Mevzuat MCP

Instructions

Implementation Reference

• mevzuat_mcp_server.py:954-1009 (handler)

  The handler and registration for the 'get_cbgenelge_content' MCP tool using FastMCP @app.tool() decorator. Defines input schema via Pydantic Field descriptions, logs parameters, delegates to MevzuatApiClientNew.get_content with mevzuat_tur=22, handles errors, and returns MevzuatArticleContent.

  @app.tool()
  async def get_cbgenelge_content(
      mevzuat_no: str = Field(
          ...,
          description="The Presidential Circular number from search results (e.g., '16', '15')"
      ),
      resmi_gazete_tarihi: str = Field(
          ...,
          description="Official Gazette date from search results in DD/MM/YYYY format (e.g., '20/09/2025')"
      ),
      mevzuat_tertip: str = Field(
          "5",
          description="Circular series from search results (e.g., '5')"
      )
  ) -> MevzuatArticleContent:
      """
      Retrieve the full content of a Turkish Presidential Circular (Cumhurbaşkanlığı Genelgesi) in Markdown format.
  
      This tool fetches the PDF document and converts it to Markdown.
      Presidential Circulars are available only as PDF files.
      Use 'search_cbgenelge' first to find the circular number and Official Gazette date.
  
      IMPORTANT: You must provide the 'resmi_gazete_tarihi' (Official Gazette date) from the search results.
      This is required to construct the correct PDF URL.
  
      Returns:
      - Full circular content formatted as Markdown (converted from PDF)
      - Ready for analysis, summarization, or question answering
  
      Example usage:
      1. Search for circulars: search_cbgenelge(baslangic_tarihi="2025")
      2. Get full content: get_cbgenelge_content(mevzuat_no="16", resmi_gazete_tarihi="20/09/2025", mevzuat_tertip="5")
      """
      logger.info(f"Tool 'get_cbgenelge_content' called: {mevzuat_no}, RG date: {resmi_gazete_tarihi}, tertip: {mevzuat_tertip}")
  
      try:
          result = await mevzuat_client.get_content(
              mevzuat_no=mevzuat_no,
              mevzuat_tur=22,  # Cumhurbaşkanlığı Genelgesi
              mevzuat_tertip=mevzuat_tertip,
              resmi_gazete_tarihi=resmi_gazete_tarihi
          )
  
          if result.error_message:
              logger.warning(f"Error fetching circular content: {result.error_message}")
  
          return result
  
      except Exception as e:
          logger.exception(f"Error in tool 'get_cbgenelge_content' for {mevzuat_no}")
          return MevzuatArticleContent(
              madde_id=mevzuat_no,
              mevzuat_id=mevzuat_no,
              markdown_content="",
              error_message=f"An unexpected error occurred: {str(e)}"
          )

• mevzuat_client.py:557-724 (helper)

  Core helper function MevzuatApiClientNew.get_content that implements the PDF download and markdown conversion for CB Genelgesi (tur=22). Constructs special PDF URL using formatted resmi_gazete_tarihi, downloads PDF, optionally uses Mistral OCR for text extraction from images, falls back to markitdown PDF conversion, supports caching.

  async def get_content(
      self,
      mevzuat_no: str,
      mevzuat_tur: int = 1,
      mevzuat_tertip: str = "3",
      resmi_gazete_tarihi: Optional[str] = None
  ) -> MevzuatArticleContent:
      """
      Download and extract content from legislation.
      Tries HTML scraping first (most reliable), then falls back to file downloads.
      For Presidential Decisions (tur=20) and Circulars (tur=22), skip HTML and go directly to PDF.
  
      Args:
          mevzuat_no: Legislation number
          mevzuat_tur: Legislation type code (1=Kanun, 20=CB Kararı, 22=CB Genelgesi, etc.)
          mevzuat_tertip: Series number
          resmi_gazete_tarihi: Official Gazette date (DD/MM/YYYY) - required for CB Genelgesi (tur=22)
      """
      # CB Kararları (tur=20) and CB Genelgesi (tur=22) are PDF-only, skip HTML scraping
      if mevzuat_tur not in [20, 22]:
          # Try HTML scraping first (most reliable method for other types)
          result = await self.get_content_from_html(mevzuat_no, mevzuat_tur, mevzuat_tertip)
          if result.markdown_content:
              return result
  
          logger.info(f"HTML scraping returned no content for {mevzuat_no}, trying file downloads")
      else:
          if mevzuat_tur == 20:
              logger.info("CB Kararı detected (tur=20), skipping HTML scraping, going directly to PDF")
          elif mevzuat_tur == 22:
              logger.info("CB Genelgesi detected (tur=22), skipping HTML scraping, going directly to PDF")
  
      cache_key = f"doc:{mevzuat_tur}.{mevzuat_tertip}.{mevzuat_no}" if self._cache_enabled else None
  
      if cache_key and self._cache:
          cached_content = self._cache.get(cache_key)
          if cached_content:
              logger.debug(f"Cache hit: {mevzuat_no}")
              return MevzuatArticleContent(
                  madde_id=mevzuat_no,
                  mevzuat_id=mevzuat_no,
                  markdown_content=cached_content
              )
  
      # Construct URLs based on mevzuat type
      if mevzuat_tur == 22:  # CB Genelgesi - special PDF URL format
          if not resmi_gazete_tarihi:
              return MevzuatArticleContent(
                  madde_id=mevzuat_no,
                  mevzuat_id=mevzuat_no,
                  markdown_content="",
                  error_message="resmi_gazete_tarihi is required for CB Genelgesi (tur=22)"
              )
          # Convert DD/MM/YYYY to YYYYMMDD
          parts = resmi_gazete_tarihi.split('/')
          if len(parts) == 3:
              date_str = f"{parts[2]}{parts[1].zfill(2)}{parts[0].zfill(2)}"
          else:
              return MevzuatArticleContent(
                  madde_id=mevzuat_no,
                  mevzuat_id=mevzuat_no,
                  markdown_content="",
                  error_message=f"Invalid date format: {resmi_gazete_tarihi}. Expected DD/MM/YYYY"
              )
          pdf_url = self.GENELGE_PDF_URL_TEMPLATE.format(date=date_str, no=mevzuat_no)
          doc_url = None  # Genelge has no DOC version
      elif mevzuat_tur == 20:  # CB Kararı - PDF only, no DOC
          doc_url = None  # CB Kararları have no DOC version
          pdf_url = self.PDF_URL_TEMPLATE.format(tur=mevzuat_tur, tertip=mevzuat_tertip, no=mevzuat_no)
      else:
          doc_url = self.DOC_URL_TEMPLATE.format(tur=mevzuat_tur, tertip=mevzuat_tertip, no=mevzuat_no)
          pdf_url = self.PDF_URL_TEMPLATE.format(tur=mevzuat_tur, tertip=mevzuat_tertip, no=mevzuat_no)
  
      # Try DOC first (skip for CB Genelgesi and CB Kararı which have no DOC version)
      if doc_url:
          try:
              logger.info(f"Trying DOC: {doc_url}")
              # Use separate headers for document download
              doc_headers = {
                  'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36',
                  'Accept': 'application/msword, */*',
              }
              response = await self._http_client.get(doc_url, headers=doc_headers)
              response.raise_for_status()
  
              doc_bytes = response.content
              logger.info(f"Downloaded DOC: {len(doc_bytes)} bytes")
  
              # DOC files from mevzuat.gov.tr are actually HTML
              if len(doc_bytes) < 100:
                  logger.warning(f"DOC file too small ({len(doc_bytes)} bytes), likely empty")
                  raise Exception("DOC file is empty or too small")
  
              doc_stream = io.BytesIO(doc_bytes)
              result = self._md_converter.convert_stream(doc_stream, file_extension=".doc")
              markdown_content = result.text_content.strip() if result and result.text_content else ""
  
              if markdown_content:
                  logger.info(f"DOC conversion successful for {mevzuat_no}")
                  if cache_key and self._cache:
                      self._cache.put(cache_key, markdown_content)
                  return MevzuatArticleContent(
                      madde_id=mevzuat_no,
                      mevzuat_id=mevzuat_no,
                      markdown_content=markdown_content
                  )
          except Exception as e:
              logger.info(f"DOC failed, trying PDF: {e}")
  
      # Try PDF fallback
      try:
          logger.info(f"Trying PDF: {pdf_url}")
  
          # For CB Kararı (tur=20) and CB Genelgesi (tur=22), ensure we have session cookies
          if mevzuat_tur in [20, 22]:
              await self._ensure_session()
              # Create temporary client with cookies to avoid deprecation warning
              async with httpx.AsyncClient(
                  headers=self.HEADERS,
                  cookies=self._cookies,
                  timeout=self._http_client.timeout,
                  follow_redirects=True
              ) as temp_client:
                  response = await temp_client.get(pdf_url)
          else:
              response = await self._http_client.get(pdf_url)
  
          response.raise_for_status()
  
          pdf_bytes = response.content
          markdown_content = ""
  
          # For CB Kararı (tur=20) and CB Genelgesi (tur=22), use Mistral OCR (handles images + text)
          if mevzuat_tur in [20, 22] and self._mistral_client:
              doc_type = "CB Kararı" if mevzuat_tur == 20 else "CB Genelgesi"
              logger.info(f"Using Mistral OCR for {doc_type} PDF")
              markdown_content = await self._ocr_pdf_with_mistral(pdf_bytes, pdf_url)
  
              # Fallback to markitdown if OCR fails
              if not markdown_content:
                  logger.warning("Mistral OCR failed, falling back to markitdown")
                  pdf_stream = io.BytesIO(pdf_bytes)
                  result = self._md_converter.convert_stream(pdf_stream, file_extension=".pdf")
                  markdown_content = result.text_content.strip() if result and result.text_content else ""
          else:
              # Use markitdown for other types
              pdf_stream = io.BytesIO(pdf_bytes)
              result = self._md_converter.convert_stream(pdf_stream, file_extension=".pdf")
              markdown_content = result.text_content.strip() if result and result.text_content else ""
  
          if markdown_content:
              logger.info(f"PDF conversion successful for {mevzuat_no}")
              if cache_key and self._cache:
                  self._cache.put(cache_key, markdown_content)
              return MevzuatArticleContent(
                  madde_id=mevzuat_no,
                  mevzuat_id=mevzuat_no,
                  markdown_content=markdown_content
              )
      except Exception as e:
          logger.error(f"PDF also failed: {e}")
  
      return MevzuatArticleContent(
          madde_id=mevzuat_no,
          mevzuat_id=mevzuat_no,
          markdown_content="",
          error_message=f"Both DOC and PDF download/conversion failed for {mevzuat_tur}.{mevzuat_tertip}.{mevzuat_no}"
      )

• mevzuat_models.py:119-125 (schema)

  Pydantic schema model MevzuatArticleContent returned by the tool, containing the extracted markdown_content of the Presidential Circular and optional error_message.

  class MevzuatArticleContent(BaseModel):
      """Model for the content of legislation (reused from old models)."""
      madde_id: str
      mevzuat_id: str
      markdown_content: str
      error_message: Optional[str] = None