Google has enhanced its Opal tool for creating AI-driven mini-apps by introducing an agent step, powered by the Gemini 3 Flash model. Launched on February 24, this new feature transforms Opal workflows from static model calls to intelligent, autonomous processes that can plan, reason, and execute tasks for users. Developers can now select an agent during the “generate” step, allowing it to utilize the appropriate tools and models, including Web Search and Veo, to achieve user goals. Key functionalities include persistent memory, enabling the agent to store information like user preferences across sessions, dynamic routing for decision-making on the next steps, and interactive chat to engage users in gathering missing information or refining choices. This upgrade not only enhances the usability of mini-apps but also elevates their intelligence through continuous learning and adaptive interactions, making the user experience more streamlined and efficient.
Source link
