Abstract
This proposal introduces AI-powered transaction decoding for OpenScan.AI: every transaction and contract interaction on the explorer gets a plain-language explanation of what it actually did — which assets moved, which contracts were called, and what the net effect was for each party.
Motivation
Raw transaction data (calldata, event logs, internal calls) is opaque to most users. Even experienced users waste time mentally simulating contract calls. Decoding transactions into human language is the core of OpenScan.AI’s “AI-powered explorer” promise and the headline feature of Phase 3 (Intelligence) of the 2026 roadmap.
Specification
- A decoding pipeline sits between the indexer and the API layer. For each transaction it assembles the full execution context: calldata, traces, event logs, and token transfers.
- A model summarizes the transaction into a structured explanation: action type (transfer, swap, approval, contract deployment, etc.), parties involved, assets and amounts, and a one-paragraph plain-language summary.
- Decoded output is exposed in the explorer UI on transaction pages and via a dedicated endpoint in the REST API v2.
- Explanations carry a confidence indicator; low-confidence decodes are labeled as such rather than hidden.
Out of scope
- Security risk scoring of contracts (covered by the Phase 3 security scanning work).
- Decoding for chains beyond XDC before multi-chain support lands (see OEP-3).
Milestones
- Decoding pipeline for native transfers and standard token operations.
- Coverage for common DeFi patterns (swaps, liquidity, staking).
- Public API endpoint and explorer UI integration.