Skip to content

CHANGELOG.md

Latest commit

 

History

History
313 lines (281 loc) · 16.9 KB

CHANGELOG.md

File metadata and controls

313 lines (281 loc) · 16.9 KB
Raw

Changelog

0.1.16

  • Switched README proof images to absolute HTTPS asset URLs so PyPI long descriptions and other external renderers can show the downstream proof strip instead of depending on repository-relative image paths.
  • Added release-check coverage to keep README image embeds PyPI-renderable for future launch hardening releases.

0.1.15

  • Updated the generated browser launchpad and checked-in public demo to lead with the real downstream AI eval PR #1, including blocked and ready CI links, before keeping the in-repo PR #17 replay as a secondary proof path.
  • Packaged the downstream PR proof strip so falsiflow try and falsiflow demo-package produce the same stronger public demo from source installs and PyPI installs.
  • Added a downstream proof-strip PNG for social preview metadata so shared public demo links show the blocked-to-ready downstream PR story.

0.1.14

  • Added a local LLM eval quickstart for turning local model outputs into Falsiflow evidence rows without sending prompts, outputs, or raw artifacts to a hosted service.
  • Added a shareable downstream PR proof strip that leads the README, launch article, demo PR playbook, and launch plan with the real blocked-to-ready AI eval PR story.
  • Added a copy-paste downstream product-metric smoke fixture that blocks placeholder launch evidence and passes only after source-backed metric rows are provided.
  • Promoted the bundled RAG quality gate acceptance path by supporting falsiflow template-check --strict and covering it in release checks.

0.1.13

  • Updated the generated launch posts, GitHub repo profile handoff, and launch execution copy to lead with the live downstream AI eval PR, blocked CI run, and ready CI run.
  • Added release-check coverage so generated launch-kit posts keep the downstream proof links and claim_check_blocked / claim_check_ready narrative in future releases.

0.1.12

  • Tightened the README/PyPI first screen around the one-sentence value proposition, three-command install path, blocked-vs-ready output, reusable GitHub Action snippet, and live downstream proof links.
  • Reworked the launch article demo section to lead with the clean downstream PR that fails on placeholder AI eval provenance before passing with source-backed evidence.

0.1.11

  • Added a maintained examples/downstream_ai_eval_smoke fixture that can be copied into a clean repository to show a placeholder AI eval claim failing CI before source-backed evidence makes the same workflow pass.
  • Linked the live downstream proof repository and PR that show the same blocked-to-ready AI eval claim gate outside the Falsiflow repository.

0.1.10

  • Moved the secondary 30-second CLI visual below the install/status block so the README first screen gets to copy-paste commands faster.
  • Added PyPI project URLs for the Live PR Story, blocked CI run, ready CI run, and launch article so package readers can jump to proof links without relying on rendered README images.
  • Pinned copy-paste GitHub Action snippets to AzurLiu/falsiflow@v0.1.10 and surfaced the six-file downstream AI eval smoke recipe from the README.

0.1.9

  • Added a shareable Live PR Story reel SVG that shows PR #17 moving from a risky AI/RAG eval claim to claim_check_blocked, source-backed repair, and claim_check_ready.
  • Embedded the reel in the README, launch article, demo PR playbook, and public demo launchpad so the blocked-to-ready story is visible without opening the full playbook first.
  • Added release-check coverage for the reel asset, README embedding, manifest inclusion, and packaged launchpad references.

0.1.8

  • Added description, Open Graph, and Twitter card metadata to the public demo launchpad so shared links preview the live PR story.
  • Enabled automatic GitHub Pages deploys on main pushes that affect the demo generator, packaged public demo, or Pages workflow.
  • Sharpened README/PyPI first-screen copy around the Live PR Story proof path.

0.1.7

  • Put the live AI eval PR story on the public demo launchpad: PR #17 now appears as a first-screen sequence from risky claim to blocked CI to source-backed ready CI.
  • Updated falsiflow try and falsiflow demo-package output so regenerated static demos keep the same real PR proof path instead of leading with only a static starter example.
  • Refreshed the checked-in docs/public_demo package with the new launchpad.

0.1.6

  • Fixed rag-eval CSV imports so manifest and artifact rows are promoted into provenance and reproducibility evidence, allowing the bundled RAG raw eval export to import and pass claim-check without a separate JSON manifest.
  • Added timestamps and explicit candidate/baseline RAG version rows to the RAG quality gate raw export used by examples and quickstart templates.
  • Added regression coverage for the real path: raw RAG eval export -> evidence import -> claim_check_ready.

0.1.5

  • Updated launch execution copy from the older v0.1.2 trust baseline to the live AI/RAG eval PR story, PyPI release path, and external-evidence workflow.
  • Switched the README/PyPI long description to a stable external-evidence workflow link so package metadata does not point at an outdated single run.
  • Refreshed PyPI trusted-publishing verification guidance so future releases check the current release version instead of carrying the original v0.1.2 recovery wording into launch copy.

0.1.4

  • Added a live public AI/RAG eval PR demo: PR #17 now shows placeholder eval evidence failing CI and source-backed evidence passing the same claim gate.
  • Added the bundled rag_quality_gate starter template plus rag-eval, local-llm-eval, and ai-eval artifact import coverage for JSON, JSONL, CSV, and manifest-backed eval outputs.
  • Added local agent integration surfaces: falsiflow/api.py, falsiflow mcp, and MCP documentation for local stdio use by AI coding agents.
  • Improved GitHub Action summaries with top blockers, evidence todo items, and next evidence actions for blocked claim checks.
  • Fixed the reusable GitHub Action so evidence: overrides the project-dir default evidence file in claim-check mode.

0.1.3

  • Sharpened the README first screen around the AI/RAG eval claim-gate story: one-line CI positioning, PyPI quickstart, blocked-vs-ready output, and a copy-paste GitHub Action snippet.
  • Reworked the public demo PR playbook into a complete failing-PR-to-ready-PR story for AI/RAG eval claims with local rehearsal commands, expected CI statuses, report artifacts, and a 30-second recording script.
  • Rewrote the launch article Stop Shipping Unverifiable AI Eval Claims around the industry failure mode first, then introduced Falsiflow as the CI evidence gate.
  • Added launch-execution copy and included it in the source distribution so public posts, channels, and metrics review notes ship with the package.

0.1.2

  • Added the product_metric_launch starter template and public casebook coverage so Falsiflow now demonstrates AI eval, product-metric, vendor, wetware, biointerface, and neural-materials claim gates.
  • Made the reusable GitHub Action install from its checked-out action path by default, keeping downstream claim gates usable before PyPI publication.
  • Updated GitHub Actions dependencies and added copy-paste AI eval action examples for downstream repositories.
  • Added a README 30-second ready-vs-blocked demo strip and release-checked it as a first-screen visual asset.
  • Added named neighboring-tool boundaries for Great Expectations, Evidently, Deepchecks, MLflow, and plain GitHub Actions.
  • Added a public demo PR playbook that rehearses an AI-eval claim moving from claim_check_blocked with placeholder evidence to claim_check_ready with source-backed evidence.
  • Sharpened launch-kit posts, announcement copy, demo script, launch metrics, and maintainer checklist around the demo PR path, PyPI status, and responsible-use reply bank.
  • Extended release-check and regression coverage so launch copy, demo PR references, sdist docs, README assets, and Markdown code fences remain checked before tagging.

0.1.1

  • Introduced the Falsiflow evidence-gated R&D workflow engine.
  • Added configurable project gates, evidence CSV validation, derived metrics, claim audits, next actions, dashboards, and portfolio aggregation.
  • Added starter templates for neural materials, biointerface coatings, RFQ vendor evidence, wetware support hardware, and AI claim evaluation.
  • Added root action.yml so downstream GitHub Actions workflows can run Falsiflow claim, template, casebook, release, adoption, quickstart, and external-readiness gates with one uses: step.
  • Added source-file provenance manifests, portable evidence bundles, bundle zip verification, and release-check quality gates.
  • Added template-check, template-scaffold, template-pack, template-registry, template-lock, and template-install flows for authoring, verifying, distributing, locking, and installing reusable external starter templates.
  • Added versioned template locks and registry source_url support so published template packs can be pinned by URL, byte count, and SHA-256 hash.
  • Added template provenance attestations with optional HMAC signatures so locked template sources can be independently verified before installation.
  • Added template-install --attestation --require-attestation so signed lockfile provenance can be enforced during template adoption.
  • Added template adoption policies with template-policy, verify-template-policy, and template-install --policy.
  • Added template release bundles with template-release, verify-template-release, and template-install --release.
  • Hardened template release verification against unsafe artifact paths, duplicate paths, unmanifested files, and embedded registry/lock mismatches.
  • Added Markdown template release verification reports for human review in CI, release-check, and adoption workflows.
  • Added an adoption-priorities document and README first-run path to keep optimization focused on open-source usability.
  • Added template-gallery JSON and Markdown outputs so cross-domain starter breadth is visible and checked by release workflows.
  • Added casebook-check JSON and Markdown outputs so public starter casebook positive demos, placeholder blockers, source provenance, and verified bundles are checked by adoption and release workflows.
  • Added casebook reviewer replay artifacts (casebook_reviewer_replay.md, .sh, and .ps1) so launch reviewers can rerun every positive demo and placeholder blocked-path proof from generated scripts.
  • Added launch metrics artifacts (launch_metrics.json and launch_metrics.md) so the 1k-star path has review windows for traffic, referrers, stars, demo visits, install signals, repeated questions, and follow-up fixes.
  • Added public package evidence checks to external-evidence, external-check, and the external evidence workflow so final external readiness requires a PyPI project URL and a pipx smoke test from the published package, not only a local checkout.
  • Added public_release_evidence.json and public_release_evidence.md to the publish kit so final public release evidence across repo, demo, PyPI, pipx, Windows, Scorecard, release-check, casebook replay, and launch metrics is reviewed from one generated ledger.
  • Added release_rehearsal.json and release_rehearsal.md to the publish kit so public release reviewers can rehearse the final command sequence, expected artifacts, success signals, and external stop conditions before announcements.
  • Added adapter profiles for generic wide CSV, vendor measurement returns, instrument exports, and plate-reader exports in evidence import and ingest-wide-csv.
  • Added audit_review.json and audit_review.md decision cards for faster human review of audit status, blockers, next actions, and release boundaries.
  • Added claim-check with claim_check.json and claim_check.md so audit, source provenance, bundle creation, and zip verification can run as one user-facing claim gate.
  • Added review artifact indexes to claim-check, bundle verification, release check, and template release verification reports so reviewers can jump between source manifests, bundle integrity, dashboards, and template-release artifacts.
  • Added claim-check --project-dir so initialized starter projects can use default project.json, evidence_pass_demo.csv, and claim_check/ paths.
  • Added quickstart with quickstart_summary.json and quickstart_summary.md so first-run users can create a starter project and verify its claim gate in one command.
  • Added next_commands to quickstart summaries so first-run users are handed directly to falsiflow doctor.
  • Added start as the beginner-friendly local app command with a free localhost port, automatic browser opening, --check, and serve_summary.json.
  • Added scripts/install_local.sh plus make install-local, make start, and make start-check so a local checkout or public Git repository can be pulled into a virtual environment and launched with one command.
  • Added scripts/install_local.ps1 for the same local install/start flow on Windows PowerShell.
  • Added pipx Makefile shortcuts for checkout-based pipx install --force . workflows.
  • Added onboard with onboard_summary.json for beginner next-step guidance.
  • Added static-demo with static_demo_summary.json so a zero-install browser demo can be exported to static hosting.
  • Added demo-package with demo_package_summary.json, .nojekyll, netlify.toml, and publish_checklist.md for a public static-demo handoff.
  • Added publish-kit with publish_handoff.json, publish_handoff.md, publish.env.example, and github_publish_commands.sh for account-bound GitHub/Pages/PyPI release handoff steps.
  • Added external-check with external_readiness.json and external_readiness.md so public repo/demo URLs, pipx, and Windows/PowerShell validation are explicit external gates.
  • Added GitHub Actions workflows for static demo deployment, Linux/macOS/Windows and pipx smoke tests, and PyPI trusted-publishing release builds.
  • Added a localhost workbench.html and API behind falsiflow start so browser users can upload project/evidence/source files, run the local evidence gate, and inspect ready/blocked results, review flow, evidence lineage, repair checklist, and linked review artifacts without touching the terminal.
  • Added discover with evidence_records.json, candidate_queue.json, ranking.md, assay_plan.md, rfq_package.md, and a placeholder project_draft/ so discovery output is structured while remaining non-AI-dependent and non-claim-ready.
  • Added agent discover, candidate rank, assay-plan, and evidence import as namespaced public interfaces for structured discovery and evidence-import workflows.
  • Added discovery schemas for evidence records, candidate recipes, and discovery summaries.
  • Added try with try_summary.json, index.html, and try_report.html so first-run users can open a local browser launchpad, proof report, and generated starter wizard before learning the full CLI.
  • Added serve with serve_summary.json so the local launchpad, try report, and wizard can be opened through a localhost browser demo.
  • Added wizard so browser-first users can draft a claim gate from plain-language presets and export the matching scaffold command, project JSON, and evidence CSV without a server.
  • Added doctor with doctor_summary.json and doctor_summary.md so users can diagnose project health and next repair actions after quickstart.
  • Added repair_checklist to doctor summaries so blocked diagnoses include a concrete command, expected artifact, and success signal.
  • Added adoption-check with adoption_check.json and adoption_check.md so the five open-source adoption priorities can be audited as adoption_ready or adoption_blocked.
  • Added repair_checklist to adoption-check summaries so blocked priorities and final ready rechecks include commands, expected artifacts, and success signals.
  • Hardened release-check distribution hygiene so local build/, dist/, and *.egg-info/ artifacts are ignored and transient build caches are cleaned.
  • Surfaced distribution hygiene in the release_and_distribution adoption priority so adoption reports show build-cache cleanup evidence directly.
  • Embedded adoption priority evidence tables in release_check.md so release reports are standalone for human review.
  • Added release_validation_status to adoption-check summaries so fast --skip-dist adoption readiness cannot be confused with full distribution validation.
  • Added top-level release_validation_status to release-check summaries, reports, and CLI output so fast release smoke tests cannot be confused with complete distribution validation.
  • Required release_validation_ready in the release checklist gate so publishing docs match the adoption-check validation boundary.