Paperlib MCP

Instructions

Implementation Reference

• src/paperlib_mcp/tools/fetch.py:363-438 (handler)

  The core handler function for the 'delete_document' tool. It handles deletion of the document from the database tables (ingest_jobs, documents, chunks, chunk_embeddings) and optionally deletes the associated PDF file from MinIO storage using delete_object. Returns a success dict with stats.

  @mcp.tool()
  def delete_document(
      doc_id: str,
      also_delete_object: bool = False,
  ) -> dict[str, Any]:
      """删除指定文档
      
      从数据库删除文档及其所有关联数据（chunks、embeddings、导入记录等）。
      可选择同时删除 MinIO 中的 PDF 文件。
      
      Args:
          doc_id: 文档的唯一标识符
          also_delete_object: 是否同时删除 MinIO 中的 PDF 文件，默认 False
          
      Returns:
          删除结果，包含删除的记录数量
      """
      try:
          # 先获取文档信息
          doc = query_one(
              "SELECT pdf_key FROM documents WHERE doc_id = %s",
              (doc_id,)
          )
          
          if not doc:
              return {
                  "success": False,
                  "error": f"Document not found: {doc_id}",
                  "doc_id": doc_id,
              }
          
          pdf_key = doc["pdf_key"]
          
          # 统计将要删除的数据
          stats = query_one(
              """
              SELECT 
                  (SELECT COUNT(*) FROM chunks WHERE doc_id = %s) as chunk_count,
                  (SELECT COUNT(*) FROM chunk_embeddings ce 
                   JOIN chunks c ON ce.chunk_id = c.chunk_id 
                   WHERE c.doc_id = %s) as embedding_count,
                  (SELECT COUNT(*) FROM ingest_jobs WHERE doc_id = %s) as job_count
              """,
              (doc_id, doc_id, doc_id)
          )
          
          # 删除导入记录
          execute("DELETE FROM ingest_jobs WHERE doc_id = %s", (doc_id,))
          
          # 删除文档（级联删除 chunks 和 embeddings）
          execute("DELETE FROM documents WHERE doc_id = %s", (doc_id,))
          
          result = {
              "success": True,
              "doc_id": doc_id,
              "deleted_chunks": stats["chunk_count"] if stats else 0,
              "deleted_embeddings": stats["embedding_count"] if stats else 0,
              "deleted_jobs": stats["job_count"] if stats else 0,
              "object_deleted": False,
          }
          
          # 可选删除 MinIO 对象
          if also_delete_object and pdf_key:
              delete_result = delete_object(pdf_key)
              result["object_deleted"] = delete_result.get("deleted", False)
              result["pdf_key"] = pdf_key
          
          return result
          
      except Exception as e:
          return {
              "success": False,
              "error": str(e),
              "doc_id": doc_id,
          }

• src/paperlib_mcp/server.py:15-36 (registration)

  Registration of the fetch tools, including delete_document, by calling register_fetch_tools(mcp) in the main MCP server setup.

  from paperlib_mcp.tools.fetch import register_fetch_tools
  from paperlib_mcp.tools.writing import register_writing_tools
  
  # M2 GraphRAG 工具
  from paperlib_mcp.tools.graph_extract import register_graph_extract_tools
  from paperlib_mcp.tools.graph_canonicalize import register_graph_canonicalize_tools
  from paperlib_mcp.tools.graph_community import register_graph_community_tools
  from paperlib_mcp.tools.graph_summarize import register_graph_summarize_tools
  from paperlib_mcp.tools.graph_maintenance import register_graph_maintenance_tools
  
  # M3 Review 工具
  from paperlib_mcp.tools.review import register_review_tools
  
  # M4 Canonicalization & Grouping 工具
  from paperlib_mcp.tools.graph_relation_canonicalize import register_graph_relation_canonicalize_tools
  from paperlib_mcp.tools.graph_claim_grouping import register_graph_claim_grouping_tools
  from paperlib_mcp.tools.graph_v12 import register_graph_v12_tools
  
  register_health_tools(mcp)
  register_import_tools(mcp)
  register_search_tools(mcp)
  register_fetch_tools(mcp)