Sprix SAGE Router

What does this agent do, and when should you use it?

Sprix SAGE Router is a Python research prototype for reconsidering an A2A task's execution route after work has begun. It evaluates the task DAG, current ownership, completed nodes, partial progress and quality, artifact portability, failures, budget, and deadline before choosing SELF, COLLABORATE, or HANDOFF. Its implementation includes SAGERouter, DAG scheduling and beam search, requirement-conditioned trust, an online success model, audit traces, and exportable learning state. The sprix_a2a.py adapter normalizes Agent Card signals and produces a transport-neutral ExecutionPlan containing ownership, assignments, dependencies, communication edges, resource estimates, and rationale. It does not send tasks, authenticate endpoints, verify signatures, stream responses, poll, cancel, or securely transfer artifacts, so deployment requires a separate A2A client and security layer. The project is an early research preview, and its published measurements come from synthetic evaluations rather than real endpoint evidence.

A caller describes candidate workers with Agent, defines a dependency-aware workload with Requirement and Task, and supplies live ownership, assignments, completed requirements, in-flight progress and quality, and artifact transferability through ExecutionState. SAGERouter.route_with_trace() applies permission and compatibility filters, performs a bounded search over owners, requirement assignments, and DAG schedules, rechecks team cost and critical-path latency, and compares SELF, COLLABORATE, and HANDOFF under one utility function. The returned trace exposes the selected Decision, assignments, topology, excluded agents, and auditable alternatives. After execution, record_outcome() accepts an ExecutionOutcome with success, requirement scores, actual cost, and actual latency to update requirement-conditioned trust, the success model, and bid fidelity; export_state() produces a learning-state snapshot. For A2A use, sprix_a2a.py validates declared skill IDs and converts the route into a transport-neutral plan, but it does not contact or operate remote endpoints.

1. An engineering team building an A2A orchestrator needs to decide at a checkpoint whether to preserve the incumbent's context or switch to a specialist.
2. A multi-agent systems researcher wants to compare SELF, COLLABORATE, and HANDOFF under identical permission, budget, and deadline constraints.
3. A marketplace team with heterogeneous specialists needs reliability estimates for each agent-requirement pair instead of one global reputation score.
4. A workflow architect designing failure recovery needs reassignment decisions that account for partial completion, reusable artifacts, and context-transfer loss.
5. A routing-algorithm team needs deterministic synthetic trajectory replay, controlled progress interventions, trust-convergence studies, and candidate-scaling measurements.

How do you install or deploy this agent?

Python 3.10+ and Git are required. The reference implementation has no runtime dependencies and documents no required API keys or other credentials. Clone the repository over the network and run the demo:

git clone https://github.com/wang2122/sprix-sage-router.git
cd sprix-sage-router
python demo.py

Run the supplied verification and benchmark commands with:

python -m unittest -v
python benchmark.py
python benchmark_dynamic.py
python benchmark_trust.py

How do you use this agent?

This example defines two candidate agents and a dependency-aware task, routes from an in-flight checkpoint, reports the decision, records the execution result, and exports learned state:

from sprix_sage import (
    Agent,
    ExecutionOutcome,
    ExecutionState,
    Requirement,
    SAGERouter,
    Task,
)

agents = [
    Agent("planner", {"planning": 0.92, "coding": 0.55}, cost=0.08, latency_ms=900),
    Agent("coder", {"planning": 0.35, "coding": 0.96}, cost=0.12, latency_ms=1200),
]

task = Task(
    "build-feature",
    requirements=(
        Requirement("planning", 0.4),
        Requirement("coding", 0.6, depends_on=("planning",)),
    ),
    value=1.0,
    budget=0.30,
    deadline_ms=4000,
    progress=0.35,
)

router = SAGERouter(agents, incumbent_id="planner")
state = ExecutionState(
    active_agents=("planner",),
    active_assignments={"planning": "planner", "coding": "planner"},
    completed_requirements=frozenset({"planning"}),
    inflight_requirement="coding",
    inflight_progress=0.35,
    inflight_quality=0.72,
    artifact_transferability={"planning": 0.95, "coding": 0.40},
)
trace = router.route_with_trace(task, state=state)
decision = trace.selected
print(decision.mode, decision.assignments, decision.topology)
print(trace.excluded_agents)

router.record_outcome(
    decision,
    ExecutionOutcome(
        success=0.9,
        requirement_scores={"planning": 0.95, "coding": 0.86},
        actual_cost=0.19,
        actual_latency_ms=1450,
    ),
)

snapshot = router.export_state()

For an A2A deployment, the caller must separately implement message/send, streaming, polling, cancellation, authentication, signature verification, and secure artifact handling.

What are this agent's strengths and limitations?

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

How does FollowAgents rate this agent?

Recent releases

Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.

1. v0.3.0 Sprix SAGE Router v0.3.0 2026-08-28
   Release notes excerpt

   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 …

   View on GitHub ↗

FAQ