ai_invoice_parser

Instructions

AI Invoice Parser: Extracts data from invoices using AI.
Ref: https://developer.pdf.co/api-reference/ai-invoice-parser.md

Implementation Reference

• pdfco/mcp/tools/apis/extraction.py:8-28 (handler)

  The main handler function for the 'ai_invoice_parser' MCP tool. It takes a URL and optional API key, constructs ConversionParams, and calls parse_invoice().

  @mcp.tool()
  async def ai_invoice_parser(
      url: str = Field(
          description="URL to the source PDF file. Supports publicly accessible links including Google Drive, Dropbox, PDF.co Built-In Files Storage. Use 'upload_file' tool to upload local files."
      ),
      api_key: str = Field(
          description="PDF.co API key. If not provided, will use X_API_KEY environment variable. (Optional)",
          default="",
      ),
  ) -> BaseResponse:
      """
      AI Invoice Parser: Extracts data from invoices using AI.
      Ref: https://developer.pdf.co/api-reference/ai-invoice-parser.md
      """
  
      # Pass arguments directly; ConversionParams now handles str with default=None
      params = ConversionParams(
          url=url,
      )
  
      return await parse_invoice(params, api_key=api_key)

• pdfco/mcp/services/pdf.py:115-116 (helper)

  Helper function parse_invoice() that makes the actual API request to the 'ai-invoice-parser' endpoint.

  async def parse_invoice(params: ConversionParams, api_key: str | None = None) -> BaseResponse:
      return await request("ai-invoice-parser", params, api_key=api_key)

• pdfco/mcp/services/pdf.py:125-153 (helper)

  The lower-level request() helper that builds the payload, calls the API via PDFCoClient, and returns a BaseResponse.

  async def request(
      endpoint: str,
      params: ConversionParams,
      custom_payload: dict | None = None,
      api_key: str | None = None,
  ) -> BaseResponse:
      payload = params.parse_payload(async_mode=True)
      if custom_payload:
          payload.update(custom_payload)
  
      try:
          async with PDFCoClient(api_key=api_key) as client:
              url = f"/v1/{endpoint}"
              print(f"Requesting {url} with payload {payload}", file=sys.stderr)
              response = await client.post(url, json=payload)
              print(f"response: {response}", file=sys.stderr)
              json_data = response.json()
              return BaseResponse(
                  status="working",
                  content=json_data,
                  credits_used=json_data.get("credits"),
                  credits_remaining=json_data.get("remainingCredits"),
                  tips=f"You **should** use the 'wait_job_completion' tool to wait for the job [{json_data.get('jobId')}] to complete if a jobId is present.",
              )
      except Exception as e:
          return BaseResponse(
              status="error",
              content=f"{type(e)}: {[arg for arg in e.args if arg]}",
          )

• pdfco/mcp/models.py:5-159 (schema)

  The BaseResponse model used as the return type, and ConversionParams model used to hold input parameters.

  class BaseResponse(BaseModel):
      status: str
      content: Any
      credits_used: int | None = None
      credits_remaining: int | None = None
      tips: str | None = None
  
  
  class ConversionParams(BaseModel):
      url: str = Field(
          description="URL to the source file. Supports publicly accessible links including Google Drive, Dropbox, PDF.co Built-In Files Storage. Use 'upload_file' tool to upload local files.",
          default="",
      )
      httpusername: str = Field(
          description="HTTP auth user name if required to access source url. (Optional)",
          default="",
      )
      httppassword: str = Field(
          description="HTTP auth password if required to access source url. (Optional)",
          default="",
      )
      pages: str = Field(
          description="Comma-separated page indices (e.g., '0, 1, 2-' or '1, 3-7'). Use '!' for inverted page numbers (e.g., '!0' for last page). Processes all pages if None. (Optional)",
          default="",
      )
      unwrap: bool = Field(
          description="Unwrap lines into a single line within table cells when lineGrouping is enabled. Must be true or false. (Optional)",
          default=False,
      )
      rect: str = Field(
          description="Defines coordinates for extraction (e.g., '51.8,114.8,235.5,204.0'). (Optional)",
          default="",
      )
      lang: str = Field(
          description="Language for OCR for scanned documents. Default is 'eng'. See PDF.co docs for supported languages. (Optional, Default: 'eng')",
          default="eng",
      )
      line_grouping: str = Field(
          description="Enables line grouping within table cells when set to '1'. (Optional)",
          default="0",
      )
      password: str = Field(
          description="Password of the PDF file. (Optional)", default=""
      )
      name: str = Field(
          description="File name for the generated output. (Optional)", default=""
      )
      autosize: bool = Field(
          description="Controls automatic page sizing. If true, page dimensions adjust to content. If false, uses worksheet’s page setup. (Optional)",
          default=False,
      )
  
      html: str = Field(
          description="Input HTML code to be converted. To convert the link to a PDF use the /pdf/convert/from/url endpoint instead.",
          default="",
      )
      templateId: str = Field(
          description="Set to the ID of your HTML template. You can find and copy the ID from HTML to PDF Templates.",
          default="",
      )
      templateData: str = Field(
          description="Set it to a string with input JSON data (recommended) or CSV data.",
          default="",
      )
      margins: str = Field(
          description="Set to CSS style margins like 10px, 5mm, 5in for all sides or 5px 5px 5px 5px (the order of margins is top, right, bottom, left). (Optional)",
          default="",
      )
      paperSize: str = Field(
          description="A4 is set by default. Can be Letter, Legal, Tabloid, Ledger, A0, A1, A2, A3, A4, A5, A6 or a custom size. Custom size can be set in px (pixels), mm or in (inches) with width and height separated by space like this: 200 300, 200px 300px, 200mm 300mm, 20cm 30cm or 6in 8in. (Optional)",
          default="",
      )
      orientation: str = Field(
          description="Set to Portrait or Landscape. Portrait is set by default. (Optional)",
          default="",
      )
      printBackground: bool = Field(
          description="true by default. Set to false to disable printing of background. (Optional)",
          default=True,
      )
      mediaType: str = Field(
          description="Uses print by default. Set to screen to convert HTML as it appears in a browser or print to convert as it appears for printing or none to set none as mediaType for CSS styles. (Optional)",
          default="",
      )
      DoNotWaitFullLoad: bool = Field(
          description="false by default. Set to true to skip waiting for full load (like full video load etc. that may affect the total conversion time). (Optional)",
          default=False,
      )
      header: str = Field(
          description="User definable HTML for the header to be applied on every page header. (Optional)",
          default="",
      )
      footer: str = Field(
          description="User definable HTML for the footer to be applied on every page footer. (Optional)",
          default="",
      )
  
      worksheetIndex: str = Field(
          description="Index of the worksheet to convert. (Optional)", default=""
      )
  
      def parse_payload(self, async_mode: bool = True):
          payload = {
              "async": async_mode,
          }
          if self.url:
              payload["url"] = self.url
          if self.httpusername:
              payload["httpusername"] = self.httpusername
          if self.httppassword:
              payload["httppassword"] = self.httppassword
          if self.pages:
              payload["pages"] = self.pages
          if self.unwrap:
              payload["unwrap"] = self.unwrap
          if self.rect:
              payload["rect"] = self.rect
          if self.lang:
              payload["lang"] = self.lang
          if self.line_grouping:
              payload["lineGrouping"] = self.line_grouping
          if self.password:
              payload["password"] = self.password
          if self.name:
              payload["name"] = self.name
          if self.autosize:
              payload["autosize"] = self.autosize
  
          if self.html:
              payload["html"] = self.html
          if self.templateId:
              payload["templateId"] = self.templateId
          if self.templateData:
              payload["templateData"] = self.templateData
          if self.margins:
              payload["margins"] = self.margins
          if self.paperSize:
              payload["paperSize"] = self.paperSize
          if self.orientation:
              payload["orientation"] = self.orientation
          if self.printBackground:
              payload["printBackground"] = self.printBackground
          if self.mediaType:
              payload["mediaType"] = self.mediaType
          if self.DoNotWaitFullLoad:
              payload["DoNotWaitFullLoad"] = self.DoNotWaitFullLoad
          if self.header:
              payload["header"] = self.header
          if self.footer:
              payload["footer"] = self.footer
  
          if self.worksheetIndex:
              payload["worksheetIndex"] = self.worksheetIndex
  
          return payload

• pdfco/mcp/server.py:1-3 (registration)

  The MCP server instance created via FastMCP, which registers the tool via the @mcp.tool() decorator on the handler function.

  from fastmcp import FastMCP
  
  mcp = FastMCP("pdfco")