SharePoint Search for a Document and Download It

arazzo: 1.0.1
info:
  title: SharePoint Search for a Document and Download It
  summary: Run a search query, take the top hit's path, and fetch the document bytes.
  description: >-
    The retrieval flow behind most SharePoint-backed assistants and RAG
    pipelines: ask the search index a question, then go get the winning
    document. The workflow keeps the search deliberately narrow with
    selectproperties so the result table is predictable, then downloads the top
    hit. Note the shape of SharePoint search results — each row is an array of
    Key/Value cells, so the Path cell is addressed positionally and depends on
    the selectproperties order declared below. Every step spells out its request
    inline so the flow can be read and executed without opening the underlying
    OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: filesApi
  url: ../openapi/sharepoint-files-api-openapi.yml
  type: openapi
- name: searchApi
  url: ../openapi/sharepoint-search-api-openapi.yml
  type: openapi
workflows:
- workflowId: search-and-download
  summary: Search SharePoint content and download the highest ranked matching document.
  description: >-
    Executes a keyword search restricted to a handful of properties, captures
    the top result's path, and downloads that file.
  inputs:
    type: object
    required:
    - queryText
    properties:
      queryText:
        type: string
        description: >-
          Keyword Query Language text, e.g. "quarterly report filetype:docx".
      rowLimit:
        type: integer
        description: Maximum number of search rows to return.
        default: 10
  steps:
  - stepId: runSearch
    description: >-
      Execute the search query. selectproperties is pinned to a fixed, ordered
      set so the returned cell positions are stable: cell 0 is Title, cell 1 is
      Path, cell 2 is FileType, cell 3 is LastModifiedTime. Change that list and
      the pointers in this step's outputs must change with it.
    operationId: searchQuery
    parameters:
    - name: querytext
      in: query
      value: $inputs.queryText
    - name: selectproperties
      in: query
      value: Title,Path,FileType,LastModifiedTime
    - name: rowlimit
      in: query
      value: $inputs.rowLimit
    successCriteria:
    - condition: $statusCode == 200
    - context: $response.body
      condition: $.PrimaryQueryResult.RelevantResults.RowCount > 0
      type: jsonpath
    outputs:
      totalRows: $response.body#/PrimaryQueryResult/RelevantResults/TotalRows
      rowCount: $response.body#/PrimaryQueryResult/RelevantResults/RowCount
      topResultCells: $response.body#/PrimaryQueryResult/RelevantResults/Table/Rows/0/Cells
      topResultTitle: $response.body#/PrimaryQueryResult/RelevantResults/Table/Rows/0/Cells/0/Value
      topResultPath: $response.body#/PrimaryQueryResult/RelevantResults/Table/Rows/0/Cells/1/Value
  - stepId: downloadTopResult
    description: >-
      Download the top ranked document using the Path cell from the search
      result. Search returns Path as an absolute URL, so a caller running this
      against a live tenant should reduce it to the server-relative form this
      operation expects.
    operationId: downloadFile
    parameters:
    - name: file_url
      in: path
      value: $steps.runSearch.outputs.topResultPath
    successCriteria:
    - condition: $statusCode == 200
  outputs:
    totalRows: $steps.runSearch.outputs.totalRows
    topResultTitle: $steps.runSearch.outputs.topResultTitle
    topResultPath: $steps.runSearch.outputs.topResultPath