Noctua MCP Server

Instructions

Implementation Reference

• src/noctua_mcp/mcp_server.py:155-240 (handler)

  The main handler function for the 'add_individual' MCP tool. It validates input, calls BaristaClient.add_individual with label checksum, handles errors/rollback, and returns the individual ID or error details.
  @mcp.tool() async def add_individual( model_id: str, class_curie: str, class_label: str, assign_var: str = "x1" ) -> Dict[str, Any]: """ Add an individual (instance) of a class to a GO-CAM model with label validation. This tool requires providing the expected label for the class to prevent accidental use of wrong IDs (e.g., GO:0003924 vs GO:0003925). The operation will automatically rollback if the created individual doesn't match the expected label. Args: model_id: The GO-CAM model identifier (e.g., "gomodel:12345") class_curie: The class to instantiate (e.g., "GO:0003674") class_label: The expected rdfs:label of the class (e.g., "molecular_function") assign_var: Variable name for referencing in the same batch Returns: Barista API response with message-type and signal fields. If validation fails, includes rolled_back=true and validation error. Examples: # Add a molecular function activity with validation add_individual("gomodel:12345", "GO:0004672", "protein kinase activity", "mf1") # Add a protein/gene product with validation add_individual("gomodel:12345", "UniProtKB:P38398", "BRCA1", "gp1") # Add a cellular component with validation add_individual("gomodel:12345", "GO:0005737", "cytoplasm", "cc1") # Add a biological process with validation add_individual("gomodel:12345", "GO:0016055", "Wnt signaling pathway", "bp1") # Add an evidence instance with validation add_individual("gomodel:12345", "ECO:0000353", "physical interaction evidence", "ev1") # Variables like "mf1", "gp1" can be referenced in subsequent # add_fact calls within the same batch operation Notes: - The label acts as a checksum to prevent ID hallucination - If the label doesn't match, the operation is automatically rolled back - This prevents corrupt models from incorrect IDs """ client = get_client() resp = client.add_individual(model_id, class_curie, assign_var, expected_label=class_label) if resp.validation_failed: return { "success": False, "error": "Validation failed", "reason": resp.validation_reason, "rolled_back": True, "expected_label": class_label, "class_curie": class_curie } if resp.error: return { "success": False, "error": resp.error, "model_id": model_id, "class_curie": class_curie } # Get the actual individual ID from model_vars or from the individuals list individual_id = assign_var if resp.model_vars and assign_var in resp.model_vars: individual_id = resp.model_vars[assign_var] elif resp.individuals and len(resp.individuals) > 0: # If model_vars is empty, get the ID from the last individual (the one just created) individual_id = resp.individuals[-1].id # Return minimal success response return { "success": True, "individual_id": individual_id, "class_curie": class_curie, "assign_var": assign_var }