Yargı MCP

Instructions

Implementation Reference

• mcp_server_main.py:1208-1222 (handler)

  MCP tool handler: Receives documentId parameter, validates it, and delegates to BedestenApiClient instance to fetch and return the document as markdown. This is the entry point for the tool.
  async def get_bedesten_document_markdown( documentId: str = Field(..., description="Document ID from Bedesten search results") ) -> BedestenDocumentMarkdown: """Get legal decision document as Markdown from Bedesten API.""" logger.info(f"Tool 'get_bedesten_document_markdown' called for ID: {documentId}") if not documentId or not documentId.strip(): raise ValueError("Document ID must be a non-empty string.") try: return await bedesten_client_instance.get_document_as_markdown(documentId) except Exception as e: logger.exception("Error in tool 'get_kyb_bedesten_document_markdown'") raise
• bedesten_mcp_module/client.py:83-147 (helper)

  BedestenApiClient.get_document_as_markdown: Fetches the document via API POST to /emsal-karar/getDocumentContent, decodes base64 content (HTML or PDF), converts to markdown using MarkItDown, handles errors, returns BedestenDocumentMarkdown.
  async def get_document_as_markdown(self, document_id: str) -> BedestenDocumentMarkdown: """ Get document content and convert to markdown. Handles both HTML (text/html) and PDF (application/pdf) content types. """ logger.info(f"BedestenApiClient: Fetching document for markdown conversion (ID: {document_id})") try: # Prepare request doc_request = BedestenDocumentRequest( data=BedestenDocumentRequestData(documentId=document_id) ) # Get document response = await self.http_client.post( self.DOCUMENT_ENDPOINT, json=doc_request.model_dump() ) response.raise_for_status() response_json = response.json() doc_response = BedestenDocumentResponse(**response_json) # Add null safety checks for document data if not hasattr(doc_response, 'data') or doc_response.data is None: raise ValueError("Document response does not contain data") if not hasattr(doc_response.data, 'content') or doc_response.data.content is None: raise ValueError("Document data does not contain content") if not hasattr(doc_response.data, 'mimeType') or doc_response.data.mimeType is None: raise ValueError("Document data does not contain mimeType") # Decode base64 content with error handling try: content_bytes = base64.b64decode(doc_response.data.content) except Exception as e: raise ValueError(f"Failed to decode base64 content: {str(e)}") mime_type = doc_response.data.mimeType logger.info(f"BedestenApiClient: Document mime type: {mime_type}") # Convert to markdown based on mime type if mime_type == "text/html": html_content = content_bytes.decode('utf-8') markdown_content = self._convert_html_to_markdown(html_content) elif mime_type == "application/pdf": markdown_content = self._convert_pdf_to_markdown(content_bytes) else: logger.warning(f"Unsupported mime type: {mime_type}") markdown_content = f"Unsupported content type: {mime_type}. Unable to convert to markdown." return BedestenDocumentMarkdown( documentId=document_id, markdown_content=markdown_content, source_url=f"{self.BASE_URL}/document/{document_id}", mime_type=mime_type ) except httpx.RequestError as e: logger.error(f"BedestenApiClient: HTTP error fetching document {document_id}: {e}") raise except Exception as e: logger.error(f"BedestenApiClient: Error processing document {document_id}: {e}") raise
• bedesten_mcp_module/models.py:87-91 (schema)

  Pydantic output schema/model for the tool response: BedestenDocumentMarkdown defining documentId, markdown_content, source_url, mime_type.
  class BedestenDocumentMarkdown(BaseModel): documentId: str = Field(..., description="The document ID (Belge Kimliği) from Bedesten") markdown_content: Optional[str] = Field(None, description="The decision content (Karar İçeriği) converted to Markdown") source_url: str = Field(..., description="The source URL (Kaynak URL) of the document") mime_type: Optional[str] = Field(None, description="Original content type (İçerik Türü) (text/html or application/pdf)")
• mcp_server_main.py:1201-1207 (registration)

  @app.tool decorator registers the get_bedesten_document_markdown function as an MCP tool with description and annotations.
  @app.tool( description="Get legal decision document from Bedesten API in Markdown format", annotations={ "readOnlyHint": True, "idempotentHint": True } )
• bedesten_mcp_module/client.py:192-195 (helper)

  BedestenApiClient.close_client_session: Closes the httpx.AsyncClient session (cleanup).
  async def close_client_session(self): """Close HTTP client session""" await self.http_client.aclose() logger.info("BedestenApiClient: HTTP client session closed.")