With the rapid development of blockchain technology, smart contracts have gradually become a common vehicle for financial fraud. Among them, Ponzi schemes in the form of smart contracts have become a typical and high-risk form of fraud in the blockchain ecosystem, resulting in significant economic losses. Recent research has made significant progress in identifying Ponzi scheme smart contracts by extracting features from contract source code and account transaction data and using them to build machine learning models. However, blockchain anonymity and transaction latency significantly limit the practical applicability of these methods. To address these issues, this study proposes a real-time identification framework for smart contract Ponzi schemes based on graphs and hypergraphs. This framework is based on the bytecode that is required to be disclosed at the time of contract deployment. Before any external transaction involving the contract occurs, the proposed framework extracts opcode sequences, constructs graph and hypergraph structures, and combines graph convolutional networks with hypergraph convolutional networks to enable real-time “detection upon deployment”. Experimental results on the latest dataset show that this framework significantly outperforms traditional methods in terms of detection accuracy and F1-score. This demonstrates that the framework can perform deep feature extraction for Ponzi scheme smart contracts without relying on source code or interaction data, providing effective support for transforming the identification of fraudulent contracts from “post-event tracing” to “pre-event auditing”.