City Gateway: Innovative, Social, and Learning

Module 1: Dynamic Edge Profiling (Privacy & Accessibility)

• Functional Requirements:

  • Perform on-device user profiling using voice-to-text semantic interfaces to assist users with low digital literacy.

  • Maintain strict GDPR/BDAR Article 25 compliance by ensuring sensitive vulnerability markers (e.g., disability status, NEET status) never leave the handset.

• Design:

  • Runtime: Deploy Google AI Edge (LiteRT) for local inference of quantized Small Language Models (SLMs).

  • Local Storage: All profiling data must be stored in a sandboxed SQLite database, encrypted with 256-bit AES via SQLCipher, with keys managed by the device's hardware-backed keystore.

  • UI Constraints: Minimum touch-target size of 48×48 logical pixels and layouts that remain stable under 200% system-level text scaling.

• Why: This approach establishes "Privacy by Design", which is a non-negotiable legal requirement for handling sensitive data from vulnerable groups, and ensures a 100% accessibility pass rate for seniors.

Module 2: Gamification Engine & WTA Core Ledger

• Functional Requirements:

  • Track and issue "Academy Credits" based on verified civic engagement.

  • Sync local points with the WorkTravel Academy (WTA) Point Matrix to move users through global career tiers (Explorer → Builder → Verified).

• Design:

  • Architecture: Use a Backend for Frontend (BFF) pattern to centralise the point matrix logic and eligibility rules, ensuring a single source of truth across client versions.

  • Data Structure: Implement structured tables designed for bidirectional sync with WTA’s pre-existing global collection formats (Activities and MemberRewards schemas).

• Why: Centralising logic on the server side reduces implementation effort by 4 times and accelerates Time to Market (TTM) from 4–5 weeks to just 1–1.5 weeks. It also bridges the gap between simple attendance and meaningful engagement.

Module 3: Organisation "Atvirumo" (Openness) Ranking

• Functional Requirements:

  • Provide a dynamic public leaderboard ranking local service providers based on their transparency and inclusivity.

  • Incentivize organisations to expose data through open standards.

• Design:

  • Scoring Formula: Reputation scores (Satvirumo​) are calculated based on Transparency (percentage of resources exposed via MCP), Participation Volume (verified transactions), and Latency (sync delays).

  • Validation: Use the ckanext-qa extension to assign a 5-star openness rating to any public data provided by the organisations.

• Why: This creates a social loyalty hub that encourages local organisations to adopt more inclusive and transparent practices to improve their community standing.

Module 4: Agentic AI Concierge ("Rietaviškis")

• Functional Requirements:

  • Provide 24/7 Lithuanian-language support to guide residents through complex municipal services.

  • Perform automated orchestration between the user and local services.

• Design:

  • Protocol 1: Use Model Context Protocol (MCP) to allow the concierge to dynamically discover and query municipal toolsets without direct database links.

  • Protocol 2: Implement Agent-to-Agent (A2A) v1.2 for secure, programmatic handshakes between the user agent and organisation agents using lightweight JSON-RPC schemas.

• Why: Standardizing communication via MCP solves the N×M integration problem, allowing the AI to work with any tool or database that supports the standard without custom "glue code".

Module 5: Proof-of-Participation (PoP) Verification

• Functional Requirements:

  • Enable physical event check-ins with a Verification Velocity of < 30 seconds.

  • Guarantee offline-first functionality for areas with poor municipal Wi-Fi.

• Design:

  • Verification: Utilize Ed25519 asymmetric verification of time-sensitive, salted QR tokens generated by physical sites.

  • Sync Logic: Use Idempotent Ingestion Design to ensure that retried sync operations upon reconnection do not result in duplicate credit issuance.

  • Alternative Input: Support BLE proximity beacons as an alternative to QR scanning for users with motor-control impairments.

• Why: High-velocity, offline-first verification is essential for user retention; any lag in issuing rewards during a physical event leads to abandonment.

Expert Summary of Architecture & Deliverables

For Phase 1 (Weeks 1–6), the priority must be the Unified Architectural Core (BFF) and the On-Device AI Runtime. The mandatory framework for the frontend is Flutter (Dart) due to its pixel-perfect control over accessibility and predictable 5-year maintenance cycle. Testing should be shifted to a centralized server-side integration suite, which is projected to reduce critical production incidents by 60%