A GitHub-ready static MVP for the SolProof AI / PROOF project.
This repository demonstrates a first-principles product baseline for verifiable AI on Solana: model selection, request validation, mock inference, proof bundle generation, local verification, token burn simulation, and replayable proof history.
## Table of Contents - [Overview](#overview) - [Project Goals](#project-goals) - [Project Status](#project-status) - [Core Features](#core-features) - [Project Highlights](#project-highlights) - [Official Links](#official-links) - [Technical Architecture](#technical-architecture) - [System Flow](#system-flow) - [Functional Modules](#functional-modules) - [Project Structure](#project-structure) - [Technology Stack](#technology-stack) - [Installation](#installation) - [Usage](#usage) - [Configuration](#configuration) - [Verification Logic](#verification-logic) - [Roadmap](#roadmap) - [FAQ](#faq) - [Current Limitations](#current-limitations) - [Future Upgrades](#future-upgrades) - [Notes](#notes) ## Overview SolProof AI is positioned as a verifiable AI experience on Solana. This MVP is intentionally lightweight and static, but it captures the minimum product loop required to make the concept understandable and demo-ready: 1. A user selects a model 2. A user submits an input payload 3. The app performs a deterministic mock inference 4. The app creates a proof bundle from the request and output 5. The bundle is verified locally 6. The app stores a replayable receipt and burn history The current version does **not** claim to provide production-grade zero-knowledge proofs or real on-chain verification. It is a clean MVP foundation designed for product validation, interface iteration, and future integration work. ## Project Goals - Demonstrate the smallest believable SolProof product loop - Make the "verifiable AI on Solana" concept tangible for demos and GitHub visitors - Keep the repository simple enough to run without a build system - Create clear seams for future wallet, API, proving, and chain integrations - Preserve product credibility by validating proof bundles locally instead of only rendering static UI ## Project Status | Area | Status | Notes | |---|---|---| | Product framing | Complete for MVP | Narrative, interface, and proof flow are visible | | Static front-end | Complete for MVP | No bundler, no framework, GitHub-friendly | | Model registry | Complete for MVP | Trial and Pro models are configurable | | Request validation | Complete for MVP | Per-model validation rules are implemented | | Mock inference engine | Complete for MVP | Deterministic heuristic outputs by model | | Proof bundle generation | Complete for MVP | Fingerprints, commitment, proof hash, tx id | | Local bundle verification | Complete for MVP | Rebuild and compare proof artifacts | | Wallet simulation | Complete for MVP | Demo wallet and PROOF balance burn flow | | On-chain verification | Planned | Not yet connected to Solana | | Real AI integration | Planned | Not yet connected to an external model API | ## Core Features - Trial and Pro model selection - Prompt input with optional file metadata - Per-model request validation rules - Deterministic inference output generation - Proof bundle construction with: - input fingerprint - output fingerprint - commitment - proof hash - public input hash - transaction id - Local bundle verification and replay - Demo wallet connection and PROOF token burn simulation - Local storage persistence for history and active bundle state - GitHub-friendly documentation and static deployment flow ## Project Highlights - **Zero build tooling** - The app is static and easy to publish on GitHub Pages or any basic static host. - **First-principles MVP design** - The implementation starts with the smallest trustworthy product loop instead of simulating a full protocol stack. - **Replayable proof receipts** - History entries store complete bundles so previous runs can be reopened and re-verified. - **Deterministic verification** - The verification layer rebuilds the bundle from the saved request and checks the critical proof fields. - **Clear upgrade path** - The repository is structured so each subsystem can be replaced incrementally with real infrastructure. ## Official Links | Resource | URL | |---|---| | Official Website | `https://proofaisol.xyz/` | | Twitter | `https://x.com/PROOF__SOL` | | GitHub Repository | `https://github.com/fm374coullon/SolProof-AI.git` | ## Technical Architecture This MVP uses a static browser architecture with three main layers: ### 1. Presentation Layer - HTML entry page - CSS-based visual system - Interactive UI rendered directly in the browser ### 2. Application Layer - Model selection - Form submission - Request validation - Wallet state - Burn accounting - History replay ### 3. Verification Layer - Deterministic inference logic - Request normalization - Fingerprint generation - Commitment and proof hash generation - Local bundle reconstruction and verification ### Architecture Diagram ```mermaid flowchart TD A["User Interface"] --> B["Application Controller"] B --> C["Model Registry"] B --> D["Request Validation"] B --> E["Inference Engine"] E --> F["Proof Builder"] F --> G["Bundle Verifier"] B --> H["Wallet and Burn State"] B --> I["Local Storage History"] G --> I H --> I ``` ## System Flow ```mermaid flowchart LR A["Select Model"] --> B["Enter Prompt or Metadata"] B --> C["Validate Request"] C --> D["Run Deterministic Inference"] D --> E["Hash Input and Output"] E --> F["Build Proof Bundle"] F --> G["Verify Bundle Locally"] G --> H["Burn Demo PROOF if Pro"] H --> I["Store Receipt and History"] ``` ## Functional Modules ### Model Registry Handles all model-facing configuration: - model id - display name - tier - burn amount - constraints - prompt hint - validation minimums ### Inference Console Provides the interactive request surface: - prompt input - optional file metadata - quick-fill seeds - run button - execution mode feedback ### Verification Pipeline Provides step-by-step visibility into the run: - request intake - inference execution - commitment generation - proof materialization - local verification - receipt archival ### Proof Artifact Panel Displays the full result object: - request envelope - inference output - fingerprints - proof fields - simulated receipt - verification status ### Wallet and Burn State Simulates token-gated access for Pro models: - connect demo wallet - initialize PROOF balance - deduct burn amount - show total burned ### History and Replay Stores and reopens previous runs: - summary list of recent runs - replay previously generated bundles - verify stored bundles again ## Project Structure ```text . |-- index.html |-- styles.css |-- app.js |-- assets | `-- logo.png |-- modules | |-- data.js | |-- engine.js | `-- state.js `-- README.md ``` ### File Responsibilities | File | Responsibility | |---|---| | `index.html` | Main page layout and application entry | | `styles.css` | Visual system, responsive layout, and component styling | | `app.js` | UI orchestration, state updates, rendering, event wiring | | `modules/data.js` | Model definitions, prompt seeds, and pipeline steps | | `modules/engine.js` | Request normalization, inference logic, proof generation, verification | | `modules/state.js` | Local storage loading, saving, and wallet identity helpers | | `assets/logo.png` | Repository logo for GitHub and project branding | ## Technology Stack ### Runtime - Browser-native JavaScript modules - HTML5 - CSS3 ### Browser APIs - `Web Crypto API` for SHA-256 hashing - `localStorage` for history persistence - `Clipboard API` for copying bundles ### Hosting Model - Static hosting - GitHub Pages compatible - No Node.js build pipeline required for runtime ## Installation ### Prerequisites - A modern browser with ES module support - Optional: a simple static file server for local preview ### Quick Start 1. Clone or download this repository. 2. Open the project folder. 3. Serve the folder with a static server. 4. Open the local URL in a browser. ### Local Preview Options - VS Code Live Server - Python: ```bash python -m http.server 8000 ``` - Any GitHub Pages-compatible static host ### Recommended Local URL ```text http://localhost:8000 ``` ## Usage ### Basic Flow 1. Open the app in a browser 2. Select a model from the registry 3. Enter a prompt or use one of the seeded examples 4. Optionally attach a file for metadata-aware runs 5. Connect the demo wallet if using a Pro model 6. Click `Generate Proof` 7. Review the verification logs and proof artifact 8. Reopen a past run from history if needed ### Trial Model Flow - No wallet required - No token burn - Best for demos and initial proof flow review ### Pro Model Flow - Requires demo wallet connection - Deducts a PROOF burn amount - Produces the same proof flow plus burn accounting ## Configuration This MVP is configured through plain JavaScript files. ### Model Configuration Edit `modules/data.js` to change: - model ids - labels - burn amounts - constraints - prompt hints - validation minimums ### Prompt Seeds Edit `PROMPT_SEEDS` in `modules/data.js` to change the quick-fill examples. ### Verification Steps Edit `PIPELINE_STEPS` in `modules/data.js` to change log stage labels. ### Wallet and Storage Behavior Edit `modules/state.js` to change: - local storage key - default state - wallet address generation ### Inference and Proof Logic Edit `modules/engine.js` to change: - heuristic inference behavior - fingerprint rules - proof field construction - verification comparison logic ## Verification Logic The MVP verification system works by reconstructing the bundle from the saved request and comparing the rebuilt fields against the stored result. ### Verified Fields - inference output - input fingerprint - output fingerprint - commitment - proof hash - public input hash - transaction id - model identity - configured burn amount ### Why This Matters Even though the proof system is simulated, the MVP still enforces internal consistency. This makes the product demo more credible than a purely decorative interface. ## Roadmap ### April 2026 - Phase 1: MVP Foundation - [x] Static repository structure - [x] Model registry - [x] Prompt submission - [x] Deterministic inference logic - [x] Proof bundle generation - [x] Local verification - [x] History replay - [x] README and branding assets ### May 2026 - Phase 2: Product Hardening - [ ] Add reset controls for wallet and history - [ ] Add richer validation feedback in the UI - [ ] Add export and import for receipts - [ ] Add dedicated views for each model type ### June 2026 and Beyond - Phase 3: Real Integrations - [ ] Connect wallet adapter - [ ] Connect real AI inference endpoints - [ ] Replace simulated proof generation with a real proving flow - [ ] Add Solana transaction submission - [ ] Add explorer-linked receipts from real network data ## FAQ ### Is this a real ZK proof implementation? No. This repository is an MVP and uses deterministic simulated proof artifacts. It is designed to show the intended product loop, not to claim production-grade cryptographic security. ### Does this app talk to Solana right now? No. The receipt and explorer link are simulated. The current version is chain-shaped, not chain-connected. ### Why use local verification if the proof is simulated? Because local verification makes the bundle internally consistent and replayable. That helps communicate how a real verification system would behave. ### Why is there no build step? The MVP is intentionally static so it can be published quickly and reviewed easily on GitHub or any simple static host. ### Where can I change model behavior? Use `modules/data.js` for model configuration and `modules/engine.js` for inference and verification behavior. ### Why are Trial and Pro models both simulated? The distinction is a product-gating concept in this MVP. Trial models are free, while Pro models require wallet connection and PROOF burn simulation. ## Current Limitations - No real AI API integration - No real Solana transaction signing or submission - No server-side proving system - No authentication or access control - No backend persistence layer - No automated test suite yet - No package manifest because the runtime is static ## Future Upgrades ### Product Layer - Add model-specific input templates - Add visual proof receipt export - Add richer error states and recovery flows ### AI Layer - Replace heuristic inference with a model API - Normalize structured output schemas across models - Add request tracing and usage metering ### Verification Layer - Integrate real proving infrastructure - Add Solana receipt writing - Add explorer-backed verification links - Persist receipts to backend storage or indexing infrastructure ## Notes - The visible model lineup in this MVP follows the models shown on the public SolProof site UI. - The contract address remains unannounced in current public materials, so this demo intentionally treats CA as unavailable. - The logo used in this repository is stored locally at `./assets/logo.png` so it renders correctly on GitHub. - The repository currently focuses on product truthfulness and upgradeability over protocol completeness. - Official project links used in this README: - Website: `https://proofaisol.xyz/` - Twitter: `https://x.com/PROOF__SOL` - Repository: `https://github.com/fm374coullon/SolProof-AI.git`