * `gpt-5-mini` The preceding list applies only to Unstructured **Let's Go** and **Pay-As-You-Go** accounts. For Unstructured **Business** accounts, to get your current list of available `` values, contact your Unstructured account administrator or Unstructured sales representative, or email Unstructured Support at [support@unstructured.io](mailto:support@unstructured.io). ### Chunker node A **Chunker** node has a `type` of `chunk`. [Learn about the available chunking strategies](/ui/chunking). #### Chunk by Character strategy ```python theme={null} chunk_by_character_chunker_workflow_node = WorkflowNode( name="Chunker", subtype="chunk_by_character", type="chunk", settings={ "unstructured_api_url": None, "unstructured_api_key": None, "include_orig_elements": , "new_after_n_chars": , "max_characters": , "overlap": , "overlap_all": , "contextual_chunking_strategy": "" } ) ``` ```json theme={null} { "name": "Chunker", "type": "chunk", "subtype": "chunk_by_character", "settings": { "unstructured_api_url": null, "unstructured_api_key": null, "include_orig_elements": , "new_after_n_chars": , "max_characters": , "overlap": , "overlap_all": , "contextual_chunking_strategy": "" } } ``` Fields for `settings` include: * `unstructured_api_url`: *Optional*. If specified, use a non-default API URL for calls to the specified chunker as needed. The default is none, which means to rely on using Unstructured's internal default API URL for the chunker. * `unstructured_api_key`: *Optional*. If specified, use a non-default API key for calls to the specified chunker as needed. The default is none, which means to rely on using Unstructured's internal default API key for the chunker. * `include_orig_elements`: *Optional*. True to have the elements that are used to form a chunk appear in `.metadata.orig_elements` for that chunk. The default is false. * `new_after_n_chars`: *Optional*. Closes new sections after reaching a length of this many characters. This is an approximate limit. The default is none. * `max_characters`: *Optional*. The absolute maximum number of characters in a chunk. The default is none. * `overlap`: *Optional*. Applies a prefix of this many trailing characters from the prior text-split chunk to second and later chunks formed from oversized elements by text-splitting. The default is none. * `overlap_all`: *Optional*. True to apply overlap to "normal" chunks formed by combining whole elements. Use with caution as this can introduce noise into otherwise clean semantic units. The default is false. * `contextual_chunking_strategy`: *Optional*. If specified, prepends chunk-specific explanatory context to each chunk. Allowed values include `v1`. The default is none. #### Chunk by Title strategy ```python theme={null} chunk_by_title_chunker_workflow_node = WorkflowNode( name="Chunker", subtype="chunk_by_title", type="chunk", settings={ "unstructured_api_url": None, "unstructured_api_key": None, "multipage_sections": , "combine_text_under_n_chars": , "include_orig_elements": , "new_after_n_chars": , "max_characters": , "overlap": , "overlap_all": , "contextual_chunking_strategy": "" } ) ``` ```json theme={null} { "name": "Chunker", "type": "chunk", "subtype": "chunk_by_title", "settings": { "unstructured_api_url": null, "unstructured_api_key": null, "multipage_sections": , "combine_text_under_n_chars": , "include_orig_elements": , "new_after_n_chars": , "max_characters": , "overlap": , "overlap_all": , "contextual_chunking_strategy": "" } } ``` Fields for `settings` include: * `unstructured_api_url`: *Optional*. If specified, use a non-default API URL for calls to the specified chunker as needed. The default is none, which means to rely on using Unstructured's internal default API URL for the chunker. * `unstructured_api_key`: *Optional*. If specified, use a non-default API key for calls to the specified chunker as needed. The default is none, which means to rely on using Unstructured's internal default API key for the chunker.- `multipage_sections`: *Optional*. ... The default is false. * `combine_text_under_n_chars`: *Optional*. Combines elements from a section into a chunk until a section reaches a length of this many characters. The default is none. * `include_orig_elements`: *Optional*. True to have the elements that are used to form a chunk appear in `.metadata.orig_elements` for that chunk. The default is false. * `new_after_n_chars`: *Optional*. Closes new sections after reaching a length of this many characters. This is an approximate limit. The default is none. * `max_characters`: *Optional*. The absolute maximum number of characters in a chunk. The default is none. * `overlap`: *Optional*. Applies a prefix of this many trailing characters from the prior text-split chunk to second and later chunks formed from oversized elements by text-splitting. The default is none. * `overlap_all`: *Optional*. True to apply overlap to "normal" chunks formed by combining whole elements. Use with caution as this can introduce noise into otherwise clean semantic units. The default is false. * `contextual_chunking_strategy`: *Optional*. If specified, prepends chunk-specific explanatory context to each chunk. Allowed values include `v1`. The default is none. #### Chunk by Page strategy ```python theme={null} chunk_by_page_chunker_workflow_node = WorkflowNode( name="Chunker", subtype="chunk_by_page", type="chunk", settings={ "unstructured_api_url": None, "unstructured_api_key": None, "include_orig_elements": , "new_after_n_chars": , "max_characters": , "overlap": , "overlap_all": , "contextual_chunking_strategy": "" } ) ``` ```json theme={null} { "name": "Chunker", "type": "chunk", "subtype": "chunk_by_page", "settings": { "unstructured_api_url": null, "unstructured_api_key": null, "include_orig_elements": , "new_after_n_chars": , "max_characters": , "overlap": , "overlap_all": , "contextual_chunking_strategy": "" } } ``` Fields for `settings` include: * `unstructured_api_url`: *Optional*. If specified, use a non-default API URL for calls to the specified chunker as needed. The default is none, which means to rely on using Unstructured's internal default API URL for the chunker. * `unstructured_api_key`: *Optional*. If specified, use a non-default API key for calls to the specified chunker as needed. The default is none, which means to rely on using Unstructured's internal default API key for the chunker.- `include_orig_elements`: *Optional*. ... The default is false. * `include_orig_elements`: *Optional*. True to have the elements that are used to form a chunk appear in `.metadata.orig_elements` for that chunk. The default is false. * `new_after_n_chars`: *Optional*. Closes new sections after reaching a length of this many characters. This is an approximate limit. The default is none. * `max_characters`: *Optional*. The absolute maximum number of characters in a chunk. The default is none. * `overlap`: *Optional*. Applies a prefix of this many trailing characters from the prior text-split chunk to second and later chunks formed from oversized elements by text-splitting. The default is none. * `overlap_all`: *Optional*. True to apply overlap to "normal" chunks formed by combining whole elements. Use with caution as this can introduce noise into otherwise clean semantic units. The default is false. * `contextual_chunking_strategy`: *Optional*. If specified, prepends chunk-specific explanatory context to each chunk. Allowed values include `v1`. The default is none. #### Chunk by Similarity strategy ```python theme={null} chunk_by_similarity_chunker_workflow_node = WorkflowNode( name="Chunker", subtype="chunk_by_similarity", type="chunk", settings={ "unstructured_api_url": None, "unstructured_api_key": None, "include_orig_elements": , "new_after_n_chars": , "max_characters": , "overlap": , "overlap_all": , "contextual_chunking_strategy": "", "similarity_threshold": } ) ``` ```json theme={null} { "name": "Chunker", "type": "chunk", "subtype": "chunk_by_similarity", "settings": { "unstructured_api_url": null, "unstructured_api_key": null, "include_orig_elements": , "new_after_n_chars": , "max_characters": , "overlap": , "overlap_all": , "contextual_chunking_strategy": "", "similarity_threshold": } } ``` Fields for `settings` include: * `unstructured_api_url`: *Optional*. If specified, use a non-default API URL for calls to the specified chunker as needed. The default is none, which means to rely on using Unstructured's internal default API URL for the chunker. * `unstructured_api_key`: *Optional*. If specified, use a non-default API key for calls to the specified chunker as needed. The default is none, which means to rely on using Unstructured's internal default API key for the chunker. * `include_orig_elements`: *Optional*. True to have the elements that are used to form a chunk appear in `.metadata.orig_elements` for that chunk. The default is false. * `new_after_n_chars`: *Optional*. Closes new sections after reaching a length of this many characters. This is an approximate limit. The default is none. * `max_characters`: *Optional*. The absolute maximum number of characters in a chunk. The default is none. * `overlap`: *Optional*. Applies a prefix of this many trailing characters from the prior text-split chunk to second and later chunks formed from oversized elements by text-splitting. The default is none. * `overlap_all`: *Optional*. True to apply overlap to "normal" chunks formed by combining whole elements. Use with caution as this can introduce noise into otherwise clean semantic units. The default is false. * `contextual_chunking_strategy`: *Optional*. If specified, prepends chunk-specific explanatory context to each chunk. Allowed values include `v1`. The default is none. * `similarity_threshold`: *Optional*. The minimum similarity that text in consecutive elements must have to be included in the same chunk. This must be a value between `0.0` and `1.0`, exclusive (`0.01` to `0.99`). The default is none. ### Embedder node An **Embedder** node has a `type` of `embed`. If you add an embedder node, you must set the workflow's chunker node's `max_characters` setting to a value at or below Unstructured's recommended maximum chunk size for your specified embedding model. [Learn more](/ui/embedding#chunk-sizing-and-embedding-models). [Learn about the available embedding providers and models](/ui/embedding). ```python theme={null} embedder_workflow_node = WorkflowNode( name="Embedder", subtype="", type="embed", settings={ "model_name": "" } ) ``` ```json theme={null} { "name": "Embedder", "type": "embed", "subtype": "", "settings": { "model_name": "" } } ``` Allowed values for `subtype` and `model_name` include the following: The following list applies only to Unstructured **Let's Go** and **Pay-As-You-Go** accounts. For Unstructured **Business** accounts, to get your current list of available models, contact your Unstructured account administrator or Unstructured sales representative, or email Unstructured Support at [support@unstructured.io](mailto:support@unstructured.io). * `"subtype": "azure_openai"` * `"model_name": "text-embedding-3-small"` * `"model_name": "text-embedding-3-large"` * `"model_name": "text-embedding-ada-002"` * `"subtype": "bedrock"` * `"model_name": "amazon.titan-embed-text-v2:0"` * `"model_name": "amazon.titan-embed-text-v1"` * `"model_name": "amazon.titan-embed-image-v1"` * `"model_name": "cohere.embed-english-v3"` * `"model_name": "cohere.embed-multilingual-v3"` * `"subtype": "ibm"` * `"model_name": "sentence-transformers/all-minilm-l6-v2"` * `"model_name": "ibm/granite-embedding-278m-multilingual"` * `"model_name": "intfloat/multilingual-e5-large"` * `"model_name": "ibm/slate-125m-english-rtrvr-v2"` * `"model_name": "ibm/slate-125m-multilingual-rtrvr-v2"` * `"subtype": "togetherai"` * `"model_name": "togethercomputer/m2-bert-80M-32k-retrieval"` * `"subtype": "voyageai"` * `"model_name": "voyage-3"` * `"model_name": "voyage-3-large"` * `"model_name": "voyage-3-lite"` * `"model_name": "voyage-code-3"` * `"model_name": "voyage-finance-2"` * `"model_name": "voyage-law-2"` * `"model_name": "voyage-code-2"` * `"model_name": "voyage-multimodal-3"` ### Extract node An **Extract** node has a `type` of `structured_data_extractor` and a `subtype` of `llm`. ```python theme={null} embedder_workflow_node = WorkflowNode( name="Extractor", subtype="llm", type="structured_data_extractor", settings={ "schema_to_extract": { "json_schema": "", "extraction_guidance": "" }, "provider": "", "model": "" } ) ``` ```json theme={null} { "name": "Extractor", "type": "structured_data_extractor", "subtype": "llm", "settings": { "schema_to_extract": { "json_schema": "", "extraction_guidance": "" }, "provider": "", "model": "" } } ``` Fields for `settings` include: * `schema_to_extract`: *Required*. The schema or guidance for the structured data that you want to extract. One (and only one) of the following must also be specified: * `json_schema`: The extraction schema, in [OpenAI Structured Outputs](https://platform.openai.com/docs/guides/structured-outputs#supported-schemas) format, for the structured data that you want to extract, expressed as a single string. * `extraction_guidance`: The extraction prompt for the structured data that you want to extract, expressed as a single string. * Allowed values for `provider` and `model` include the following: * `"provider": "anthropic"` * `"model": "claude-opus-4-5-20251101"` * `"model": "claude-sonnet-4-5-20250929"` * `"model": "claude-haiku-4-5-20251001"` * `"model": "claude-3-7-sonnet-20250219"` * `"model": "claude-sonnet-4-20250514"` * `"provider": "azure_openai"` * `"model": "gpt-5-mini"` * `"model": "gpt-4o"` * `"model": "gpt-4o-mini"` * `"provider": "bedrock"` * `"model": "us.anthropic.claude-opus-4-20250514-v1:0"` * `"model": "us.anthropic.claude-sonnet-4-20250514-v1:0"` * `"model": "us.anthropic.claude-3-7-sonnet-20250219-v1:0"` * `"model": "us.anthropic.claude-sonnet-4-5-20250929-v1:0"` * `"provider": "openai"` * `"model": "gpt-4o"` * `"model": "gpt-5-mini"` * `"model": "gpt-4o-mini"` [Learn more](/ui/data-extractor). ## List templates To list templates, use the `UnstructuredClient` object's `templates.list_templates` function (for the Python SDK) or the `GET` method to call the `/templates` endpoint (for `curl` or Postman). ```python theme={null} import os import json from unstructured_client import UnstructuredClient from unstructured_client.models.operations import ListTemplatesRequest client = UnstructuredClient( api_key_auth=os.getenv("UNSTRUCTURED_API_KEY") ) response = client.templates.list_templates( request=ListTemplatesRequest() ) print(json.dumps(response.response_list_templates, indent=4)) ``` ```python theme={null} import os import json import asyncio from unstructured_client import UnstructuredClient from unstructured_client.models.operations import ListTemplatesRequest async def list_templates(): client = UnstructuredClient( api_key_auth=os.getenv("UNSTRUCTURED_API_KEY") ) response = client.templates.list_templates( request=ListTemplatesRequest() ) print(json.dumps(response.response_list_templates, indent=4)) asyncio.run(list_templates()) ``` ```bash theme={null} curl --request 'GET' --location \ "$UNSTRUCTURED_API_URL/templates" \ --header "unstructured-api-key: $UNSTRUCTURED_API_KEY" \ --header 'accept: application/json' ``` 1. In the method drop-down list, select **GET**. 2. In the address box, enter the following URL: ```text theme={null} {{UNSTRUCTURED_API_URL}}/templates ``` 3. On the **Headers** tab, enter the following headers: * **Key**: `unstructured-api-key`, **Value**: `{{UNSTRUCTURED_API_KEY}}` * **Key**: `accept`, **Value**: `application/json` 4. Click **Send**. ## Get a template To get information about a template, use the `UnstructuredClient` object's `templates.get_template` function (for the Python SDK) or the `GET` method to call the `/templates/` endpoint (for `curl` or Postman), replacing `` with the template's unique ID. To get this ID, see [List templates](#list-templates). ```python theme={null} import os import json from unstructured_client import UnstructuredClient from unstructured_client.models.operations import GetTemplateRequest client = UnstructuredClient( api_key_auth=os.getenv("UNSTRUCTURED_API_KEY") ) response = client.templates.get_template( request=GetTemplateRequest( template_id="" ) ) print(json.dumps(response.response_get_template, indent=4)) ``` ```python theme={null} import os import json import asyncio from unstructured_client import UnstructuredClient from unstructured_client.models.operations import GetTemplateRequest async def get_template(): client = UnstructuredClient( api_key_auth=os.getenv("UNSTRUCTURED_API_KEY") ) response = client.templates.get_template( request=GetTemplateRequest( template_id="" ) ) print(json.dumps(response.response_get_template, indent=4)) asyncio.run(get_template()) ``` ```bash theme={null} curl --request 'GET' --location \ "$UNSTRUCTURED_API_URL/templates/" \ --header "unstructured-api-key: $UNSTRUCTURED_API_KEY" \ --header 'accept: application/json' ``` 1. In the method drop-down list, select **GET**. 2. In the address box, enter the following URL: ```text theme={null} {{UNSTRUCTURED_API_URL}}/templates/ ``` 3. On the **Headers** tab, enter the following headers: * **Key**: `unstructured-api-key`, **Value**: `{{UNSTRUCTURED_API_KEY}}` * **Key**: `accept`, **Value**: `application/json` 4. Click **Send**.