llm-echo 0.5 counts tool-call and reasoning tokens

llm-echo exists as a debugging aid rather than a real model: it echoes prompts back, so developers can inspect what the LLM tool is actually sending without paying for or waiting on a model call. The 0.5 release narrows a gap in that simulation. Version 0.5 now factors tool calls, system prompts and reasoning tokens into the simulated token counts, so the numbers it reports better reflect the inputs a real model would be charged for. Willison notes he revisited the plugin while testing PRs for an upcoming LLM release, which suggests the token-accounting change was driven by his own testing needs around that release.