DPrompt
DPrompt (revox.ai/dprompt) is REVOX’s on-chain AI inference layer, designed to make advanced AI models natively accessible within smart contracts. It enables decentralized applications (dApps) to query machine learning models and receive verifiable results directly on-chain, without relying on centralized providers. By combining off-chain computation with cryptographic proofs, DPrompt bridges the gap between AI and Web3 in a trust-minimized way.
Unlike traditional oracles that deliver only static data (e.g., price feeds), DPrompt delivers dynamic, context-aware AI inferences—allowing smart contracts to ask open-ended questions, perform classifications, or run predictive models, then act upon the results autonomously.
Product Experience
The DPrompt interface reflects its goal of simplicity and accessibility, both for developers and end users.
Model Selection: Users can choose between different AI models (e.g., GPT) depending on their task requirements.
Chain Integration: DPrompt currently supports BNB Smart Chain and is designed for multi-chain expansion.
Custom and Hot Prompts:
Custom Prompt allows users to input free-form instructions for the AI model.
Hot Prompts provide pre-built templates for common tasks (e.g., risk checks, trading signals, classification).
Result Verification: The returned inference is displayed alongside a transaction hash, ensuring traceability and verifiability.
Smart Contract Integration: Each DPrompt instance is tied to a contract address, enabling dApps to reference it directly within workflows.
This streamlined UI (see figure above) illustrates the workflow: Users → AI Contract → AI Provider → Result (with Proofs).
Key Features
On-Chain AI Requests
Smart contracts submit inference requests directly, embedding AI reasoning into automated workflows.
Verifiable Outputs
Results are accompanied by cryptographic proofs (zkML, stake-based consensus, or hybrid schemes).
Gas-Only Model
Users only pay gas fees, lowering friction and removing hidden subscription models.
Multi-Model Support
Compatible with large language models (LLMs), classification models, and prediction engines.
Composable Workflows
DPrompt can be used as a node in REVOX workflows, combined with Primitive Agents, Plugins, and Context Knowledge.
Technical Architecture
Request Layer: Smart contracts encode prompts and send them to the DPrompt contract.
Computation Layer (Off-Chain AI Providers): Nodes execute AI inference off-chain to optimize cost and latency.
Verification Layer: Each inference is accompanied by:
Proof-of-Stake Consensus: Multiple providers must agree on results.
Zero-Knowledge Proofs (zkML): Where available, cryptographic guarantees prove that the model ran faithfully on given inputs.
Delivery Layer: Verified results are written back on-chain, available to dApps, agents, and workflows.
Developer and User Value
Developers
Direct access to AI in smart contracts.
Ability to build autonomous agents and dApps without external trust assumptions.
Simple integration with SDKs and Studio for workflow design.
Users
Lower transaction complexity: only gas fees, no external API keys.
Transparency and immutability: every inference can be audited on-chain.
Confidence that AI outputs are verifiable and tamper-resistant.
Future Roadmap
Expanded Model Support: Integrating multi-modal models (text, image, audio).
Cross-Chain Availability: Extend DPrompt beyond BNB Smart Chain into Ethereum, Arbitrum, and other L2s.
User-Submitted Models: Enable developers to deploy and monetize their own AI models in the DPrompt network.
Workflow Integration: Full compatibility with REVOX Studio for drag-and-drop inference nodes in workflow design.
Performance Optimization: Lower latency pipelines, batching requests, and improving zkML proving efficiency.
Last updated