Skip to content

Releases: mivertowski/RustKernels

v0.4.0 — Python Bindings & Production-Ready Execution

07 Feb 20:02
@mivertowski mivertowski
v0.4.0
3f72594

Choose a tag to compare

View all tags
v0.4.0 — Python Bindings & Production-Ready Execution Latest
Latest

Highlights

  • Python bindings — pip install rustkernels for native Python access to all 106 batch kernels via PyO3
  • Production-ready execution — all 106 kernels fully implemented with real dispatch (no stubs)
  • Deep RingKernel 0.4.2 integration — bidirectional domain conversion, enterprise re-exports
  • Enterprise modules — security, observability, resilience, runtime, memory management
  • Ecosystem service layer — Axum REST, Tower middleware, Tonic gRPC, Actix actors

Python Bindings (rustkernel-python)

import rustkernels

reg = rustkernels.KernelRegistry()
print(reg.kernel_ids)                # all 106 kernel IDs
result = reg.execute("graph/betweenness_centrality", {
    "num_nodes": 4,
    "edges": [[0,1], [1,2], [2,3], [0,3]],
    "normalized": True,
})
  • KernelRegistry class with discovery (kernel_ids, by_domain, search) and execute()
  • Module-level execute() convenience function with cached default registry
  • Catalog API: list_domains(), total_kernel_count(), enabled_domains()
  • Full exception hierarchy mapping all KernelError variants
  • GIL-releasing async bridge for concurrent kernel execution
  • Feature pass-through: --features full enables all 14 domains

Enterprise Modules

  • Security: JWT/API key auth, RBAC, multi-tenancy, secrets management
  • Observability: Prometheus metrics, OTLP tracing, structured logging, SLO alerting
  • Resilience: circuit breakers, exponential retry, timeout propagation, health probes
  • Runtime: lifecycle state machine, graceful shutdown, production config presets
  • Memory: size-stratified pools, pressure handling, inter-phase reductions

New Kernels

  • Graph: GNNInference, GraphAttention
  • ML: EmbeddingGeneration, SemanticSimilarity, SecureAggregation, DrugInteractionPrediction, ClinicalPathwayConformance, StreamingIsolationForest, AdaptiveThreshold, SHAPValues, FeatureImportance
  • Process Intelligence: DigitalTwin, NextActivityPrediction, EventLogImputation

Workspace

20 crates total (19 Rust + 1 Python), 106 kernels across 14 domains.

Full Changelog: v0.2.0...v0.4.0

Assets 2
Loading

RustKernels v0.2.0

19 Jan 20:12
@mivertowski mivertowski
v0.2.0
a63f55c

Choose a tag to compare

View all tags
RustKernels v0.2.0

RustKernels v0.2.0

Major upgrade to ringkernel 0.3.1 with enterprise features.

🎯 Highlights

🔐 Enterprise Security

  • Authentication & RBAC: Kernel-aware auth with domain-based access policies
  • TLS & Encryption: Secure communication and GPU memory encryption
  • Secrets Management: Integration with secret stores for kernel credentials
  • Multi-tenancy: Per-tenant kernel isolation and resource quotas

📊 Production Observability

  • OTLP Tracing: Distributed tracing for kernel execution with context propagation
  • Structured Logging: Per-domain log configuration with audit logging
  • Prometheus Metrics: Per-kernel performance metrics and GPU utilization
  • Alerting: SLO violation alerts with Slack/PagerDuty integration

🛡️ Resilience Patterns

  • Circuit Breakers: Per-kernel and domain-level failure protection
  • Health Checks: Liveness, readiness, and startup probes
  • Timeout Management: Configurable execution timeouts with deadline propagation
  • Recovery Policies: Checkpoint/restart support with GPU state recovery

⚙️ Runtime Lifecycle

  • Graceful Shutdown: Drain period for in-flight operations
  • Configuration Hot Reload: Update settings without restart
  • Development/Production Modes: Pre-configured runtime builders

🆕 New Kernels

Graph Analytics

  • GNNInference - Message passing neural network inference
  • GraphAttention - Graph Attention Network with multi-head attention

Statistical ML

  • EmbeddingGeneration - Hash-based text embeddings for NLP
  • SemanticSimilarity - Multi-metric semantic similarity search
  • SecureAggregation - Federated learning with differential privacy
  • SHAPValues - Kernel SHAP for feature explanations
  • FeatureImportance - Permutation-based feature importance
  • DrugInteractionPrediction - Multi-drug interaction prediction
  • ClinicalPathwayConformance - Treatment guideline checking
  • StreamingIsolationForest - Online anomaly detection
  • AdaptiveThreshold - Self-adjusting thresholds with drift detection

Process Intelligence

  • DigitalTwin - Monte Carlo process simulation
  • NextActivityPrediction - Markov/N-gram next activity prediction
  • EventLogImputation - Event log quality detection and repair

📦 New Crate

  • rustkernel-ecosystem: Web framework integrations (Actix, Axum, Tower, gRPC)

📈 Stats

  • 106 kernels across 14 domain crates
  • 19 crates published to crates.io

⚠️ Breaking Changes

  • Minimum ringkernel version: 0.3.1
  • New trait methods on GpuKernel: health_check(), shutdown(), refresh_config()
  • Updated BatchKernel with execute_with_context() and execute_with_timeout()
  • New CheckpointableKernel trait for recovery support

📚 Installation

[dependencies]
rustkernels = "0.2.0"

# Or individual crates
rustkernel-core = "0.2.0"
rustkernel-graph = "0.2.0"
rustkernel-ml = "0.2.0"
# ... etc

🔗 Links

Loading