Large Workflows with Local LLMs

## Executive summary — This post describes the design, implementation, and testing of a lightweight Python library (TS_LLMLib) and example scripts to orchestrate local LLMs and MCP endpoints for tasks such as Ghidra-based function annotation and C/C++ file scanning; it focuses on context-size limits, tool sandboxing, retry/timeouts, and practical test outcomes rather than describing an active cyber incident.