Repository navigation
Releases: TheTrueSCU/hexaqueue
Releases · TheTrueSCU/hexaqueue
Release list
v0.5.0
v0.4.0
v0.3.0
v0.2.0
Release v0.2.0: Hexaflow distributed cluster adapter and workflow CLI
v0.1.0
CHANGELOG
v0.1.0 (2026-09-07)
Highlights & Features
- Modular Monorepo Architecture: 13 cohesive, decoupled hexagonal architecture packages (
hexaqueue-core,hexaqueue-cli,hexaqueue-server,hexaqueue-worker,hexaqueue-collateral,hexaqueue-scanner,hexaqueue-monitor,hexaqueue-dashboard,hexaqueue-github-runner,hexaqueue-gitlab-runner,hexaqueue-kueue,hexaqueue-workflow, andhexaqueueumbrella). - Core Kernel & State Machine Rollup: Deterministic state progression for leaf jobs (
SUBMITTED->BLOCKED->PENDING->RUNNING->DONE) and aggregated root runs (SUBMITTED,BLOCKED,RUNNING,DONE) with terminal outcomes (COMPLETED,FAILED,CANCELLED,TIMED_OUT,PREEMPTED). - DAG Dependency & Cycle Engine: Kahn's algorithm topological sorting with cycle detection and rich edge trigger rules (
AFTER_OK,AFTER_NOT_OK,AFTER_ANY,AFTER_CORR). - Hierarchical Job Groups & Parameter Sweeps: Arbitrary recursive group nesting with variable inheritance, template interpolation (
{{ var }}), explicit sequences (params), and Cartesian product matrix sweeps (matrix). - Content-Addressable Collateral (CAS): Local filesystem CAS staging at
~/.hexaqueue/collateralwith streaming SHA256 integrity verification, quarantine states, and automatic retention tiers. - Hexagonal Ports & Zero-Cost Developer Loop: Comprehensive abstract port interfaces for queueing, runtime execution, isolated storage, budget accounting, and log streaming with zero-dependency local adapters and
FREE_TIERresource clamping ($0 cloud spend). - Realistic Example Applications: End-to-end batch ETL pipeline (
examples/data-pipeline) and stochastic multi-regime Monte Carlo simulation (examples/monte-carlo) with programmatic Python DAG builders and CLI collateral scripts. - Developer Experience CLI (
hq): Local developer commands for run submission, status monitoring, and log streaming. - 360° Quality Rigor: >90% branch test coverage, Google-style docstrings with
Notes/Architectural Intent:, 1:1 unit test parity enforcement, Hypothesis property-based fuzzing, and hardened multi-tier CI/CD pipelines.