Update biopython/mcp_output/mcp_plugin/mcp_service.py

biopython/mcp_output/mcp_plugin/mcp_service.py

@@ -359,38 +359,37 @@ def blast_get_result(payload: dict):
 @mcp.tool(name="blast_search", description="Search NCBI database using BLAST to find similar sequences")
 def blast_search(payload: dict):
     """
-    使用 BLAST 搜索相似序列
+    使用 BLAST 搜索相似序列（稳定版）
     Required fields: sequence
-    Optional fields: database (nt/nr/refseq_rna), program (blastn/blastp), hitlist_size (default 20), expect (default 1e-10)
-    
-    Database recommendations:
-    - refseq_rna: Faster, curated sequences (recommended for genes)
-    - nt: Comprehensive but VERY SLOW (can take 5-15 minutes)
-    - nr: For protein sequences
+    Optional fields: 
+      - database (nt/nr/refseq_rna, default refseq_rna)
+      - program (blastn/blastp, default blastn)
+      - hitlist_size (default 20)
+      - expect (default 1e-10)
     """
     try:
         sequence = str(payload.get("sequence", "")).strip()
         if not sequence:
             return {"success": False, "result": None, "error": "sequence is required"}
-        
-        # Clean sequence
+
+        # 清理序列
         sequence = "".join(sequence.split()).upper()
-        
-        # 默认使用更快的 refseq_rna 数据库
+
+        # 参数
         database = payload.get("database", "refseq_rna")
         program = payload.get("program", "blastn")
         hitlist_size = int(payload.get("hitlist_size", 20))
         expect = float(payload.get("expect", 1e-10))
-        
-        # 数据库选择建议
+
+        # 提示信息
         if database == "nt":
-            print("⚠️  Warning: 'nt' database is very large and may take 5-15 minutes!")
-            print("💡 Consider using 'refseq_rna' for faster results (curated sequences)")
-        
+            print("⚠️ Warning: 'nt' database is very large and may take 5-15 minutes!")
+            print("💡 Consider using 'refseq_rna' for faster results")
+
         print(f"🔍 Submitting BLAST search: {len(sequence)} bp, database={database}, hits={hitlist_size}")
-        print(f"⏳ Please wait... This may take 1-15 minutes depending on database size")
-        
-        # Submit BLAST request with timeout handling
+        print("⏳ Please wait...")
+
+        # 提交 BLAST
         try:
             handle = NCBIWWW.qblast(
                 program=program,
@@ -400,47 +399,45 @@ def blast_search(payload: dict):
                 expect=expect,
                 format_type="XML"
             )
-            
-            blast_record = NCBIXML.read(handle)
-            handle.close()
-            
-            print(f"✅ BLAST search completed: {len(blast_record.alignments)} alignments found")
         except Exception as blast_error:
-            # 如果使用 nt 失败，建议使用更快的数据库
-            if database == "nt":
-                return {
-                    "success": False,
-                    "result": None,
-                    "error": f"BLAST timeout or error with 'nt' database. Please try 'refseq_rna' instead. Original error: {str(blast_error)}"
-                }
-            else:
-                raise blast_error
-        
+            return {
+                "success": False,
+                "result": None,
+                "error": f"BLAST submission failed: {str(blast_error)}"
+            }
+
+        # 迭代解析 XML
         hits = []
-        for idx, alignment in enumerate(blast_record.alignments[:hitlist_size]):
-            print(idx)
-            print(alignment)
-            if not alignment.hsps:
-                continue
-            print("✅✅✅")
-            hsp = alignment.hsps[0]
-            align_len = int(hsp.align_length)
-            identities = int(hsp.identities)
-            identity_pct = (identities / align_len * 100.0) if align_len else 0
-            
-            hits.append({
-                "rank": idx + 1,
-                "accession": str(alignment.accession),
-                "hit_id": str(alignment.hit_id),
-                "hit_def": str(alignment.hit_def),
-                "length": int(alignment.length),
-                "bit_score": float(hsp.bits),
-                "evalue": float(hsp.expect),
-                "identity_pct": float(round(identity_pct, 2)),
-                "identities": int(identities),
-                "align_length": int(align_len)
-            })
-        print("✅✅✅✅✅✅")
+        for blast_record in NCBIXML.parse(handle):
+            for idx, alignment in enumerate(blast_record.alignments[:hitlist_size]):
+                if not alignment.hsps:
+                    continue
+
+                hsp = alignment.hsps[0]
+                align_len = int(hsp.align_length)
+                identities = int(hsp.identities)
+                identity_pct = round((identities / align_len) * 100, 2) if align_len else 0
+
+                hits.append({
+                    "rank": idx + 1,
+                    "accession": str(alignment.accession),
+                    "hit_id": str(alignment.hit_id),
+                    "hit_def": str(alignment.hit_def),
+                    "length": int(alignment.length),
+                    "bit_score": float(hsp.bits),
+                    "evalue": float(hsp.expect),
+                    "identity_pct": float(identity_pct),
+                    "identities": int(identities),
+                    "align_length": int(align_len)
+                })
+
+                # 输出进度，保持 SSE 活跃
+                print(f"Processing alignment {idx+1}/{min(hitlist_size, len(blast_record.alignments))}")
+
+        handle.close()
+        print(f"✅ BLAST search completed: {len(hits)} alignments found")
+
+        # 构建返回结果
         result = {
             "query_length": int(len(sequence)),
             "database": str(database),
@@ -448,8 +445,9 @@ def blast_search(payload: dict):
             "total_hits": int(len(hits)),
             "hits": hits
         }
-        
+
         return {"success": True, "result": result, "error": None}
+
     except Exception as e:
         return {"success": False, "result": None, "error": str(e)}