How do you write a regression suite when your product is a conversation with an LLM? You can't `expect(response).toBe('hello')`. Built one with simulated users + a 3-tier judge (rules → soft rules → LLM). Pass = avg ≥ 6.5 && min ≥ 5. Architecture:
🎯 Built a real bridge between composio.dev & @elizaos
✅ 250+ integrations (GitHub, Slack, Notion, Drive…)
✅ Auth & tokens managed via Composio
✅ Native function calling inside Eliza agents
✅ bun add @standujar/plugin-composio
🔗 github.com/standujar/plug…
🌟 Découvrez la première néobanque intégrant un portefeuille non-custodial dans son application ! Gérez vos Euros et Cryptos de manière fluide et sécurisée avec Deblock. 🚀 #Deblock
En savoir plus sur deblock.com