Skip to main content

ml_pipeline

Version: v1.3.0 | Status: Experimental | Last Updated: July 2026

Overview​

Lightweight interface for producing pipeline-definition and execution-shaped receipts. The two functions are Python exports and MCP tools. They are stateless pass-through adapters—not an ML execution engine. For executable workflow orchestration, use the orchestrator module.

Key Components​

ComponentFileDescription
ml_pipeline_createmcp_tools.pyMCP tool: define a pipeline from name + steps
ml_pipeline_executemcp_tools.pyMCP tool: execute a pipeline with inputs

Quick Start​

from codomyrmex.ml_pipeline import ml_pipeline_create, ml_pipeline_execute

result = ml_pipeline_create("my_pipeline", [{"name": "train", "epochs": 3}])
output = ml_pipeline_execute("my_pipeline", {"data": "/path/to/data"})

[!NOTE] Both functions return echo responses. No validation, persistence, or ML execution occurs.

MCP Tools​

ToolDescription
ml_pipeline_createCreate a pipeline definition
ml_pipeline_executeExecute a named pipeline

Directory Contents​

FilePurpose
mcp_tools.pyMCP tool definitions (2 tools, 34 lines)
__init__.pyExplicit exports for both MCP-backed functions