Sponsor: Teleport — AI agents (and the sub-agents they spawn) are granted the same access as humans, but they never sleep. Learn how to monitor and govern them using isolated ephemeral trusted runtimes.
October 18, 2026
Release llm-echo 0.5 — a debug plugin providing an echo model for LLM
I’m working towards a new release of LLM, and while testing a PR, I revisited the llm-echo debug plugin. It provides a “model” that doesn’t involve an LLM at all—it simply echoes back the prompt you give it. This new release updates its simulated token usage counting to account for tool calls, system prompts, and reasoning tokens.
Example usage:
uvx --with llm-echo llm -m echo 'hi there' -o thinking 1 --usageOutput:
First I consider the prompt, then I decide what to say.{ "prompt": "hi there", "system": "", "attachments": [], "stream": true, "previous": []}Token usage: 2 input, 24 outputPosted on October 18, 2026 at 11:16 PM
Recent Posts
- Hello Python 3.15, goodbye 3.10, now I can use 3.11 - October 12, 2026
- A new blog feature built using my voice - October 9, 2026
- Claude Haiku 5.5 - October 7, 2026
- We need default hard budget caps for almost everything - October 3, 2026
This is a beat by Simon Willison, posted on October 18, 2026.