sprix-sage-router: SAGE Router – State‑Aware Routing for A2A Agent Networks

• Fact SAGE Router evaluates three routing modes (SELF, COLLABORATE, HANDOFF) in a single auditable objective.[1]AISAGE decides which feasible agent configuration should execute the task, in which mode, and why.
• Fact It preserves completed requirements and estimates remaining work based on artifact portability.[1]AIThe router preserves completed requirements and combines in-flight completion, observed partial quality, current ownership, and artifact portability to estimate how much work each candidate must actually redo.
• Fact SAGE uses requirement‑conditioned trust rather than a single reputation score.[1]AIReliability is tracked per agent and requirement instead of assuming that one reputation score transfers across skills.
• Fact A separate evaluator replays checkpoints to score artifact reuse and recovery latency without invoking SAGE’s switching equation.[1]AIA separate evaluator replays checkpoints and scores artifact reuse, added cost, recovery latency, wasted work, and final quality without calling SAGE's switching equation.
• Fact The algorithm jointly searches calibrated requirement owners and their schedule, treating work assigned to one agent as serial.[1]AISAGE jointly searches calibrated requirement owners and their schedule. Work assigned to one agent is seriali
• Fact The GitHub repository wang2122/sprix-sage-router was created on 18 Aug 2026 and had 4,261 stars on 2 Oct 2026.[1]
• Fact Its description reads: “Sprix AI at 屿智同行 — state-aware SELF/COLLABORATE/HANDOFF routing for A2A agent networks.”[1]
• Fact That is 4,261 stars in 46 days, an average of 93.4 per day.[1]

• Analysis Integrate SAGE Router into existing A2A protocols to enable state‑aware routing decisions, or use its open‑source implementation as a research baseline.AI
• Analysis Automatically detected from public activity data. Not yet reviewed by an analyst.

• Fact SAGE Router evaluates three routing modes (SELF, COLLABORATE, HANDOFF) in a single auditable objective.[1]AISAGE decides which feasible agent configuration should execute the task, in which mode, and why.
• Fact It preserves completed requirements and estimates remaining work based on artifact portability.[1]AIThe router preserves completed requirements and combines in-flight completion, observed partial quality, current ownership, and artifact portability to estimate how much work each candidate must actually redo.
• Fact SAGE uses requirement‑conditioned trust rather than a single reputation score.[1]AIReliability is tracked per agent and requirement instead of assuming that one reputation score transfers across skills.
• Fact A separate evaluator replays checkpoints to score artifact reuse and recovery latency without invoking SAGE’s switching equation.[1]AIA separate evaluator replays checkpoints and scores artifact reuse, added cost, recovery latency, wasted work, and final quality without calling SAGE's switching equation.
• Fact The algorithm jointly searches calibrated requirement owners and their schedule, treating work assigned to one agent as serial.[1]AISAGE jointly searches calibrated requirement owners and their schedule. Work assigned to one agent is seriali

Architecture

Insufficient evidenceNo architecture has been documented for this signal.

Insufficient evidenceDeveloper implications have not been assessed yet.

Insufficient evidenceNo adoption indicators recorded.

Timeline

1. 2 Oct 2026

   First seenFirst seen by the Astrava pipeline[1]

• Fact The repository does not claim invention of underlying techniques such as Noisy‑OR or beam search.[1]AINoisy-OR, beam search, a linear utility, Beta beliefs, online logistic regression, and DAG-induced communication edges are not claimed as inventions.
• Analysis Generated from activity metrics only; verify the linked sources before acting on it.

Insufficient evidenceNo competing approaches documented yet.

Insufficient evidenceNo independent projects or organisations are documented as building on this yet.