PrivaFlow-AI

• 

  PrivaFlow-AI v3.0: Sovereign Multi-Agent Zero-Knowledge Attestation Protocol

• 

  Client Ingestion Terminal: Zero-Transit Raw Artifact Analysis

• 

  ZK Redaction Chamber: Entity Sanitization & Threshold Compliance Audit

• 

  Midnight Verifiable Credential Passport & Halo2 Proof Envelope

• 

  Enterprise Verifier Portal: Real-Time On-Chain Proof Validation

Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for PrivaFlow-AI

Inspiration

In modern compliance workflows (mortgage approvals, regulatory filings, employment verification), users are forced into full-disclosure dilemmas—uploading entire unencrypted documents (pay stubs, tax records, medical files) just to prove single numeric predicates. This creates multi-gigabyte PII attack surfaces for enterprises. PrivaFlow-AI bridges client-isolated multi-agent AI reasoning with Midnight Network's compact Zero-Knowledge state machine to deliver sovereign, tamper-proof attestations with Zero Raw PII Exposure.

What it does

• Client-Side Isolated Ingestion: Sensitive documents are processed entirely within a client-side memory enclave.
• Tri-Agent AI Consensus: Google Gemini-powered agents (Redaction Agent, Constraint Auditor, and Cryptographic Arbiter) tokenize all identity artifacts (SSN, names, account IDs) and derive compliance invariants.
• Midnight Compact ZK Engine: Compiles validation proofs into on-chain zero-knowledge witnesses, anchoring Blake2b prompt hashes and deriving anti-replay Nullifier Trees.
• Enterprise Verifier Portal: Third-party auditors verify policy satisfaction against the Midnight ledger in sub-second latency without ever accessing underlying raw records.

How we built it

• Midnight Compact: Architected PrivaFlowAttestation.compact smart contracts enforcing private witness states, public nullifiers, and anti-replay trees.
• AI Enclave: Integrated Google Gemini API for multi-agent constraint extraction, entity sanitization, and structured cryptographic payload generation.
• Backend & Relayer: FastAPI microservice handling BLAKE2b/Poseidon hashing, consensus root calculation, and testnet RPC communication.
• Frontend HUD: Next.js 14, Tailwind CSS, and Lucide Icons crafted with an APEX OLED glass-morphic terminal interface.

Challenges we ran into

• Designing an immutable anti-replay nullifier mechanism that prevents identical credentials from being double-spent without linking back to user identity.
• Ensuring deterministic JSON schemas and high confidence (>99.8%) from multi-agent reasoning pipelines prior to zero-knowledge witness compilation.

Accomplishments that we're proud of

• Achieving zero-latency, full-fidelity PII redaction combined with mathematical Zero-Knowledge proofs on Midnight Network.
• Building an institutional-grade, ready-to-deploy user experience and verifier matrix.

What we learned

• Deep operational mastery of Midnight's Compact language, private/public state segregation, and zero-knowledge circuit optimization.

What's next for PrivaFlow-AI

• Direct in-browser WASM proof generation for 100% offline zero-knowledge witness computation.
• Cross-chain Merkle proof relayers to EVM and Solana enterprise ecosystems.

Built With

• compact-zk
• fastapi
• google-gemini-api
• midnight-network
• next.js
• python
• react
• tailwind-css
• typescript
• zero-knowledge-proofs