llm-echo 0.5a0: Revolutionizing Automated Testing for LLMs

⚡

Key Takeaways

1Version 0.5a0 of the llm-echo plugin introduces an option to test LLM 0.32a0 and above.

2This plugin uses a mock model 'echo' to facilitate writing automated tests without a real LLM.

3The uvx command allows simulating a reasoning block before returning a JSON of the prompt.

💡Why it matters — This plugin simplifies the development and debugging of LLM models, thereby accelerating testing and innovation processes.

⚡Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

📄

Full Analysis

llm-echo 0.5a0: Revolutionizing Automated Testing for LLM

The llm-echo debugging plugin has recently been updated to version 0.5a0, bringing a crucial new feature for developers working with LLM language models. This update introduces the -o thinking 1 option, specifically designed to test versions 0.32a0 and above of LLM.

The llm-echo plugin offers a mock model named "echo," which does not require the use of a real language model. This feature is particularly useful for writing automated tests, allowing developers to simulate interactions without mobilizing significant resources. To use this functionality, the following command can be executed:

uvx --with llm==0.32a1 --with llm-echo==0.5a0 llm -m echo hi -o thinking 1

This command simulates a reasoning block for the standard error before returning a JSON that reflects the initial prompt. This allows for effective testing of the expected behaviors of LLM models without the need for an operational model.

Version History

LLM version 0.32a0, released on April 29, 2026, represents a major overhaul while remaining backward compatible. This update was preceded by the obsolescence of OpenAI Microsoft’s AGI clause on April 27, 2026. Additionally, DeepSeek V4, launched on April 24, 2026, offers near-state-of-the-art performance at a reduced cost.

This article was written by Simon Willison and published on May 5, 2026.

⚡

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.

📰 Voir toutes les actus IA →