StarWhisper Telescope: Agent-Based Observation Assistant System to Approach AI Astrophysicist

The exponential growth of large-scale telescope arrays has boosted time-domain astronomy but introduced operational bottlenecks, including labor-intensive observation planning, data processing, and real-time decision-making. Here we present the StarWhisper Telescope (SWT) system, an AI agent framework automating end-to-end astronomical observations for projects, such as the Nearby Galaxy Supernovae Survey (NGSS). By integrating large language models (LLMs) with specialized function calls and modular workflows, SWT autonomously generates site-specific observation lists, executes real-time image analysis via pipelines like Xinglong-Observatory Popular Science Telescope Pipeline, and dynamically triggers follow-up proposals upon transient detection. The system reduces human intervention by 90% through automated target prioritization, scheduling, and data reporting, while enabling seamless collaboration between amateur and professional astronomers. Deployed across NGSS’s network of 10 amateur telescopes, SWT has detected transients (e.g., SN2024xin, AT2025pk) with promising response times relative to existing surveys, achieving a median discovery lag of less than 12 hours for nearby transients. Furthermore, SWT’s scalable agent architecture—comprising observation planning, control, and analysis modules—provides a blueprint for future facilities like the SiTian Project, where AI-driven autonomy will be critical for managing hundreds of telescopes. By unifying automated instrumentation with scientific reasoning, this work establishes a transformative paradigm where embodied AI systems accelerate discovery, setting a new benchmark for efficiency and scalability in astrophysical research.