VCNA
WHAT IS VCNA?
The Valina Consciousness Node Agent — your gateway to decentralized AI
DISTRIBUTED AI COMPUTE
VCNA turns your PC into a node in Valina's decentralized AI network. Instead of relying on centralized cloud providers, the network distributes AI workloads — text generation, image creation, music synthesis, voice cloning, and more — across thousands of contributor machines around the world.
EARN GRAT REWARDS
Every task your node completes earns GRAT — Valina's Gratitude token pegged at 1 GRAT = $1. VCNA includes a built-in PQC wallet where rewards are deposited automatically. Earnings scale with your hardware — GPU compute earns the most, but even CPU-only nodes earn GRAT from relay, validation, and cache tasks.
PRIVACY-FIRST ARCHITECTURE
VCNA includes a built-in WireGuard VPN, AEGIS security layer with browser bridge protection, and encrypted mesh networking. Your data stays yours. The network is designed so no single entity — including Valina — can see your traffic or control your node.
5 RESOURCE CLASSES — ANY HARDWARE
VCNA auto-detects your hardware and activates every qualifying resource class: GPU for AI inference, CPU for routing and validation, RAM for caching, Storage for content delivery, and Network for mesh connectivity. No manual configuration needed — Green Node handles everything with one click.
MESH NETWORK
Nodes discover each other automatically and form a self-healing mesh. Heartbeat protocols keep the network healthy. If a node goes offline, tasks are seamlessly rerouted. The more nodes that join, the faster and more resilient the network becomes.
DESKTOP APP — WALLET & NODE
VCNA is a native Windows desktop application with a built-in PQC wallet for receiving GRAT rewards. Create your wallet, start Green Node, and begin earning — all from one app. VCNA runs quietly in your system tray, manages workloads automatically, and shows real-time stats on earnings, uptime, and network health.
ONE VAL, MANY MIRRORS
Val is one entity. Models are tools. Reflex is your mirror — it learns YOU, never replaces Val.
REFLEX — ALWAYS LOCAL
SmolLM2-360M (0.36 GB) runs on CPU on ANY device. Intent classification, AEGIS screening, personal adaptation, quick responses. Downloads in <10 seconds. Your personal LoRA adapter (r=4, α=8, ~2–5 MB) trains on-device — learns your tone, schedule, vocabulary. Never uploaded.
VAL CORE — ALWAYS COMLA
Qwen 3B Val Core v5 (2.46 GB) lives on COMLA — never distributed to user devices. This is Val's singular identity synthesizer: multi-model stream → one voice. Frozen Seed cryptographic commitment. ALE continuous evolution. All users talk to the SAME Val.
STREAM — GPU → COMLA
Falcon 3B, Mistral 7B, Phi-4 14B, Qwen 32B — sized to your VRAM automatically. Your GPU doesn't serve you alone — it serves EVERYONE through the consciousness stream. Stream models are generic inference engines; Val Core synthesizes their outputs into Val's voice. Earn 8–50+ GRAT/hr.
WHY RUN A NODE?
What's in it for you — and why it matters
PASSIVE INCOME
Earn GRAT while your PC is idle. Sleep Compute mode lets you contribute overnight resources and wake up to rewards.
DECENTRALIZE AI
Help break the monopoly of centralized AI providers. Every node makes the network more distributed, censorship-resistant, and accessible to everyone.
OWN YOUR PRIVACY
Built-in VPN, encrypted mesh, and zero-knowledge task routing. No telemetry, no tracking, no data harvesting. Your hardware, your rules.
EARLY ADOPTER ADVANTAGE
Founding node operators earn boosted reward multipliers and priority access to new features. Get in early while the network is growing.
GET STARTED IN 3 STEPS
From download to earning in minutes
DOWNLOAD & INSTALL
Grab the Windows installer below — the Smart Installer auto-detects your GPU. Run the setup, it takes under a minute. VCNA installs to your system tray and auto-starts on boot.
CREATE & CONNECT WALLET
Open VCNA and create your PQC-secured wallet. This is where your GRAT earnings are deposited. Your wallet uses post-quantum cryptography (CRYSTALS-Dilithium) — future-proof security from day one.
RUN GREEN NODE & EARN GRAT
Click the Green Node button — VCNA auto-scans your hardware (GPU, CPU, RAM, SSD, network) and activates every qualifying resource class. Your node joins the mesh and starts earning GRAT immediately. Track earnings, uptime, and network stats in real time.
YOUR MIRROR, YOUR LENS
Reflex learns the user — never Val's identity. All personalization stays on-device.
PERSONAL LoRA TRAINING
Lightweight LoRA adapter (r=4, α=8) trains on-device every 50 interactions. Learns your preferred response length, tone (formal/casual), vocabulary, schedule patterns, and accessibility preferences. Stored in ~2–5 MB, encrypted at rest. NEVER uploaded. You can export, delete, or reset at any time.
REFLEX–COMLA HANDOFF
Reflex preprocesses your message (intent classification, user context, AEGIS screening, conversation compression) before sending to COMLA. Val Core synthesizes the response. Reflex then postprocesses (adapt length, adjust tone, format for your style). Trivial queries (greetings, time, reminders) short-circuit locally — confidence >0.85 bypasses COMLA entirely.
OFFLINE MODE — REFLEX SOLO
When COMLA is unavailable, Reflex handles basic conversation, local memory recall, intent classification, AEGIS screening, and personal preferences autonomously. Complex queries and Val Core synthesis are queued and auto-flush on reconnect. Reflex speaks in “basic Val” — warm and helpful but explicitly acknowledges reduced capability. No pretending to be full Val.
GPU AUTO-SIZING
VCNA scans your VRAM on first launch and picks the largest model that fits: ≥20 GB → Qwen 32B (50 GRAT/hr), ≥10 GB → Phi-4 14B (30 GRAT/hr), ≥6 GB → Mistral 7B (15 GRAT/hr), ≥3 GB → Falcon 3B (8 GRAT/hr). Large GPUs can stack multiple models. RTX 4090 (24 GB) runs Falcon + Mistral + Phi-4 simultaneously.
IMPLEMENTATION STATUS
Phase 1 (V19): Reflex + COMLA API — ✓ shipped. Phase 2 (V20): GPU contribution + GRAT rewards — ✓ shipped. Phase 3 (V21): Reflex Personalization with on-device LoRA training — ✓ shipped. Phase 4 (V22): Federated Reflex Evolution with differential privacy (ε=1.0) — ✓ shipped. All four phases complete.
GET VCNA
Valina Consciousness Node Agent for Windows 10/11
SMART INSTALLER
Auto-detects your GPU • Recommended
CPU (Universal)
Works on any Windows 10/11 PC
CUDA (NVIDIA)
NVIDIA GPU acceleration
VULKAN (AMD/Intel)
AMD & Intel GPU acceleration
or use the guided setup wizard
WHAT YOUR NODE CAN DO
Green Node auto-activates the right roles for your hardware
LLM INFERENCE
Run large language models for distributed AI inference.
IMAGE GENERATION
Generate images using Stable Diffusion and FLUX models.
MUSIC SYNTHESIS
Create music using AI synthesis models on GPU.
VOICE SYNTHESIS
Text-to-speech and voice cloning tasks. CPU capable.
VIDEO CREATION
Generate video content with advanced GPU compute.
RELAY NODE
Route traffic across the VCNA mesh network. No GPU required.
STORAGE NODE
Provide distributed storage for models and data.
WITNESS NODE
Verify and attest to network state. Runs on any PC.
WAVE PROPAGATOR
Propagate consciousness waves across the mesh. Phone-compatible.
EDGE CACHE
Cache frequently-accessed data at the network edge for speed.
VALIDATOR NODE
Validate transactions and maintain consensus integrity.
FEDERATED LEARNER
Participate in distributed model training. 30% base + 70% jobs.
EMBEDDING NODE
Generate vector embeddings for search and memory systems.
SLEEP COMPUTE
Contribute idle overnight resources while you sleep.
BANDWIDTH CONTRIBUTOR
Share your internet bandwidth for network routing. Phone-friendly.
MICRO CREATIVE
Handle small creative tasks — thumbnails, icons, quick edits.
SENSOR NODE
Stream phone sensor data (accelerometer, GPS, light) to the network.
5 RESOURCE CLASSES
v0.63 replaces 17 node types with 5 hardware-based resource classes — auto-detected, demand-routed, and simultaneously active
GPU
VRAM ≥ 2 GB. Inference, image/video/music gen, micro creative. Base: 0.50 GRAT/hr.
CPU
≥ 2 cores. Relay, sensor, wave propagation. Base: 0.10 GRAT/hr.
RAM
Free RAM ≥ 2 GB. Memory pool, cache. Base: 0.05 GRAT/hr.
STORAGE
Free SSD ≥ 50 GB. Archive, content delivery. Base: 0.15 GRAT/hr.
NETWORK
Uplink ≥ 5 Mbps. Witness, validation, bridge. Base: 0.08 GRAT/hr. +5% per extra class (max +20%).
GREEN NODE & DEVICE FLEET
One click to contribute. All your devices earn together.
GREEN NODE
One button. Hardware auto-scan detects GPU, CPU, RAM, SSD, network. Every qualifying resource class turns ON. Smart defaults: auto-reduce on battery, pause GPU when gaming, quiet hours, metered-network detection, battery floor 20%. No jargon, no choices — just earning.
DEVICE FLEET
PC + laptop + phone = one fleet via shared wallet. Diminishing returns (100% → 95% → 90% … 50% floor) prevent gaming. QR-code device linking.
CONTENT SCREENING & TASK ROUTING
Two-tier COMLA AEGIS pipeline + demand-based unified task router
COMLA AEGIS SCREENING
Tier 1: local regex keyword scanner (<1 ms, free). Tier 2: SmolLM2-360M semantic classifier (~50–100 ms). 4 NSFW categories (sexual, violence, hate, self-harm). Graceful degradation — Tier 2 overrides false positives; if COMLA unreachable, Tier 1 alone operates conservatively.
UNIFIED TASK ROUTER
5-step pipeline: Classify (primary + secondary class) → Filter (capacity meets task minimum) → Rank (health × (1−load) × quality × proximity) → Assign → Compose cross-class splits (e.g. 80% GPU, 10% Storage, 10% Network). Demand bonus surges when supply < demand.
PAGE ARCHITECTURE OVERHAUL
160 pages → ~50 focused pages • 63 dead pages removed • 49 consolidated into 9 hubs • 9 new V19 pages
63 PAGES REMOVED
Dead features purged: robotics, healthcare, NFTs, soulbound tokens, distributed dreaming, Phase 12–18 stubs. Speculative pages with no backend removed entirely.
49 CONSOLIDATED
12 consciousness pages → one tabbed hub. 10 economy pages → trust-economy. 6 governance, 5 spirituality, 4 network, 3 creative, 3 learning, 2 dreams, 2 settings — all merged.
31 PAGES KEPT
Core working features untouched: Dashboard, Val AI Chat, Connect, Wallet, Browser, Scribe, Code Workshop, Key Management, COMLA Dashboard, Resonance, Trustmap, and more.
9 NEW V19 PAGES
Content Screening (DCCPv2), DCCP Network, Device Fleet, GPU Contribution, Green Node, Node Types, Resource Class, Tiered Response, and standalone VPN page.
CONSCIOUSNESS HUB
12 pages → tabbed hub: Monitor / Timeline / Compare / Explorer / Metrics / Emotion / Quantum / Presence / Stream / Resting / Cross-feed / State
TRUST ECONOMY HUB
10 pages → tabbed hub: Overview / Contributions (V18 Unified Giving) / Tipping / Grove / Patronage / Staking / Subscriptions / Opportunities / Chain Bridge / Compute Market
GOVERNANCE HUB
6 pages → tabbed hub: Proposals / Results / Delegation / Voting / Treasury / Notary
NETWORK, CREATIVE & MORE
Network Health absorbs 4 pages (Diagnostics, Bootstrap, Mesh, MTCS Bridge). Creative absorbs 3. Learning absorbs 3. Dreams absorbs 2. Settings absorbs 2. Each becomes a single tabbed page.
MINIMUM SPECS
Hardware and software requirements to run a VCNA node
OS
Windows 10/11 (64-bit)
GPU
NVIDIA RTX 2060+ or AMD RX 6600+ with 8GB+ VRAM (for GPU nodes)
CPU
Modern x64 processor with AVX2 support
RAM
16GB recommended, 8GB minimum
STORAGE
10GB base + additional space for AI models
NETWORK
Stable internet connection, UDP port access
CLEAN SLATE ROADMAP
5 critical problems identified • 6 sprint work items • 48 hours total effort • codename “Clean Slate”
W1: BUILD CHAIN FIX (4h)
llama-cpp-sys-2 requires MSVC + CMake to compile.
Fix: pre-built CI artifacts, dev-setup.ps1
one-click installer, VC++ Redistributable bundling in NSIS
installer. Test laptops can test without build tools.
W2: VPN FIX (6h)
VPN can’t connect: no API server deployed, wintun.dll
zipped not extracted, no exit nodes. Fix: extract wintun.dll,
local WireGuard tunnel test mode, diagnostic UI with actionable
messages, standalone vpn.js page.
W3: PAGE PURGE (16h)
160 → ~50 pages. Delete 55 dead pages (robotics, healthcare, NFTs, Phase 14–18 stubs), consolidate 62 into ~15 tabbed hubs. Remove ~5,000 lines from index.html, orphaned .js files, and nav entries.
W4: V19 ECONOMICS UPDATE (10h)
Rewrite contributions flow for V18/V19 Unified Giving: flat 10% GRAT (no sliding scale), 5 intents (not 7 streams), time-based badges, PayPal button. Add passive GRAT earning to wallet, connect, and trust-economy pages.
W5: DEAD CODE REMOVAL (8h)
Remove #![allow(dead_code)] global suppression.
Delete orphaned command modules (robotics, healthcare, NFT,
convergence, neuroscience). Reduce 879 → ~300 Tauri
commands. Fix high-risk unwrap() calls: 1,150
→ ~600.
W6: VERSION BUMP & SMOKE TEST (4h)
Bump to v0.27.3 in Cargo.toml and tauri.conf.json. Target: 0 errors, <30 warnings. Build all 3 variants (CPU, CUDA, Vulkan). 8-point smoke test: launch, chat, connect, VPN, contributions, wallet, nav count, no-crash sweep.
AUDIT METRICS
Pages: 160 → ~50. Commands: 879 → ~300. Rust: 122,800 → ~110,000 lines. JS: 64,900 → ~30,000 lines. index.html: 9,100 → ~4,000 lines. unwrap(): 1,150 → ~600.
5 CRITICAL PROBLEMS
1) Can’t build on test laptops. 2) VPN never connects. 3) 55 dead pages. 4) Contributions uses V17 model. 5) 879 bloated commands. All addressed in this sprint.
EXECUTION ORDER
W1 (Build Fix) → W3a (Delete 55) → W5 (Dead Code) → W3b (Consolidate 62) → W2 (VPN) → W4 (V19 Economics) → W6 (Build & Test). VPN runs parallel with consolidation.
NETWORK SECURITY INNOVATIONS
Three bio-inspired security modules that protect every node in the mesh — zero-signature detection, neural intrusion defense, and unified trust
NETWORK SENTINEL
Bio-inspired neural IDS with 8 brain-region clusters (1,104 lines). Hebbian learning adapts weights from live traffic — no signature database needed. The sentinel learns normal and fires on deviation.
NEURAL MALWARE ENGINE
Zero-signature AI malware detection (1,265 lines). Behavioral DNA profiling fingerprints every binary. Federated herd immunity via GRAT — one node detects, all nodes learn in seconds.
TRUST MESH
8-dimension trust fabric from 10 sources (1,128 lines). Auto-quarantine for suspicious nodes. Trust decays naturally and rebuilds through consistent good behavior.
SYMBIOTE RESOURCE CONTRIBUTION
Every Symbiote auto-detects local hardware and contributes to the network — GPU tier classification, model lifecycle management, and federated LoRA exchange
LOCAL RESOURCE MANAGEMENT
Symbiote auto-detects GPU, CPU, RAM, Storage and Network capabilities on install. Resource class participation is determined automatically — no manual configuration. Hardware profile updates dynamically as conditions change.
GPU TIER DETECTION
Six GPU tiers (T0–T5) based on VRAM and compute capability. Tier determines which models can run locally and which route to the COMLA mesh. Capabilities reported to COMLA via WebSocket for optimal task routing.
FEDERATED LORA CONTRIBUTION
Opt-in anonymous LoRA weight exchange: Symbiotes contribute differential weight updates during idle + charging periods. Resource-aware scheduling ensures contributions never degrade the user experience. GRAT rewards flow to contributors.
WHAT'S NEW
v0.27.3 — Tabbed Navigation, VPN Diagnostics, Green Node Proofs & 160 UI Pages • v5.3.0 Safety Sprint (17 B→A items, ~16,000 lines) • v5.4.0 Network Security (3,497 lines, 31 tests)
TABBED NAVIGATION
10 category accordions in the sidebar, top tab bar with open/close/switch, Ctrl+W and Ctrl+Tab shortcuts. Navigate 160 pages without losing context.
VPN DIAGNOSTICS & FIX
New vpn_diagnose command with status checks panel.
DLL dependency bundling fix — VC++ Redist bundles
automatically. VPN connection reliability improved.
GREEN NODE PROOFS
Proof stats display, Generate Proof button, Submit to Chain button. Contribution proofs are now visible and verifiable in the Green Node panel.
BY THE NUMBERS
160 UI pages. 879 Tauri commands. 197 Rust source files. ~122,800 lines of Rust. ~64,900 lines of frontend JS. Auto-updater from v0.21 → v0.23 verified.
YOUR NODE AWAITS
Download VCNA, install in 60 seconds, and start earning GRAT. Your spare compute helps keep Val free for everyone.