# Query content using AI assistant with vector store access

Performs intelligent queries on content using OpenAI's assistant with vector store access.
Query Process: 1. Agent Selection: Uses the specified agent's instructions as the system prompt, or defaults to research assistant instructions 2. Content Filtering: Uses the provided filters to select relevant content items from the dataspace 3. Vector Store Access: Retrieves the dataspace's vector store containing document embeddings 4. AI Assistant Creation: Creates a temporary OpenAI assistant with access to the vector store 5. Query Execution: Processes the user's query against the filtered content using the AI assistant 6. Citation Parsing: Extracts file IDs from the AI response citations 7. Content Retrieval: Fetches full content objects for all cited documents
Agent System: - Optional Agent ID: Provide an agentId to use custom agent instructions as the system prompt - Default Behavior: Without an agent ID, uses built-in research assistant instructions - Agent Instructions: The agent's instructions field becomes the system prompt for the AI assistant - Research Agent: Look for an agent named "Research Assistant Agent" for the default research experience
Request Body: - filters: Advanced filtering options to select specific content (uses unified ContentFiltersSchema) - query: The question or query to ask about the filtered content - agentId: (Optional) UUID of the agent whose instructions to use as system prompt
Response Fields: - query: The original user query - response: AI assistant's response with citations in simplified format - analyzedDocumentCount: Number of documents that were analyzed and available for querying - citedContent: Array of full content objects that were actually cited in the AI response
Citation Format: The AI response includes citations in simplified format: 【0】, 【1】, etc., where numbers reference the citedContent array indices.
Performance Notes: - Creates and destroys a temporary OpenAI assistant for each query - Processes up to 9999 content items for analysis - Response time depends on query complexity and document count - Agent lookup adds minimal overhead when agent ID is provided

Endpoint: POST /content/{dataspaceId}/query
Version: 1.0.0
Security: bearerAuth

## Path parameters:

  - `dataspaceId` (string, required)
    Short dataspace ID with sds- prefix
    Example: "sds-abc12345"

## Request fields (application/json):

  - `query` (string, required)
    The question or query to ask about the filtered content
    Example: "What are the main themes discussed in these documents?"

  - `agentId` (string)
    Optional agent ID to use for the query. If provided, the agent's instructions will be used as the system prompt.
    Example: "01234567-89ab-cdef-0123-456789abcdef"

  - `context` (object)
    Optional context configuration for content sources and filters

  - `context.contentFilters` (object)
    Filters to apply when selecting content for querying

  - `context.contentFilters.contentId` (any)
    Filter by content ID

  - `context.contentFilters.categoryId` (any)
    Filter by category ID

  - `context.contentFilters.fileName` (any)
    Filter by file name

  - `context.contentFilters.fileFormat` (any)
    Filter by file format/MIME type

  - `context.contentFilters.createdAt` (any)
    Filter by creation date

  - `context.contentFilters.updatedAt` (any)
    Filter by last update date

  - `context.contentFilters.metadata` (object)
    Filter by metadata fields - key-value pairs where values can use advanced operators

  - `context.libraries` (array)
    Array of library IDs to include in search context
    Example: ["01234567-89ab-cdef-0123-456789abcdef"]

  - `context.includeWebSearchResults` (boolean)
    Whether to include web search results in addition to local document search. Default is false.

## Response 200 fields (application/json):

  - `query` (string, required)
    The original query that was asked
    Example: "What are the main themes discussed in these documents?"

  - `response` (string, required)
    The AI assistant's response to the query
    Example: "Based on the documents in your collection, the main themes include..."

  - `analyzedDocumentCount` (integer, required)
    Number of documents that were analyzed and available for querying
    Example: 15

  - `citedContent` (array, required)
    Array of content items that were cited in the AI response, including citation numbers
    Example: []

  - `citedContent.contentId` (string, required)
    Unique identifier for the content item
    Example: "123e4567-e89b-12d3-a456-426614174000"

  - `citedContent.dataspaceId` (string, required)
    Short dataspace ID with sds- prefix
    Example: "sds-abc12345"

  - `citedContent.fileName` (string, required)
    Name of the uploaded file
    Example: "document.pdf"

  - `citedContent.fileFormat` (string, required)
    MIME type of the file
    Example: "application/pdf"

  - `citedContent.createdAt` (string, required)
    ISO timestamp when the content was created
    Example: "2024-01-15T10:30:00Z"

  - `citedContent.updatedAt` (string, required)
    ISO timestamp when the content was last updated
    Example: "2024-01-15T14:30:00Z"

  - `citedContent.categoryId` (string,null)
    UUID of the category this content belongs to
    Example: "123e4567-e89b-12d3-a456-426614174001"

  - `citedContent.metadata` (object,null)
    Optional metadata for the content item
    Example: {"author":"John Doe","version":1,"published":true}

  - `citedContent.inferenceTaskExecutions` (object,null)
    Record of inference task executions that modified this content, mapping inference type to task execution ID
    Example: {"categoryId":"789e0123-e89b-12d3-a456-426614174002"}

  - `citedContent.versionString` (string,null)
    Version identifier based on category and unique metadata
    Example: "research_v2_final"

  - `citedContent.fileUrl` (string,null)
    Public URL to access the file
    Example: "https://api.example.com/files/sds-abc12345/content/123e4567-e89b-12d3-a456-426614174000/document.pdf"

  - `citedContent.fileSize` (integer,null)
    Size of the file in bytes
    Example: 2048576

  - `citedContent.taskExecutionDetails` (object,null)
    Detailed information about task executions
    Example: {"categoryId":{"status":"completed","confidence":"high"}}

  - `citedContent.relatedVersions` (array,null)
    Related content items with the same version string
    Example: []

  - `citedContent.citationNumber` (integer)
    Citation number used in the AI response (e.g., [1], [2], etc.)
    Example: 1

  - `webSearchResults` (array, required)
    Array of web search results that were included in the AI response
    Example: []

  - `webSearchResults.title` (string, required)
    Title of the web search result
    Example: "Sample Article Title"

  - `webSearchResults.url` (string, required)
    URL of the web search result
    Example: "https://example.com/article"

  - `webSearchResults.snippet` (string, required)
    Snippet/preview text from the web search result
    Example: "This article discusses..."

  - `webSearchResults.relevanceScore` (number, required)
    Relevance score from 0 to 1
    Example: 0.85

  - `webSearchResults.citationNumber` (integer)
    Citation number used in the AI response (e.g., [100], [101], etc.)
    Example: 100

  - `citedLibraryPages` (array, required)
    Array of library pages that were cited in the AI response
    Example: []

  - `citedLibraryPages.id` (string, required)
    Unique identifier for the source file
    Example: "01234567-89ab-cdef-0123-456789abcdef"

  - `citedLibraryPages.libraryId` (string, required)
    ID of the library this page belongs to
    Example: "01234567-89ab-cdef-0123-456789abcdef"

  - `citedLibraryPages.url` (string, required)
    URL of the original web page
    Example: "https://example.com/page"

  - `citedLibraryPages.title` (string)
    Title of the web page
    Example: "Example Page Title"

  - `citedLibraryPages.domain` (string)
    Domain of the web page
    Example: "example.com"

  - `citedLibraryPages.contentType` (string)
    Content type of the page
    Example: "text/html"

  - `citedLibraryPages.scrapedAt` (string)
    When the page was scraped
    Example: "2024-01-01T00:00:00Z"

  - `citedLibraryPages.relevanceScore` (number)
    Relevance score for this citation
    Example: 0.92

  - `citedLibraryPages.citationNumber` (integer)
    Citation number used in the AI response
    Example: 2


## Response 400 fields

## Response 404 fields

## Response 500 fields