export const MODELS = [ { id: "gemini-flash-zkml", name: "Gemini Flash ZKML", description: "Fast text classification with lightweight proof generation.", tier: "trial", proofBurn: 0, constraints: 524288, latency: "1.5s - 3.0s", modeLabel: "Free local verification", minPromptLength: 20, promptHint: "Use this model for text categorization and broad intent labeling.", attachmentPolicy: "optional" }, { id: "gemini-sentiment", name: "Gemini Sentiment", description: "Multi-language sentiment and confidence scoring for social text.", tier: "trial", proofBurn: 0, constraints: 262144, latency: "1.2s - 2.4s", modeLabel: "Free local verification", minPromptLength: 12, promptHint: "Best for social posts, launch feedback, and market-tone analysis.", attachmentPolicy: "optional" }, { id: "gemini-vision", name: "Gemini Vision", description: "Image-or-scene classification for richer, higher-cost verification flows.", tier: "pro", proofBurn: 120, constraints: 1048576, latency: "3.0s - 5.2s", modeLabel: "Wallet + token burn", minPromptLength: 18, promptHint: "Provide a scene description and optionally attach a file for metadata-aware classification.", attachmentPolicy: "recommended" }, { id: "gemini-fraud", name: "Gemini FraudDetect", description: "Transaction risk scoring for suspicious patterns and abnormal activity.", tier: "pro", proofBurn: 200, constraints: 2097152, latency: "4.2s - 6.8s", modeLabel: "Wallet + token burn", minPromptLength: 24, promptHint: "Use structured transaction context, addresses, routing clues, and velocity markers.", attachmentPolicy: "optional" } ]; export const PROMPT_SEEDS = { defi: "Wallet 7pV... deposited 125,000 USDC into a lending pool, borrowed 80,000 against a thin collateral asset, then bridged funds out within 5 minutes. Assess the risk level and explain why.", sentiment: "The launch looks sharp, the product finally feels real, and the team shipped faster than most Solana AI projects this cycle.", fraud: "txn_id=8fc29 amount=48950 token=USDC route=unknown_pool velocity=14m recipient_age=2h flagged_mixers=true" }; export const PIPELINE_STEPS = [ "Input committed", "Inference executed", "Commitment hashed", "Proof materialized", "Bundle verified", "Receipt archived" ];