Sprix SAGE Router v0.3.0

Benchmark rigor and scalability

• structurally held-out geometric evaluator with independently specified latent skills
• random and greedy team baselines
• learning, exploration, initialization, and early/late ablations
• bounded top-k candidate prefilter and request-local scoring caches
• workload-sensitive cost and explicit authorized degraded-route flags
• evidence-gated pair synergy updates
• related-work comparisons, module split, ruff/mypy CI, and release automation

The benchmark remains an author-designed synthetic suite, not evidence of real-world superiority. PyPI upload stays gated until the project owner configures Trusted Publishing; the attached wheel and source distribution passed twine check and clean-environment installation.

Sprix SAGE Router v0.1.0

Research preview

The first public research release of Sprix SAGE Router, an open-source algorithmic output of Sprix AI at 屿智同行.

Included

• State-aware SELF, COLLABORATE, and HANDOFF routing
• Permission-first eligibility and progress-aware switching cost
• Submodular capability coverage and greedy team construction
• Trust-calibrated bids and online reliability / pair-synergy learning
• Executable demo, deterministic benchmark, and behavioral tests
• Enterprise-style contribution, security, governance, and citation metadata

This release is an early-stage research preview. It is not a production SLA or a peer-reviewed result.