Skip to content

Releases: CavinKrenik/QRES

QRES v19.0.1

02 Feb 03:21
@github-actions github-actions
v19.0.1
bc2ba92

Choose a tag to compare

View all tags
QRES v19.0.1 Latest
Latest

QRES: Decentralized Neural Swarm Operating System for Edge IoT

Cavin Krenik — Olympic College | Published February 2026

📄 Updated: v19.0.1 SecureANDsafe Hardening Complete

v19.0.1 Release Notes: "Secure & Safe"

Release Date: February 2, 2026

Overview

v19.0.1 completes the advanced hardening and algorithmic refinement phase. The swarm is now cryptographically verifiable, formally verified for liveness, and resistant to sophisticated Sybil attacks.

Test Status

81/81 tests passing (cargo test --all --features std)

New Features

1. ZK Transition Proofs (Cryptographic Truth)

  • Sigma protocol proofs for weight transitions using Fiat-Shamir transform (BLAKE3)
  • ZkProtocol trait with prove_transition() method
  • Forged prev_hash causes verification failure - malicious neurons rejected

2. Reputation Tracker (Sybil Resistance)

  • ReputationTracker with per-peer scores: +0.02 for valid ZKP, -0.08 for drift, -0.15 for ZKP failure
  • Ban threshold: score < 0.2
  • Result: 50/50 Sybil nodes banned within 4 rounds, 0% final drift

3. Mid-Flight Join TLA+ Specification (Formal Verification)

  • Full TLA+ spec in research/MidFlightJoin.tla with states: Offline → Joining → Receiving_Summary → Synced
  • Liveness PROVEN under 90% packet loss

4. PreStorm Regime Detection (Predictive Intelligence)

  • 3-point moving average for entropy calculation
  • Entropy derivative triggers "Pre-Storm" state
  • 4-tick early warning before Storm mode

5. Bottleneck Autoencoder (Extreme Efficiency)

  • 4-layer architecture: Input(D) → Hidden(D/2) → Bottleneck(B) → Hidden(D/2) → Output(D)
  • 6.7x compression (22 bytes vs 148 bytes for Summary Genes)

6. BFP-16 VarianceMonitor (Signal Recovery)

  • Auto-tuning for vanishing gradients (<10^-7)
  • Bit-shift correction to re-center precision window
  • Non-zero learning velocity maintained with extremely small weights

v19.0.0 Release Notes: "The Immune System II"

Release Date: February 1, 2026

Overview

v19.0.0 marks the completion of the "Adversarial Hardening" phase. The core network protocol has been upgraded to survive "Inlier Bias" attacks and "Vanishing Gradients" in hostile environments.

Critical Security Constraints

Warning: Mixing v19.0 nodes with v18.0 nodes is NOT supported due to the BFP-16 header format change in the Summary Gene.

New Features

1. Robust Aggregation (TrimmedMeanByz)

Replaces the Krum aggregator.

  • Problem: Krum selects a single existing vector, making it vulnerable if all vectors are slightly biased (Inlier Bias).
  • Solution: Coordinate-wise Trimmed Mean (crates/qres_core/src/aggregation.rs) sorts values per dimension and removes the top/bottom $f$ outliers.
  • Verification: Zero drift observed in Golden Run scenarios where Krum failed.

2. Block Floating Point (BFP-16)

Solves the precision bottleneck of I16F16.

  • Problem: I16F16 has a minimum step of $1.5 \times 10^{-5}$. Gradients at $LR=10^{-5}$ rounded to zero.
  • Solution: Bfp16Vec uses a shared 8-bit exponent and 16-bit integers.
  • Result: Dynamic range of f32 with the storage density of i16.

3. Mid-Flight Onboarding (Summary Gene)

Allows new nodes to join without replaying history.

  • Protocol: Peers exchange a 74-byte SummaryGene containing:
    • Current Consensus State (BFP-16)
    • Variance/Risk Metric (BFP-16)
    • History Hash & Round Index
  • Performance: >99% bandwidth reduction vs v18.0 full sync.

Key Metrics (The "Hero" Stats)

Metric Value
Compression Ratio 31.8x (Peak)
Nodes Simulated 10,000 (Azure Verified)
RAM Overhead < 1 KB per Node
Protocol Success 100%

Abstract

Constrained edge devices in IoT networks face severe limitations in bandwidth and reliability that make traditional Federated Learning (sending MBs of weights) impossible. QRES (Quantum-Relational Encoding System) is a decentralized operating system that replaces heavy weight synchronization with deterministic "silent" consensus.

By combining a Q16.16 fixed-point core with biologically inspired Lamarckian persistence, QRES guarantees bit-perfect reproducibility across heterogeneous hardware (ARM/x86). We empirically verified the system on Microsoft Azure, scaling to 10,000 concurrent nodes on a single commodity vCPU with negligible memory impact ($O(1)$ amortized growth). The system achieves up to 31.8x compression on telemetry data, outperforming standard algorithms like Zstd while maintaining 100% consensus reliability.

Key Contributions

1. "Silent Consensus" via Bit-Perfect Determinism

Replaced non-deterministic floating-point math with a custom Q16.16 Fixed-Point Engine. This allows 10,000+ devices to agree on a model state without transmitting raw weights—if the predictive error is zero, zero bandwidth is used.

2. Massive Scalability ($O(1)$ Memory)

Engineered a no_std Rust actor runtime that leverages allocator amortization. Azure stress tests proved the system can manage 10,000 nodes with <0.70 KB of RAM overhead per node, effectively eliminating memory fragmentation risks for long-running swarms.

3. Lamarckian Persistence (Self-Healing)

Introduced a "GeneStorage" layer that persists learned behaviors across power cycles. Unlike stateless FL clients, QRES nodes recover 100% of their intelligence instantly after a reboot, critical for energy-harvesting IoT hardware.

Experimental Evaluation (v18.0)

1. Verified Scalability (Azure Standard_D2s)

Stress test of the consensus runtime on a single 2-vCPU Cloud VM.

Simulated Nodes Total RAM (MB) RAM / Node Success Rate
1,000 1.72 MB 1.76 KB 100%
5,000 24.64 MB 5.05 KB 100%
10,000 25.83 MB 0.70 KB 100%

2. Compression Efficiency vs. Industry Standard

QRES "Prediction-as-Compression" vs. Zstandard (Facebook).

Dataset Domain QRES Ratio Zstd Ratio Gain
SmoothSine Telemetry 31.8x 2.1x 15x
Wafer Manufacturing 4.98x 3.55x 1.4x
ECG5000 Medical 4.98x 1.8x 2.7x

Technical Stack (v18.0)

Component Technology
Core Rust (no_std, Tokio Async Runtime)
Math Custom Q16.16 Fixed-Point Engine
Infrastructure Azure Cloud (Standard_D2s_v3)
Privacy Differential Privacy ($\epsilon=1.0$) + ECDH Masking

v18.0.0: The Neural Swarm Pivot

Version: v18.0.0 | Released: 2026-01-16

This release pivots from v17.0's deterministic compression to a fully decentralized neural swarm architecture. The system now demonstrates emergent self-healing behavior through hardware-constrained gene gossip and persistent evolutionary memory.

Highlights

Emergent Intelligence

  • Swarm Simulator: Bevy-based God View visualization of 100 nodes in a 10x10 grid
  • Self-Healing Networks: Red (panicked) nodes automatically request cure genes from purple (evolved) neighbors
  • Gossip Protocol: Decentralized gene propagation under MTU fragmentation constraints
  • Visible Evolution: Watch as a single mutation spontaneously appears and spreads to heal the network

Hippocampus: Persistent Evolutionary Memory

  • GeneStorage Trait: Abstract persistence interface (no_std compatible)
  • DiskGeneStorage: Saves evolved genes to ./swarms_memory/ directory
  • Auto-Loading on Spawn: Nodes check for saved genes and spawn as evolved immediately
  • Periodic Persistence: Every 5 seconds, calm evolved nodes save bytecode to disk
  • Lamarckian Evolution: Learned strategies survive simulation restarts and reboots

No_Std Deterministic Core

  • SwarmNeuron Trait: Abstract interface for neural processors across embedded/desktop
  • LinearNeuron: 8-lag linear predictor with entropy tracking and refractory periods
  • Q16.16 Fixed-Point: All math is integer-based for cross-platform determinism
  • Regime Switching: Automatic Calm/Storm/Adapting states based on entropy thresholds

Breaking Changes

  • Predictor Trait Removed: Replaced with SwarmNeuron trait offering broader interface
  • Gene Format: Now supports install_gene() for persistent bytecode loading
  • Simulator Location: Moved from examples/ to tools/swarm_sim/ as full-fledged crate
  • Storage Module: New qres_core/src/cortex/storage.rs adds GeneStorage abstraction

Performance

  • Convergence: Swarm reaches consensus on learned model in <30 seconds under noise
  • Bandwidth: 8 KB/day per node with gene gossip optimization
  • Mutation Rate: ~5% probability per epoch triggers evolution under stress

Migration Guide

  1. Update imports: use qres_core::cortex::{SwarmNeuron, LinearNeuron, GeneStorage}
  2. For custom neurons: implement SwarmNeuron trait instead of Predictor
  3. For storage: implement GeneStorage or use DiskGeneStorage reference implementation
  4. Simulator: cargo run -p swarm_sim --release (previously: cargo run --example swarm_sim)

v17.0.0 Release Notes

Version: v17.0.0 | Released: 2026-01-14

This release introduces Federated Learning capabilities, enabling the swarm to converge on a shared intelligence ("Meta-Brain") through reputation-weighted aggregation.

Highlights

Federated Learning (The Singularity)

  • Reputation-Weighted Averaging: Model updates are weighted by peer reputation and freshness decay
  • Kahan Summation: Prevents floating-point drift during aggregation across thousands of parameters
  • Epoch-Based Aggregation: Updates are buffered and aggregated every 5 seconds for stability
  • Singularity Detection: Automatic detection when global error rate drops below 0.01

Adaptive Precision Switching

  • *...
Read more
Assets 5
Loading

QRES v19.0.0

02 Feb 01:56
@CavinKrenik CavinKrenik
v19.0.0
bdc7bb9

Choose a tag to compare

View all tags
QRES v19.0.0

QRES v19.0.0 - Neural Swarm Operating System

Highlights

  • TrimmedMeanByz aggregation for Byzantine fault tolerance
  • BFP-16 floating-point format for deterministic edge computing
  • Summary gene persistence for Lamarckian evolution
  • Full determinism for bit-perfect consensus

See CHANGELOG.md for full details.

Loading

QRES v18.0.0

17 Jan 03:23
@github-actions github-actions
v18.0.0
94d7db1

Choose a tag to compare

View all tags
QRES v18.0.0

QRES Release Notes

v18.0.0: The Neural Swarm Pivot

Version: v18.0.0 | Released: 2026-01-15

This release pivots from v17.0's deterministic compression to a fully decentralized neural swarm architecture. The system now demonstrates emergent self-healing behavior through hardware-constrained gene gossip and persistent evolutionary memory.

Highlights

Emergent Intelligence

  • Swarm Simulator: Bevy-based God View visualization of 100 nodes in a 10x10 grid
  • Self-Healing Networks: Red (panicked) nodes automatically request cure genes from purple (evolved) neighbors
  • Gossip Protocol: Decentralized gene propagation under MTU fragmentation constraints
  • Visible Evolution: Watch as a single mutation spontaneously appears and spreads to heal the network

Hippocampus: Persistent Evolutionary Memory

  • GeneStorage Trait: Abstract persistence interface (no_std compatible)
  • DiskGeneStorage: Saves evolved genes to ./swarms_memory/ directory
  • Auto-Loading on Spawn: Nodes check for saved genes and spawn as evolved immediately
  • Periodic Persistence: Every 5 seconds, calm evolved nodes save bytecode to disk
  • Lamarckian Evolution: Learned strategies survive simulation restarts and reboots

No_Std Deterministic Core

  • SwarmNeuron Trait: Abstract interface for neural processors across embedded/desktop
  • LinearNeuron: 8-lag linear predictor with entropy tracking and refractory periods
  • Q16.16 Fixed-Point: All math is integer-based for cross-platform determinism
  • Regime Switching: Automatic Calm/Storm/Adapting states based on entropy thresholds

Breaking Changes

  • Predictor Trait Removed: Replaced with SwarmNeuron trait offering broader interface
  • Gene Format: Now supports install_gene() for persistent bytecode loading
  • Simulator Location: Moved from examples/ to tools/swarm_sim/ as full-fledged crate
  • Storage Module: New qres_core/src/cortex/storage.rs adds GeneStorage abstraction

Performance

  • Convergence: Swarm reaches consensus on learned model in <30 seconds under noise
  • Bandwidth: 8 KB/day per node with gene gossip optimization
  • Mutation Rate: ~5% probability per epoch triggers evolution under stress

Migration Guide

  1. Update imports: use qres_core::cortex::{SwarmNeuron, LinearNeuron, GeneStorage}
  2. For custom neurons: implement SwarmNeuron trait instead of Predictor
  3. For storage: implement GeneStorage or use DiskGeneStorage reference implementation
  4. Simulator: cargo run -p swarm_sim --release (previously: cargo run --example swarm_sim)

v17.0.0 Release Notes

Version: v17.0.0 | Released: 2026-01-14

This release introduces Federated Learning capabilities, enabling the swarm to converge on a shared intelligence ("Meta-Brain") through reputation-weighted aggregation.

Highlights

Federated Learning (The Singularity)

  • Reputation-Weighted Averaging: Model updates are weighted by peer reputation and freshness decay
  • Kahan Summation: Prevents floating-point drift during aggregation across thousands of parameters
  • Epoch-Based Aggregation: Updates are buffered and aggregated every 5 seconds for stability
  • Singularity Detection: Automatic detection when global error rate drops below 0.01

Adaptive Precision Switching

  • Calm Mode (I16F16): Full precision for normal operation
  • Storm Mode (I8F8): Reduced precision during high-throughput events
  • Automatic Switching: Based on entropy and throughput thresholds

Enhanced Security

  • ZK Proofs: Curve25519-based zero-knowledge proofs for model updates
  • Differential Privacy: ε-DP with configurable privacy budget
  • Reputation Gating: Trust-based acceptance of proof-less messages

Performance Improvements

  • Bandwidth: 8KB/day vs 2.3GB/day for traditional federated learning
  • Convergence: <30 epochs for swarm consensus
  • Determinism: Bit-perfect reproducibility across platforms

Breaking Changes

  • Updated all crate versions to 17.0.0
  • Removed internal codenames from release artifacts

QRES v16.5.0 Release Notes

Codename: "The Immune System" | Released: 2026-01-14

"Identity without Exposure. Trust without Centralization."

This release introduces the QRES Immune System—a comprehensive security stack designed to protect the decentralized "Living Brain" from adversarial attacks while preserving the privacy of edge contributors.

Highlights

The Ghost Protocol (Privacy Stack)

We have implemented a Defense-in-Depth privacy layer that ensures no single peer or component can see the raw model updates:

  1. Differential Privacy (Noise Layer): Deterministic Gaussian noise is added to the I16F16 weights before they leave the device.
  2. Secure Aggregation (Masking Layer): Peers establish pairwise shared secrets (X25519) to mask their updates. The Aggregator sees only the global sum, as individual masks cancel out mathematically.
  3. Zero-Knowledge Proofs (Verification Layer): Peers attach NormProofs (Pedersen Commitments) proving their masked update is bounded (not garbage) without revealing the update itself.

Trust & Reputation (The Gatekeeper)

The swarm now actively filters participation based on "Mathematical Merit":

  • Reputation Manager: A persistent trust score tracks peer behavior.
    • Accepted Update: +0.01 Trust
    • Krum Rejection: -0.1 Trust
    • Ban Threshold: Trust < 0.2
  • Identity Binding: Aggregation results are now cryptographically bound to the sender's Ed25519 identity, enabling long-term accountability.

Hardened Federated Dreaming

  • Sanity Checks: The "Dreaming" process (Generative Replay) now validates synthetic weights against a local buffer of real data before applying them, preventing "hallucinations" or model poisoning via synthesis.

Changes

Core (qres_core)

  • Added privacy module with add_noise_fixed for I16F16 support.
  • Added secure_agg module with mask_update_fixed and strict X25519 key agreement.
  • Added zk_proofs module with ProofBundle and verify_batch.
  • Added packet module defining the GhostUpdate structure.

Daemon (qres_daemon)

  • Integrated ReputationManager into AppState.
  • Updated BrainAggregator to return accepted/rejected peer lists for scoring.
  • Updated SwarmP2P message loop to handle reputation rewards/punishments.

Breaking Changes

  • Protocol Update: The peer-to-peer message format has changed to support GhostUpdate packets. v16.5 nodes cannot federate with v16.0 nodes.
  • Config: reputation.json is now required (automatically created if missing).

Upgrade Guide

# Update Rust Toolchain
rustup update stable

# Pull latest
git pull origin main

# Build
cargo build --release

QRES v16.0.0 Release Notes

v16.0.0 - The "Systems" Update

Release Date: January 13, 2026
Focus: Determinism, Safety, and Zero-Copy Performance.

Major Changes

  • Breaking: compress_chunk now requires a pre-allocated &mut [u8] buffer (Zero-Copy).
  • Feat: Replaced floating-point math with fixed::types::I16F16 for bit-perfect cross-arch consensus.
  • Security: Removed all panic paths (unwrap, expect) from the no_std core.
  • Structure: Monorepo split into crates/ (Production) and research/ (Experiments).

Bug Fixes

  • Fixed "Link Explosion" in P2P sync by implementing Deterministic Seed Sync (8 KB/day).
  • Fixed "Expansion Problem" via Hybrid Gatekeeper (fallback to bit-packing on high entropy).

QRES v16.0.0 - Pre-Release Notes

Date: January 13, 2026
Title: QRES: Adapter Hybrid Compression System

Major Features

1. Hybrid Conditional Pipeline

QRES now dynamically switches between two codec paths based on real-time data entropy (< 7.5 bits/byte threshold):

  • Bit-Packing Path: High-speed Delta+ZigZag+BitPack algorithm. (Used for Grid/Noise data)
  • Neural-Enhanced Path: Neural residual prediction for structured data. (Used for Weather/ECG)

2. Validated Benchmarks (2.75x - 24.9x)

Comprehensive benchmarking across 7 diverse datasets confirms QRES outperforms standard predictors:

  • SmoothSine: 24.9x
  • Jena Climate: 4.9x
  • ItalyPower: 4.6x
  • Wafer: 4.2x
  • ECG5000: 4.0x
  • ETTh1: 2.8x

3. Production-Ready Core

  • bitpack.rs: Integrated validated bit-packing logic directly into qres_core.
  • qres_core API: exposed compress_adaptive and decompress_adaptive for easy integration.
  • Fixed-Point Arithmetic: Q16.16 math ensures cross-platform determinism (x86/ARM/WASM).

4. Documentation Overhaul

  • New Paper: "QRES: An Adaptive Hybrid Compression System for Edge IoT" (PDF available in docs/paper/)
  • Theory Docs: "Living Brain" architecture details moved to docs/THEORY.md.
  • Roadmap: v16 milestones marked complete.

Fixes

  • Fixed "Metric Fallacy" in benchmarks (now measuring against raw 4-byte f32).
  • Fixed CI/CD failures related to missing data directories.
  • Resolved cargo fmt and clippy lints.
  • Hotfix: Restored compress_adaptive Python alias for backward compatibility.
  • Hotfix: Resolved Tauri plugin version mismatch.

QRES v15.4.0 Release Notes

Release Date: January 11, 2026

Overview

v15.4.0 introduces Hardware-in-the-Loop Simulation using real-world climate data, along with major visualization upgrades to the Hive Mind and Neural Graph pages.

New Features

Weather Replay Engine

  • Real-World Data: Integrates the Jena Climate Dataset (Max Planck Institute) for high-fidelity sensor simulation
  • Storm Detection: Maps atmospheric pressure drops to vibration spikes, triggering LEARNING mode
  • Debug Panel: Real-time display of Frame index, Pressure (mbar), and Compression ratio

Hive Mind: Intera...

Read more
Loading

QRES v17.0.0

14 Jan 20:57
@github-actions github-actions
v17.0.0
f90608a

Choose a tag to compare

View all tags
QRES v17.0.0

QRES v17.0.0 Release Notes

Version: v17.0.0 | Released: 2026-01-14

This release introduces Federated Learning capabilities, enabling the swarm to converge on a shared intelligence ("Meta-Brain") through reputation-weighted aggregation.

Highlights

Federated Learning (The Singularity)

  • Reputation-Weighted Averaging: Model updates are weighted by peer reputation and freshness decay
  • Kahan Summation: Prevents floating-point drift during aggregation across thousands of parameters
  • Epoch-Based Aggregation: Updates are buffered and aggregated every 5 seconds for stability
  • Singularity Detection: Automatic detection when global error rate drops below 0.01

Adaptive Precision Switching

  • Calm Mode (I16F16): Full precision for normal operation
  • Storm Mode (I8F8): Reduced precision during high-throughput events
  • Automatic Switching: Based on entropy and throughput thresholds

Enhanced Security

  • ZK Proofs: Curve25519-based zero-knowledge proofs for model updates
  • Differential Privacy: ε-DP with configurable privacy budget
  • Reputation Gating: Trust-based acceptance of proof-less messages

Performance Improvements

  • Bandwidth: 8KB/day vs 2.3GB/day for traditional federated learning
  • Convergence: <30 epochs for swarm consensus
  • Determinism: Bit-perfect reproducibility across platforms

Breaking Changes

  • Updated all crate versions to 17.0.0
  • Removed internal codenames from release artifacts

QRES v16.5.0 Release Notes

Codename: "The Immune System" | Released: 2026-01-14

"Identity without Exposure. Trust without Centralization."

This release introduces the QRES Immune System—a comprehensive security stack designed to protect the decentralized "Living Brain" from adversarial attacks while preserving the privacy of edge contributors.

Highlights

The Ghost Protocol (Privacy Stack)

We have implemented a Defense-in-Depth privacy layer that ensures no single peer or component can see the raw model updates:

  1. Differential Privacy (Noise Layer): Deterministic Gaussian noise is added to the I16F16 weights before they leave the device.
  2. Secure Aggregation (Masking Layer): Peers establish pairwise shared secrets (X25519) to mask their updates. The Aggregator sees only the global sum, as individual masks cancel out mathematically.
  3. Zero-Knowledge Proofs (Verification Layer): Peers attach NormProofs (Pedersen Commitments) proving their masked update is bounded (not garbage) without revealing the update itself.

Trust & Reputation (The Gatekeeper)

The swarm now actively filters participation based on "Mathematical Merit":

  • Reputation Manager: A persistent trust score tracks peer behavior.
    • Accepted Update: +0.01 Trust
    • Krum Rejection: -0.1 Trust
    • Ban Threshold: Trust < 0.2
  • Identity Binding: Aggregation results are now cryptographically bound to the sender's Ed25519 identity, enabling long-term accountability.

Hardened Federated Dreaming

  • Sanity Checks: The "Dreaming" process (Generative Replay) now validates synthetic weights against a local buffer of real data before applying them, preventing "hallucinations" or model poisoning via synthesis.

Changes

Core (qres_core)

  • Added privacy module with add_noise_fixed for I16F16 support.
  • Added secure_agg module with mask_update_fixed and strict X25519 key agreement.
  • Added zk_proofs module with ProofBundle and verify_batch.
  • Added packet module defining the GhostUpdate structure.

Daemon (qres_daemon)

  • Integrated ReputationManager into AppState.
  • Updated BrainAggregator to return accepted/rejected peer lists for scoring.
  • Updated SwarmP2P message loop to handle reputation rewards/punishments.

Breaking Changes

  • Protocol Update: The peer-to-peer message format has changed to support GhostUpdate packets. v16.5 nodes cannot federate with v16.0 nodes.
  • Config: reputation.json is now required (automatically created if missing).

Upgrade Guide

# Update Rust Toolchain
rustup update stable

# Pull latest
git pull origin main

# Build
cargo build --release

QRES v16.0.0 Release Notes

v16.0.0 - The "Systems" Update

Release Date: January 13, 2026
Focus: Determinism, Safety, and Zero-Copy Performance.

Major Changes

  • Breaking: compress_chunk now requires a pre-allocated &mut [u8] buffer (Zero-Copy).
  • Feat: Replaced floating-point math with fixed::types::I16F16 for bit-perfect cross-arch consensus.
  • Security: Removed all panic paths (unwrap, expect) from the no_std core.
  • Structure: Monorepo split into crates/ (Production) and research/ (Experiments).

Bug Fixes

  • Fixed "Link Explosion" in P2P sync by implementing Deterministic Seed Sync (8 KB/day).
  • Fixed "Expansion Problem" via Hybrid Gatekeeper (fallback to bit-packing on high entropy).

QRES v16.0.0 - Pre-Release Notes

Date: January 13, 2026
Title: QRES: Adapter Hybrid Compression System

Major Features

1. Hybrid Conditional Pipeline

QRES now dynamically switches between two codec paths based on real-time data entropy (< 7.5 bits/byte threshold):

  • Bit-Packing Path: High-speed Delta+ZigZag+BitPack algorithm. (Used for Grid/Noise data)
  • Neural-Enhanced Path: Neural residual prediction for structured data. (Used for Weather/ECG)

2. Validated Benchmarks (2.75x - 24.9x)

Comprehensive benchmarking across 7 diverse datasets confirms QRES outperforms standard predictors:

  • SmoothSine: 24.9x
  • Jena Climate: 4.9x
  • ItalyPower: 4.6x
  • Wafer: 4.2x
  • ECG5000: 4.0x
  • ETTh1: 2.8x

3. Production-Ready Core

  • bitpack.rs: Integrated validated bit-packing logic directly into qres_core.
  • qres_core API: exposed compress_adaptive and decompress_adaptive for easy integration.
  • Fixed-Point Arithmetic: Q16.16 math ensures cross-platform determinism (x86/ARM/WASM).

4. Documentation Overhaul

  • New Paper: "QRES: An Adaptive Hybrid Compression System for Edge IoT" (PDF available in docs/paper/)
  • Theory Docs: "Living Brain" architecture details moved to docs/THEORY.md.
  • Roadmap: v16 milestones marked complete.

Fixes

  • Fixed "Metric Fallacy" in benchmarks (now measuring against raw 4-byte f32).
  • Fixed CI/CD failures related to missing data directories.
  • Resolved cargo fmt and clippy lints.
  • Hotfix: Restored compress_adaptive Python alias for backward compatibility.
  • Hotfix: Resolved Tauri plugin version mismatch.

QRES v15.4.0 Release Notes

Release Date: January 11, 2026

Overview

v15.4.0 introduces Hardware-in-the-Loop Simulation using real-world climate data, along with major visualization upgrades to the Hive Mind and Neural Graph pages.

New Features

Weather Replay Engine

  • Real-World Data: Integrates the Jena Climate Dataset (Max Planck Institute) for high-fidelity sensor simulation
  • Storm Detection: Maps atmospheric pressure drops to vibration spikes, triggering LEARNING mode
  • Debug Panel: Real-time display of Frame index, Pressure (mbar), and Compression ratio

Hive Mind: Interactive Neural Swarm

  • Infinite Canvas: Zoom (0.1x-8x) and pan controls for exploring large networks
  • Node Inspector HUD: Click any node to view IP, CPU load, Memory, and Status
  • Gradient Packets: Animated particles flow between nodes when streaming is active

Neural Graph: Deep Learning Visualization

  • Layered Architecture: 5-layer deep network (Input → Hidden A/B → Attention → Output)
  • Live Spike Propagation: Visual pulses travel from input sensors to output nodes
  • Reactive to Data: Input nodes flash based on real telemetry intensity

Improvements

UI/UX Enhancements

  • Single Connect Button: Removed duplicate header button; swarm toggle in Edge Swarm panel only
  • Clean Sidebar: Text-only navigation labels (no icons)
  • No-Scroll Layout: Dashboard now fits entirely in viewport

Architecture

  • Simulated Compression: Browser mode uses realistic compression ratios (~4-6:1) without requiring WASM
  • ResizeObserver: Charts properly resize and fill available space
  • Vite Config: Updated server.fs.allow for WASM file access

Documentation

  • README: Added "Hardware-in-the-Loop Simulation" section
  • Release Notes: Updated v15.3.0 notes with simulation features

Upgrade Instructions

# 1. Pull latest
git pull origin main

# 2. Install dependencies
cd web && npm install

# 3. (Optional) Fetch weather data
python3 scripts/fetch_weather_replay.py

# 4. Launch dashboard
npm run dev

Metrics

Metric v15.3.0 v15.4.0
Startup Time ~1.6s ~1.5s
Bundle Size 1.4MB 1.5MB
Visualization FPS 30 60

Full Changelog: v15.3.0...v15.4.0

Full Changelog: v16.5.0...v17.0.0

Loading

QRES v16.5.0 - The Hybrid Era

14 Jan 15:19
@github-actions github-actions
v16.5.0
de8b7d8

Choose a tag to compare

View all tags
QRES v16.5.0 - The Hybrid Era

QRES v16.5.0 Release Notes

Codename: "The Immune System" | Released: 2026-01-14

"Identity without Exposure. Trust without Centralization."

This release introduces the QRES Immune System—a comprehensive security stack designed to protect the decentralized "Living Brain" from adversarial attacks while preserving the privacy of edge contributors.

Highlights

The Ghost Protocol (Privacy Stack)

We have implemented a Defense-in-Depth privacy layer that ensures no single peer or component can see the raw model updates:

  1. Differential Privacy (Noise Layer): Deterministic Gaussian noise is added to the I16F16 weights before they leave the device.
  2. Secure Aggregation (Masking Layer): Peers establish pairwise shared secrets (X25519) to mask their updates. The Aggregator sees only the global sum, as individual masks cancel out mathematically.
  3. Zero-Knowledge Proofs (Verification Layer): Peers attach NormProofs (Pedersen Commitments) proving their masked update is bounded (not garbage) without revealing the update itself.

Trust & Reputation (The Gatekeeper)

The swarm now actively filters participation based on "Mathematical Merit":

  • Reputation Manager: A persistent trust score tracks peer behavior.
    • Accepted Update: +0.01 Trust
    • Krum Rejection: -0.1 Trust
    • Ban Threshold: Trust < 0.2
  • Identity Binding: Aggregation results are now cryptographically bound to the sender's Ed25519 identity, enabling long-term accountability.

Hardened Federated Dreaming

  • Sanity Checks: The "Dreaming" process (Generative Replay) now validates synthetic weights against a local buffer of real data before applying them, preventing "hallucinations" or model poisoning via synthesis.

Changes

Core (qres_core)

  • Added privacy module with add_noise_fixed for I16F16 support.
  • Added secure_agg module with mask_update_fixed and strict X25519 key agreement.
  • Added zk_proofs module with ProofBundle and verify_batch.
  • Added packet module defining the GhostUpdate structure.

Daemon (qres_daemon)

  • Integrated ReputationManager into AppState.
  • Updated BrainAggregator to return accepted/rejected peer lists for scoring.
  • Updated SwarmP2P message loop to handle reputation rewards/punishments.

Breaking Changes

  • Protocol Update: The peer-to-peer message format has changed to support GhostUpdate packets. v16.5 nodes cannot federate with v16.0 nodes.
  • Config: reputation.json is now required (automatically created if missing).

Upgrade Guide

# Update Rust Toolchain
rustup update stable

# Pull latest
git pull origin main

# Build
cargo build --release

QRES v16.0.0 Release Notes

v16.0.0 - The "Systems" Update

Release Date: January 13, 2026
Focus: Determinism, Safety, and Zero-Copy Performance.

Major Changes

  • Breaking: compress_chunk now requires a pre-allocated &mut [u8] buffer (Zero-Copy).
  • Feat: Replaced floating-point math with fixed::types::I16F16 for bit-perfect cross-arch consensus.
  • Security: Removed all panic paths (unwrap, expect) from the no_std core.
  • Structure: Monorepo split into crates/ (Production) and research/ (Experiments).

Bug Fixes

  • Fixed "Link Explosion" in P2P sync by implementing Deterministic Seed Sync (8 KB/day).
  • Fixed "Expansion Problem" via Hybrid Gatekeeper (fallback to bit-packing on high entropy).

QRES v16.0.0 - Pre-Release Notes

Date: January 13, 2026
Title: QRES: Adapter Hybrid Compression System

Major Features

1. Hybrid Conditional Pipeline

QRES now dynamically switches between two codec paths based on real-time data entropy (< 7.5 bits/byte threshold):

  • Bit-Packing Path: High-speed Delta+ZigZag+BitPack algorithm. (Used for Grid/Noise data)
  • Neural-Enhanced Path: Neural residual prediction for structured data. (Used for Weather/ECG)

2. Validated Benchmarks (2.75x - 24.9x)

Comprehensive benchmarking across 7 diverse datasets confirms QRES outperforms standard predictors:

  • SmoothSine: 24.9x
  • Jena Climate: 4.9x
  • ItalyPower: 4.6x
  • Wafer: 4.2x
  • ECG5000: 4.0x
  • ETTh1: 2.8x

3. Production-Ready Core

  • bitpack.rs: Integrated validated bit-packing logic directly into qres_core.
  • qres_core API: exposed compress_adaptive and decompress_adaptive for easy integration.
  • Fixed-Point Arithmetic: Q16.16 math ensures cross-platform determinism (x86/ARM/WASM).

4. Documentation Overhaul

  • New Paper: "QRES: An Adaptive Hybrid Compression System for Edge IoT" (PDF available in docs/paper/)
  • Theory Docs: "Living Brain" architecture details moved to docs/THEORY.md.
  • Roadmap: v16 milestones marked complete.

Fixes

  • Fixed "Metric Fallacy" in benchmarks (now measuring against raw 4-byte f32).
  • Fixed CI/CD failures related to missing data directories.
  • Resolved cargo fmt and clippy lints.
  • Hotfix: Restored compress_adaptive Python alias for backward compatibility.
  • Hotfix: Resolved Tauri plugin version mismatch.

QRES v15.4.0 Release Notes

Release Date: January 11, 2026

Overview

v15.4.0 introduces Hardware-in-the-Loop Simulation using real-world climate data, along with major visualization upgrades to the Hive Mind and Neural Graph pages.

New Features

Weather Replay Engine

  • Real-World Data: Integrates the Jena Climate Dataset (Max Planck Institute) for high-fidelity sensor simulation
  • Storm Detection: Maps atmospheric pressure drops to vibration spikes, triggering LEARNING mode
  • Debug Panel: Real-time display of Frame index, Pressure (mbar), and Compression ratio

Hive Mind: Interactive Neural Swarm

  • Infinite Canvas: Zoom (0.1x-8x) and pan controls for exploring large networks
  • Node Inspector HUD: Click any node to view IP, CPU load, Memory, and Status
  • Gradient Packets: Animated particles flow between nodes when streaming is active

Neural Graph: Deep Learning Visualization

  • Layered Architecture: 5-layer deep network (Input → Hidden A/B → Attention → Output)
  • Live Spike Propagation: Visual pulses travel from input sensors to output nodes
  • Reactive to Data: Input nodes flash based on real telemetry intensity

Improvements

UI/UX Enhancements

  • Single Connect Button: Removed duplicate header button; swarm toggle in Edge Swarm panel only
  • Clean Sidebar: Text-only navigation labels (no icons)
  • No-Scroll Layout: Dashboard now fits entirely in viewport

Architecture

  • Simulated Compression: Browser mode uses realistic compression ratios (~4-6:1) without requiring WASM
  • ResizeObserver: Charts properly resize and fill available space
  • Vite Config: Updated server.fs.allow for WASM file access

Documentation

  • README: Added "Hardware-in-the-Loop Simulation" section
  • Release Notes: Updated v15.3.0 notes with simulation features

Upgrade Instructions

# 1. Pull latest
git pull origin main

# 2. Install dependencies
cd web && npm install

# 3. (Optional) Fetch weather data
python3 scripts/fetch_weather_replay.py

# 4. Launch dashboard
npm run dev

Metrics

Metric v15.3.0 v15.4.0
Startup Time ~1.6s ~1.5s
Bundle Size 1.4MB 1.5MB
Visualization FPS 30 60

Full Changelog: v15.3.0...v15.4.0

Full Changelog: v16.0.0...v16.5.0

Loading

QRES v16.5-ghost - The Hybrid Era

14 Jan 19:22
@github-actions github-actions
v16.5-ghost
a94b648

Choose a tag to compare

View all tags
QRES v16.5-ghost - The Hybrid Era

QRES v16.5.0 Release Notes

Codename: "The Immune System" | Released: 2026-01-14

"Identity without Exposure. Trust without Centralization."

This release introduces the QRES Immune System—a comprehensive security stack designed to protect the decentralized "Living Brain" from adversarial attacks while preserving the privacy of edge contributors.

Highlights

The Ghost Protocol (Privacy Stack)

We have implemented a Defense-in-Depth privacy layer that ensures no single peer or component can see the raw model updates:

  1. Differential Privacy (Noise Layer): Deterministic Gaussian noise is added to the I16F16 weights before they leave the device.
  2. Secure Aggregation (Masking Layer): Peers establish pairwise shared secrets (X25519) to mask their updates. The Aggregator sees only the global sum, as individual masks cancel out mathematically.
  3. Zero-Knowledge Proofs (Verification Layer): Peers attach NormProofs (Pedersen Commitments) proving their masked update is bounded (not garbage) without revealing the update itself.

Trust & Reputation (The Gatekeeper)

The swarm now actively filters participation based on "Mathematical Merit":

  • Reputation Manager: A persistent trust score tracks peer behavior.
    • Accepted Update: +0.01 Trust
    • Krum Rejection: -0.1 Trust
    • Ban Threshold: Trust < 0.2
  • Identity Binding: Aggregation results are now cryptographically bound to the sender's Ed25519 identity, enabling long-term accountability.

Hardened Federated Dreaming

  • Sanity Checks: The "Dreaming" process (Generative Replay) now validates synthetic weights against a local buffer of real data before applying them, preventing "hallucinations" or model poisoning via synthesis.

Changes

Core (qres_core)

  • Added privacy module with add_noise_fixed for I16F16 support.
  • Added secure_agg module with mask_update_fixed and strict X25519 key agreement.
  • Added zk_proofs module with ProofBundle and verify_batch.
  • Added packet module defining the GhostUpdate structure.

Daemon (qres_daemon)

  • Integrated ReputationManager into AppState.
  • Updated BrainAggregator to return accepted/rejected peer lists for scoring.
  • Updated SwarmP2P message loop to handle reputation rewards/punishments.

Breaking Changes

  • Protocol Update: The peer-to-peer message format has changed to support GhostUpdate packets. v16.5 nodes cannot federate with v16.0 nodes.
  • Config: reputation.json is now required (automatically created if missing).

Upgrade Guide

# Update Rust Toolchain
rustup update stable

# Pull latest
git pull origin main

# Build
cargo build --release

QRES v16.0.0 Release Notes

v16.0.0 - The "Systems" Update

Release Date: January 13, 2026
Focus: Determinism, Safety, and Zero-Copy Performance.

Major Changes

  • Breaking: compress_chunk now requires a pre-allocated &mut [u8] buffer (Zero-Copy).
  • Feat: Replaced floating-point math with fixed::types::I16F16 for bit-perfect cross-arch consensus.
  • Security: Removed all panic paths (unwrap, expect) from the no_std core.
  • Structure: Monorepo split into crates/ (Production) and research/ (Experiments).

Bug Fixes

  • Fixed "Link Explosion" in P2P sync by implementing Deterministic Seed Sync (8 KB/day).
  • Fixed "Expansion Problem" via Hybrid Gatekeeper (fallback to bit-packing on high entropy).

QRES v16.0.0 - Pre-Release Notes

Date: January 13, 2026
Title: QRES: Adapter Hybrid Compression System

Major Features

1. Hybrid Conditional Pipeline

QRES now dynamically switches between two codec paths based on real-time data entropy (< 7.5 bits/byte threshold):

  • Bit-Packing Path: High-speed Delta+ZigZag+BitPack algorithm. (Used for Grid/Noise data)
  • Neural-Enhanced Path: Neural residual prediction for structured data. (Used for Weather/ECG)

2. Validated Benchmarks (2.75x - 24.9x)

Comprehensive benchmarking across 7 diverse datasets confirms QRES outperforms standard predictors:

  • SmoothSine: 24.9x
  • Jena Climate: 4.9x
  • ItalyPower: 4.6x
  • Wafer: 4.2x
  • ECG5000: 4.0x
  • ETTh1: 2.8x

3. Production-Ready Core

  • bitpack.rs: Integrated validated bit-packing logic directly into qres_core.
  • qres_core API: exposed compress_adaptive and decompress_adaptive for easy integration.
  • Fixed-Point Arithmetic: Q16.16 math ensures cross-platform determinism (x86/ARM/WASM).

4. Documentation Overhaul

  • New Paper: "QRES: An Adaptive Hybrid Compression System for Edge IoT" (PDF available in docs/paper/)
  • Theory Docs: "Living Brain" architecture details moved to docs/THEORY.md.
  • Roadmap: v16 milestones marked complete.

Fixes

  • Fixed "Metric Fallacy" in benchmarks (now measuring against raw 4-byte f32).
  • Fixed CI/CD failures related to missing data directories.
  • Resolved cargo fmt and clippy lints.
  • Hotfix: Restored compress_adaptive Python alias for backward compatibility.
  • Hotfix: Resolved Tauri plugin version mismatch.

QRES v15.4.0 Release Notes

Release Date: January 11, 2026

Overview

v15.4.0 introduces Hardware-in-the-Loop Simulation using real-world climate data, along with major visualization upgrades to the Hive Mind and Neural Graph pages.

New Features

Weather Replay Engine

  • Real-World Data: Integrates the Jena Climate Dataset (Max Planck Institute) for high-fidelity sensor simulation
  • Storm Detection: Maps atmospheric pressure drops to vibration spikes, triggering LEARNING mode
  • Debug Panel: Real-time display of Frame index, Pressure (mbar), and Compression ratio

Hive Mind: Interactive Neural Swarm

  • Infinite Canvas: Zoom (0.1x-8x) and pan controls for exploring large networks
  • Node Inspector HUD: Click any node to view IP, CPU load, Memory, and Status
  • Gradient Packets: Animated particles flow between nodes when streaming is active

Neural Graph: Deep Learning Visualization

  • Layered Architecture: 5-layer deep network (Input → Hidden A/B → Attention → Output)
  • Live Spike Propagation: Visual pulses travel from input sensors to output nodes
  • Reactive to Data: Input nodes flash based on real telemetry intensity

Improvements

UI/UX Enhancements

  • Single Connect Button: Removed duplicate header button; swarm toggle in Edge Swarm panel only
  • Clean Sidebar: Text-only navigation labels (no icons)
  • No-Scroll Layout: Dashboard now fits entirely in viewport

Architecture

  • Simulated Compression: Browser mode uses realistic compression ratios (~4-6:1) without requiring WASM
  • ResizeObserver: Charts properly resize and fill available space
  • Vite Config: Updated server.fs.allow for WASM file access

Documentation

  • README: Added "Hardware-in-the-Loop Simulation" section
  • Release Notes: Updated v15.3.0 notes with simulation features

Upgrade Instructions

# 1. Pull latest
git pull origin main

# 2. Install dependencies
cd web && npm install

# 3. (Optional) Fetch weather data
python3 scripts/fetch_weather_replay.py

# 4. Launch dashboard
npm run dev

Metrics

Metric v15.3.0 v15.4.0
Startup Time ~1.6s ~1.5s
Bundle Size 1.4MB 1.5MB
Visualization FPS 30 60

Full Changelog: v15.3.0...v15.4.0

Full Changelog: v16.5.0...v16.5-ghost

Loading

QRES v16.0.0

13 Jan 23:47
@github-actions github-actions
v16.0.0
a18966a

Choose a tag to compare

View all tags
QRES v16.0.0

QRES v16.0.0 Release Notes

Date: January 13, 2026
Title: QRES: Adapter Hybrid Compression System

Major Features

1. Hybrid Conditional Pipeline

QRES now dynamically switches between two codec paths based on real-time data entropy (< 7.5 bits/byte threshold):

  • Bit-Packing Path: High-speed Delta+ZigZag+BitPack algorithm. (Used for Grid/Noise data)
  • Neural-Enhanced Path: Neural residual prediction for structured data. (Used for Weather/ECG)

2. Validated Benchmarks (2.75x - 24.9x)

Comprehensive benchmarking across 7 diverse datasets confirms QRES outperforms standard predictors:

  • SmoothSine: 24.9x
  • Jena Climate: 4.9x
  • ItalyPower: 4.6x
  • Wafer: 4.2x
  • ECG5000: 4.0x
  • ETTh1: 2.8x

3. Production-Ready Core

  • bitpack.rs: Integrated validated bit-packing logic directly into qres_core.
  • qres_core API: exposed compress_adaptive and decompress_adaptive for easy integration.
  • Fixed-Point Arithmetic: Q16.16 math ensures cross-platform determinism (x86/ARM/WASM).

4. Documentation Overhaul

  • New Paper: "QRES: An Adaptive Hybrid Compression System for Edge IoT" (PDF available in docs/paper/)
  • Theory Docs: "Living Brain" architecture details moved to docs/THEORY.md.
  • Roadmap: v16 milestones marked complete.

Fixes

  • Fixed "Metric Fallacy" in benchmarks (now measuring against raw 4-byte f32).
  • Fixed CI/CD failures related to missing data directories.
  • Resolved cargo fmt and clippy lints.
  • Hotfix: Restored compress_adaptive Python alias for backward compatibility.
  • Hotfix: Resolved Tauri plugin version mismatch.

QRES v15.4.0 Release Notes

Release Date: January 11, 2026

Overview

v15.4.0 introduces Hardware-in-the-Loop Simulation using real-world climate data, along with major visualization upgrades to the Hive Mind and Neural Graph pages.

New Features

Weather Replay Engine

  • Real-World Data: Integrates the Jena Climate Dataset (Max Planck Institute) for high-fidelity sensor simulation
  • Storm Detection: Maps atmospheric pressure drops to vibration spikes, triggering LEARNING mode
  • Debug Panel: Real-time display of Frame index, Pressure (mbar), and Compression ratio

Hive Mind: Interactive Neural Swarm

  • Infinite Canvas: Zoom (0.1x-8x) and pan controls for exploring large networks
  • Node Inspector HUD: Click any node to view IP, CPU load, Memory, and Status
  • Gradient Packets: Animated particles flow between nodes when streaming is active

Neural Graph: Deep Learning Visualization

  • Layered Architecture: 5-layer deep network (Input → Hidden A/B → Attention → Output)
  • Live Spike Propagation: Visual pulses travel from input sensors to output nodes
  • Reactive to Data: Input nodes flash based on real telemetry intensity

Improvements

UI/UX Enhancements

  • Single Connect Button: Removed duplicate header button; swarm toggle in Edge Swarm panel only
  • Clean Sidebar: Text-only navigation labels (no icons)
  • No-Scroll Layout: Dashboard now fits entirely in viewport

Architecture

  • Simulated Compression: Browser mode uses realistic compression ratios (~4-6:1) without requiring WASM
  • ResizeObserver: Charts properly resize and fill available space
  • Vite Config: Updated server.fs.allow for WASM file access

Documentation

  • README: Added "Hardware-in-the-Loop Simulation" section
  • Release Notes: Updated v15.3.0 notes with simulation features

Upgrade Instructions

# 1. Pull latest
git pull origin main

# 2. Install dependencies
cd qres-studio && npm install

# 3. (Optional) Fetch weather data
python3 scripts/fetch_weather_replay.py

# 4. Launch dashboard
npm run dev

Metrics

Metric v15.3.0 v15.4.0
Startup Time ~1.6s ~1.5s
Bundle Size 1.4MB 1.5MB
Visualization FPS 30 60

Full Changelog: v15.3.0...v15.4.0

Full Changelog: v15.4.0...v16.0.0

Loading

QRES v15.4.0

11 Jan 22:13
@github-actions github-actions
v15.4.0
b0553d6

Choose a tag to compare

View all tags
QRES v15.4.0

QRES v15.4.0 Release Notes

Release Date: January 11, 2026
Codename: Neural Swarm

🎉 Overview

v15.4.0 introduces Hardware-in-the-Loop Simulation using real-world climate data, along with major visualization upgrades to the Hive Mind and Neural Graph pages.

✨ New Features

🌡️ Weather Replay Engine

  • Real-World Data: Integrates the Jena Climate Dataset (Max Planck Institute) for high-fidelity sensor simulation
  • Storm Detection: Maps atmospheric pressure drops to vibration spikes, triggering LEARNING mode
  • Debug Panel: Real-time display of Frame index, Pressure (mbar), and Compression ratio

🕸️ Hive Mind: Interactive Neural Swarm

  • Infinite Canvas: Zoom (0.1x-8x) and pan controls for exploring large networks
  • Node Inspector HUD: Click any node to view IP, CPU load, Memory, and Status
  • Gradient Packets: Animated particles flow between nodes when streaming is active

🧠 Neural Graph: Deep Learning Visualization

  • Layered Architecture: 5-layer deep network (Input → Hidden A/B → Attention → Output)
  • Live Spike Propagation: Visual pulses travel from input sensors to output nodes
  • Reactive to Data: Input nodes flash based on real telemetry intensity

🔧 Improvements

UI/UX Enhancements

  • Single Connect Button: Removed duplicate header button; swarm toggle in Edge Swarm panel only
  • Clean Sidebar: Text-only navigation labels (no icons)
  • No-Scroll Layout: Dashboard now fits entirely in viewport

Architecture

  • Simulated Compression: Browser mode uses realistic compression ratios (~4-6:1) without requiring WASM
  • ResizeObserver: Charts properly resize and fill available space
  • Vite Config: Updated server.fs.allow for WASM file access

📝 Documentation

  • README: Added "Hardware-in-the-Loop Simulation" section
  • Release Notes: Updated v15.3.0 notes with simulation features

📦 Upgrade Instructions

# 1. Pull latest
git pull origin main

# 2. Install dependencies
cd qres-studio && npm install

# 3. (Optional) Fetch weather data
python3 scripts/fetch_weather_replay.py

# 4. Launch dashboard
npm run dev

📊 Metrics

Metric v15.3.0 v15.4.0
Startup Time ~1.6s ~1.5s
Bundle Size 1.4MB 1.5MB
Visualization FPS 30 60

Full Changelog: v15.3.0...v15.4.0

Full Changelog: v15.3.0...v15.4.0

Loading

QRES v15.3.0 - Edge Monitor

11 Jan 06:19
@github-actions github-actions
v15.3.0
553dff2

Choose a tag to compare

View all tags
QRES v15.3.0 - Edge Monitor

QRES Edge Monitor v15.3.0

Release Date: January 11, 2026
Codename: Edge Visualization

Overview

v15.3.0 transforms QRES Studio into a Real-Time IoT Edge Dashboard, replacing the file compression interface with live sensor stream visualization.

New Features

📡 Real-time IoT Streaming

  • Live Telemetry: 10Hz sensor data from simulated ESP32, Pi-4 Cluster, and Jetson-Nano
  • D3.js Bandwidth Chart: Scrolling Raw vs. QRES compressed bandwidth visualization
  • Connect to Swarm Toggle: One-click sensor stream activation

🧠 Neural Graph Visualization

  • Interactive Topology: Force-directed graph (Swarm, Mixer, Root, QNN, SNN nodes)
  • Real-time Updates: Graph reflects current MetaBrain activity

⚡ MetaBrain State Monitoring

  • SNN Spike Visualizer: Canvas-based spiking neuron animation
  • Regime Change Detection: Visual feedback for neural adaptation

Improvements

  • Simplified Header: "QRES Edge" branding with clean badge
  • Navigation-Only Sidebar: Icon-based (📡 🕸️ 🧠) replacing file controls
  • Window Size: 1200x800 default for better visibility
  • Flexbox Layout: More reliable viewport sizing

Bug Fixes

  • WebSocket connection stability for long-running streams
  • TypeScript status mapping (IDLE → OFFLINE)
  • A11y form label compliance

Known Issues

Feature Native Mode Browser Mode
Swarm Toggle ✅ Full ⚠️ Simulated
Hive Mind Sync ✅ Full ❌ Disabled

Note: Browser sandboxing prevents P2P socket connections. Use npm run tauri dev for full functionality.

Upgrade

git pull origin main
cd qres-studio && npm install
npm run dev

Full changelog: v15.2.0...v15.3.0

Full Changelog: v15.2.0...v15.3.0

Loading

QRES v15.2.0

09 Jan 07:08
@github-actions github-actions
v15.2.0
5c3dac1

Choose a tag to compare

View all tags
QRES v15.2.0

QRES v15.2.0

Release Date: January 8, 2026

Overview

v15.2.0 focuses on Scientific Reproducibility and Validation, It includes comprehensive benchmarks, scalability analysis, and complete theoretical documentation.

Key Features

📊 Comprehensive Benchmarks (docs/BENCHMARKS.md)

  • Scalability: Validated 100-node swarm performance (~9.4MB protocol state).
  • Throughput: ~15ms update latency (10 nodes).
  • Privacy Cost: Quantified 3.1x runtime overhead for Full Privacy Stack.
  • Compression: ~22:1 ratio on IoT telemetry.

📚 Documentation Overhaul

  • Theory: New THEORY.md detailing privacy composition and Byzantine proofs.
  • Related Work: RELATED_WORK.md with 30+ citations.
  • Clean Structure: Pruned obsolete files, organized docs/archive.

🧪 Reproducibility Code

  • Docker: Production-grade Dockerfile.
  • Scripts: run_all_benchmarks.sh for one-click validation.
  • Examples: examples/swarm_scale.rs for load testing.

What's Next (v16.0)

  • Post-Quantum Security: Dilithium signatures.
  • FPGA Acceleration: Hardware offload for SNN inference.

See CHANGELOG.md for full details.

Full Changelog: v15.0.0...v15.2.0

Loading