THIS COMPUTER FIXTURE
—
Waiting for the analysis event
ONE REQUEST. VISIBLE DECISIONS.
See how the work gets done.
Follow the plan, each task, and the evidence required for acceptance.
Synthetic data throughout. No model, runtime or provider execution.
No measured performance or larger-than-VRAM result.
This bounded preview uses the supplied sample request and illustrative outputs.
MODEL FOOTPRINT FIXTURE
—
Model-weight bytes · not runtime peak memory
RUNTIME PLAN
Awaiting analysis
Candidate decisions appear when the plan event arrives.
Mock execution bridge has not started.
Runtime candidates
Planning feasibility onlyWaiting for the plan event.
From request to accepted outcome
Loading fixture…
IntakeNo events played
Workload input ↓ analyze → compare candidates → decompose
Deterministic
Rules & supplied facts
Exact Reuse
Identical approved input
Local AI
Selected local runtime
Frontier AI
Gated external capability · simulated
Task outputs converge ↓
Final answer ILLUSTRATIVE
Service acceptance: not startedThe answer appears after the merge event. Acceptance remains pending until verification and service acceptance pass.
Semantic non-regression: not measured. Fixture acceptance demonstrates UI behavior only; mock completion is not actual execution or service-quality evidence.
A. Workload-control effect
Fixture completed tasks by route
Actual model calls: 0 · Actual provider calls: 0
No savings or improvement multiplier is inferred.
B. Local-execution effect
Profile and plan fixtures only
Runtime execution: not performed
Peak memory, latency and throughput: not measured
Physical execution and service quality need separate evidence.
Planning evidence & event history
Open fixture planning JSONInspect mock adapter lifecycle → · Open bounded local execution evidence → · Open Inference → Execution evidence →