v1.0.0 — A/B Test Research Designer

Demo Links

Hosted demo: <placeholder — fill after fly deploy>
Docker image: <placeholder — after docker push>

Executive Summary

v1.0.0 is the first stable release of AB Test Research Designer. It brings the full BCG Phase 1..5 wave into a single supported baseline: teams can define experiment inputs, review deterministic statistical guidance, save project history locally, and export decision-ready reports from one workspace.

This release also turns the product from an internal prototype into a releasable local tool. The frontend was reworked into a more navigable dashboard, project workflows gained templates, filters, revisions, and audit visibility, and the statistical surface now covers SRM checks, Bayesian sizing, sequential monitoring, CUPED, and guardrail planning.

Operationally, v1.0.0 adds stronger release hygiene around accessibility and quality gates. Automated axe coverage now spans the major interactive surfaces, Lighthouse CI is wired into the shipped frontend flow, and the verify pipeline remains the required gate before any future release or push.

Capability Matrix

Feature	Status	Notes
Experiment design wizard and deterministic sample-size planning	GA	Stable for local experiment planning across the main wizard flow
Statistical decision support: SRM, Bayesian, sequential, CUPED, guardrails	GA	Exposed through the backend API and surfaced in the redesigned results UI
Saved projects, revisions, filters, templates, and audit log	GA	Local SQLite workspace with revision history and operational traceability
Report export: HTML, Markdown, PDF, CSV, XLSX	GA	Supports both immediate shareable exports and stored project report downloads
Accessibility and Lighthouse quality gates	GA	Automated axe and Lighthouse CI coverage are part of the release hardening path
AI advice via the local orchestrator adapter	Beta	Optional capability; deterministic planning works without it
Docker packaging for local deployment	Beta	Supported for local verification, but production hosting still needs external TLS termination

Known Limitations

AI advice depends on a separately running local orchestrator reachable through AB_LLM_BASE_URL; when it is unavailable, the deterministic planning path remains available but advisory output is limited.
Automated axe checks are in place, but manual screen-reader regression passes were not completed for this release.
The provided Docker runtime does not enable HTTPS by default; production deployments still need a reverse proxy or platform-managed TLS in front of the container.
Fly.io demo hosting is prepared as a single-machine SQLite deployment; horizontal scaling needs Postgres or LiteFS and is out of scope for v1.0.0.

Upgrade Path

No migration required - v1.0.0 is the first stable release.

Verification Steps For Publishing

Confirm the hosted demo placeholder now points at the live Fly URL and that /health, /readyz, and /api/v1/diagnostics return 200.
Confirm the Docker image placeholder now points at the pushed registry tag or digest.
Re-run scripts\verify_all.cmd --with-e2e.
Attach the assets listed below.
Replace any remaining placeholders before clicking Publish release.

Verification Commands

python scripts/generate_api_docs.py --check
scripts\verify_all.cmd --with-e2e
scripts\verify_all.cmd --with-docker

Assets To Attach

ab-test-research-designer_1.0.0.tar.gz (repo HEAD tarball plus fly.toml)
docs/RELEASE_NOTES_v1.0.0.md
docs/DEPLOY.md

How to publish

gh release create v1.0.0 --draft --notes-file docs/RELEASE_NOTES_v1.0.0-github-draft.md