> ## Documentation Index
> Fetch the complete documentation index at: https://composio-preview.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# VLM Run

> VLM Run provides multimodal agents, structured extraction, predictions, files, skills, feedback, and evaluation APIs.

![VLM Run logo](https://mintcdn.com/composio-preview/DMB8UrgJmE7r-ZQ5/images/toolkits/vlm_run.svg?fit=max\&auto=format\&n=DMB8UrgJmE7r-ZQ5\&q=85\&s=ac3f018c74c4e5ab75ba602226fa7f9f)

* **Category:** ai document extraction
* **Auth:** API\_KEY
* **Composio-managed OAuth available?** N/A
* **Tools:** 11
* **Triggers:** 0
* **Slug:** `VLM_RUN`
* **Version:** 00000000\_00

## Tools

> [accordion] Create Skill
>
> **Slug:** `VLM_RUN_CREATE_SKILL`
>
> Create a reusable skill from exactly one uploaded zip, prompt, or chat session.
>
> #### Input Parameters
>
> **name** `string`
>
> Skill name; required with file\_id and optional for prompt or session sources.
>
> **prompt** `string`
>
> Prompt that becomes SKILL.md; provide exactly one skill source.
>
> **file\_id** `string`
>
> Uploaded skill zip ID; provide exactly one of file\_id, prompt, or session\_id.
>
> **is\_public** `boolean`
>
> Whether the new skill is publicly accessible.
>
> **session\_id** `string`
>
> Chat session whose history becomes the skill source; provide exactly one skill source.
>
> **description** `string`
>
> Skill description; required with file\_id and optional for prompt or session sources.
>
> **json\_schema** `object`
>
> Optional output JSON schema; allowed only with a prompt source.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Discover Extraction Schemas
>
> **Slug:** `VLM_RUN_DISCOVER_EXTRACTION_SCHEMAS`
>
> List supported structured-extraction domains, or return the full JSON schema for one domain when domain is provided.
>
> #### Input Parameters
>
> **config** `object`
>
> Typed generation options to apply when resolving a domain schema. Used only when domain is provided.
>
> **domain** `string`
>
> Domain identifier to resolve, such as document.invoice. Omit it to list all supported extraction domains.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Execute Agent
>
> **Slug:** `VLM_RUN_EXECUTE_AGENT`
>
> Start a VLM Run agent execution from an existing agent name or inline configuration over multimodal inputs. Execution may consume credits and is asynchronous by default; poll the returned ID with VLM\_RUN\_GET\_RUN.
>
> #### Input Parameters
>
> **batch** `boolean`
>
> Run asynchronously when true.
>
> **model** `string`
>
> Orion model override; omit to use the skill or agent default.
>
> **config** `object`
>
> Inline execution behavior used to identify or configure the agent.
>
> **inputs** `object`
>
> Named typed media content or plain JSON values matching the agent input schema.
>
> **metadata** `object`
>
> Caller metadata and data-governance settings.
>
> **agent\_name** `string`
>
> Name and optional version of an existing agent, such as agent-name:latest.
>
> **request\_id** `string`
>
> Caller-selected unique request identifier.
>
> **callback\_url** `string`
>
> HTTPS URL called when asynchronous execution completes.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Extract Structured JSON
>
> **Slug:** `VLM_RUN_EXTRACT_STRUCTURED_JSON`
>
> Start structured JSON extraction from images, a document, a video, or audio using a domain, custom schema, or skill. Extraction may consume credits; document, video, and audio runs are asynchronous by default, so poll the returned prediction ID with VLM\_RUN\_GET\_RUN.
>
> #### Input Parameters
>
> **url** `string`
>
> Public media URL; for document, video, or audio use exactly one of url or file\_id.
>
> **batch** `boolean`
>
> Override asynchronous processing; omit to preserve the modality-specific provider default.
>
> **model** `string`
>
> Structured-extraction model or quality variant, such as vlm-1:fast.
>
> **config** `object`
>
> Generation, schema, skill, quality, and media options.
>
> **domain** `string`
>
> Extraction domain; required for audio and optional for other modalities when config supplies a skill or custom schema.
>
> **images** `array`
>
> Image URLs or base64-encoded images; required only for image modality.
>
> **file\_id** `string`
>
> Uploaded file ID; for document, video, or audio use exactly one of file\_id or url.
>
> **metadata** `object`
>
> Metadata and data-handling preferences for the prediction.
>
> **modality** `string ("image" | "document" | "video" | "audio")` *required*
>
> Media type to process; selects the matching generation endpoint.
>
> **request\_id** `string`
>
> Caller-selected request identifier, sent upstream as id.
>
> **callback\_url** `string`
>
> Callback URL for asynchronous completion delivery.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Find Files
>
> **Slug:** `VLM_RUN_FIND_FILES`
>
> List uploaded files or find one by file ID or MD5 hash. In list mode, use offset and limit until has\_more is false.
>
> #### Input Parameters
>
> **limit** `integer`
>
> Maximum files in list mode (1-1000).
>
> **offset** `integer`
>
> Zero-based list offset; sent upstream as skip and used only when no selector is supplied.
>
> **file\_id** `string`
>
> Exact uploaded file ID; mutually exclusive with md5\_hash.
>
> **md5\_hash** `string`
>
> Exact 32-character MD5 hex digest used for deduplication lookup; mutually exclusive with file\_id.
>
> **generate\_public\_url** `boolean`
>
> Whether ID lookup should generate a temporary public URL.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Find Skills
>
> **Slug:** `VLM_RUN_FIND_SKILLS`
>
> List VLM Run skills or find one exact skill by ID, name, and optional version. In list mode, continue from next\_offset while has\_more is true.
>
> #### Input Parameters
>
> **name** `string`
>
> Exact skill name; mutually exclusive with skill\_id.
>
> **limit** `integer`
>
> Maximum skills to return in list mode (1-1000).
>
> **offset** `integer`
>
> Zero-based list offset; used only when no selector is supplied.
>
> **grouped** `boolean`
>
> Request grouped skill results in list mode.
>
> **version** `string`
>
> Optional exact skill version used only with name.
>
> **order\_by** `string ("created_at" | "updated_at" | "name")`
>
> Sort field in list mode.
>
> **skill\_id** `string`
>
> Exact skill ID; mutually exclusive with name.
>
> **descending** `boolean`
>
> Sort descending in list mode.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Get Run
>
> **Slug:** `VLM_RUN_GET_RUN`
>
> Get the current status and result of one structured-extraction prediction or agent execution; call repeatedly to poll asynchronous work.
>
> #### Input Parameters
>
> **run\_id** `string` *required*
>
> Prediction or agent execution ID returned by a run or list tool.
>
> **run\_type** `string ("prediction" | "agent_execution")` *required*
>
> Kind of run to retrieve.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] List Agents
>
> **Slug:** `VLM_RUN_LIST_AGENTS`
>
> Return agents available to the connected account for selection before execution.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] List Artifacts
>
> **Slug:** `VLM_RUN_LIST_ARTIFACTS`
>
> List artifact metadata belonging to exactly one chat session or agent execution. Use offset and limit to traverse pages until has\_more is false.
>
> #### Input Parameters
>
> **limit** `integer`
>
> Maximum artifacts to return (1-1000).
>
> **offset** `integer`
>
> Zero-based artifact offset sent upstream as offset.
>
> **session\_id** `string`
>
> Parent chat session ID; provide exactly one of session\_id or execution\_id.
>
> **execution\_id** `string`
>
> Parent agent execution ID; provide exactly one of execution\_id or session\_id.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] List Runs
>
> **Slug:** `VLM_RUN_LIST_RUNS`
>
> List structured-extraction predictions or agent executions for the connected account. Use offset and limit to traverse pages until has\_more is false.
>
> #### Input Parameters
>
> **limit** `integer`
>
> Maximum runs to return (1-1000).
>
> **offset** `integer`
>
> Zero-based number of runs to skip; sent upstream as skip.
>
> **run\_type** `string ("prediction" | "agent_execution")` *required*
>
> Kind of run to list.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not

> [accordion] Upload File
>
> **Slug:** `VLM_RUN_UPLOAD_FILE`
>
> Upload a local file to VLM Run for extraction, agent input, or skill creation. Retain the returned file ID for tools that consume uploaded files.
>
> #### Input Parameters
>
> **file** `object` *required*
>
> Local file to upload through Composio's file store.
>
> **purpose** `string ("assistants" | "batch" | "fine-tune" | "vision" | "datasets")`
>
> Provider purpose assigned to the uploaded file.
>
> **generate\_public\_url** `boolean`
>
> Whether to include a temporary public URL in the response.
>
> #### Output
>
> **data** `string` *required*
>
> Data from the action execution
>
> **error** `string`
>
> Error if any occurred during the execution of the action
>
> **successful** `boolean` *required*
>
> Whether or not the action execution was successful or not
