42/42
Engine source tests
Core manifest, context, wiring, threshold, bridge, and integration checks passed.
Skip to lab contentInspect JSON proof
FrootAI Lab
/labA transparent evidence-work register for FrootAI claims. Current benchmark and dataset surfaces are illustrative fixtures unless a page explicitly carries verified measurement evidence.
3 Benchmarks3 Datasets
One job
Show what supports a claim — and what does not
Lab is the evidence library for Orchard and FAI Engine. It separates executable source proof, expired or blocked verification, and interface fixtures so they cannot be mistaken for one another.
Operational evidence snapshot
What ran successfully, and what is blocked
97/97
MCP tool unit tests
Tool-layer tests passed; external MCP reachability was not evaluated here.
Blocked
Play 01 direct check
8 declared references remain unresolved.
0 current
Evaluation receipts
95 design-only records are active; 5 evaluation receipts expired.
Interface fixtures
Validate reporting UX with supplied sample data
These cards test charts, tables, downloads, and reproduction instructions. They do not prove comparative cost, carbon, or speed.
Cost Report Fixture
Eight cached examples for validating cost-comparison charts and tables.
8 fixture categoriesInspect fixture →
Regional Intensity Fixture
Twelve supplied regional values for validating carbon-report presentation.
12 fixture regionsInspect fixture →
Pipeline Timing Fixture
Pre-filled stage timings for validating the comparison report.
3 fixture implementationsInspect fixture →
Fixtures, not marketing proof
Current fixture pages exercise reporting and reproduction plumbing with deterministic samples. They are not evidence for cost, carbon, or performance superiority.
What remains unproven
Lab is not yet a hosted workspace where visitors design and run benchmark jobs, and the public site does not issue production Engine verdicts.
Illustrative Evidence Fixtures
Deterministic sample inputs that exercise Lab charts and report plumbing. These are not citation-grade production benchmarks.
bar
Cost Comparison Fixture: AVM-Composed vs Hand-Authored Bicep
Eight cached category examples used to exercise the cost-comparison report. This fixture does not establish average savings across 100 workloads.
Fixture categories: 8Inspect fixture
bar
Regional Grid-Intensity Fixture
Twelve pre-filled regional grid-intensity values used to exercise carbon-report presentation. No 200-composition deployment study is attached.
Fixture regions: 12Inspect fixture
horizontal-bar
Harvest Timing Fixture: Node, Python, and azd
Pre-filled stage timings used to exercise the pipeline comparison report. Raw executions across 50 repositories are not included.
Fixture implementations: 3Inspect fixture
Illustrative Data Fixtures
Small CC0 CSV fixtures with schemas for validating data contracts and UI flows. They are not complete production exports.
Harvested Plays Catalog
Ten illustrative rows showing the intended harvested-play schema. Commit hashes and measurements are fixtures, not production provenance.
Illustrative rows for schema and UI testing. Not a complete production export.
10 rows · Updated
0 downloads · CC0-1.0
AVM Module Taxonomy
Twenty illustrative AVM taxonomy rows used to validate schema and presentation. This is not the full upstream module catalog.
Illustrative rows for schema and UI testing. Not a complete production export.
20 rows · Updated
0 downloads · CC0-1.0
WAF Pillar Results
Twenty illustrative WAF result rows used to validate schema and presentation. They are not current compliance verdicts.
Illustrative rows for schema and UI testing. Not a complete production export.
20 rows · Updated
0 downloads · CC0-1.0
Method
Requirements for publishable evidence
A measured result belongs here only when its inputs, runner, statistics, limitations, raw data, and reproduction path are inspectable.
- 01
Pin inputs
Corpus, repository revisions, regions, policies, tool and model versions.
- 02
Run repeatedly
Use declared sample sizes and report median, p95, variance, and confidence.
- 03
Publish evidence
Keep raw data, chart source, methodology, citations, and runnable scripts.
- 04
Expire honestly
Attach freshness dates; archive or rerun evidence when assumptions change.
What the current fixtures demonstrate
- Cost: paired category-chart and delta-calculation behavior.
- Carbon: regional-value sorting and presentation behavior.
- Timing: stage-duration chart and report behavior.
What they do not prove
- They do not establish measured workload savings, emissions, or speed.
- They do not replace Azure pricing, emissions, or deployment validation.
- They do not establish product superiority.