Cytellect 0.1.0-local.17

The connected image, statistics and graph workspace is included in this Windows release. It preserves stain roles, adopted mask edits, object-level GFP selection and source references through editable figures and reproducibility export.

Source: 021081271803b80a5c92563ec253ce0d8b644081.

The exact ZIP passed fresh and repeat setup with pinned Python and actual Fiji, all 26 installed browser cases (zero failed, skipped or flaky), independent numerical replay and clean launcher shutdown. The manifest, payloads and acceptance-harness hash were independently checked before publication.

Use the same installation destination for updates. Compatible pinned Fiji/Python runtimes are reused; application-only updates do not download another Fiji. Research data stays outside application versions.

Private image analysis runs in the installed API/worker or Docker. Vercel provides the public UI and download. Biological segmentation accuracy and independent researcher usability remain separate evaluations; automated arithmetic and transport checks do not establish them.

Acceptance: https://github.com/genellect/cytellect/actions/runs/37694037264
Exact-source CI: https://github.com/genellect/cytellect/actions/runs/37689312905

Final publication verification: the download metadata was merged through PR #49 as aadc2b47914d54ca83d7c26739ec5b581bf7cc19. All five required main checks passed in https://github.com/genellect/cytellect/actions/runs/37700545704 . Vercel production deployment dpl_FdKU2ctWPWysZUyjRswjQCpB8R6N is READY on that source; actual Edge desktop/mobile checks confirmed local.17 downloads, the public workspace, no overflow/page errors and no unrequested paid requests. The anonymously downloaded archive matches SHA-256 1532f1e0b32eb7f9f457f82b6a9847e85c9a919a178c766b27f835cdcc0df51c.

Docker was built from the same final main, updated while preserving the research volume and authentication, and opened in Edge. Its actual API/Fiji/worker startup and image/statistics/graph modes passed the local smoke check. Docker Web image: sha256:456c66ebaca04e98aab0cca451f0ef39b67f03de793e6a735b2096c74a238cf8. Biological accuracy and independent researcher usability remain separate evaluations.

Cytellect 0.1.0-local.16

Windows package from source dc278d5e0bfefcb058a30167ffc715ad21a299e8.

Fixes persisted Undo/adopted revisions across tabs, explicit failed-input exclusion, descriptive-plot recovery, and stale-comparison guards. The unified workspace is the default entry point. Proposal-service budgets and production-key activation are documented separately.

Acceptance: fresh and repeat installation with pinned Fiji; 26 installed browser cases passed with zero failed, skipped or flaky cases; independent numerical replay and launcher shutdown passed. All 231 payload files and the archive/source/harness identities were independently checked.

SHA-256: 7f3770416a27fa3f9e2ba70edec8d1050c49565182a57e31f5a4b1b422d152a6

Archive size: 13,818,022 bytes.

The subsequent wording improvement in PR #40 is not in this archive; it will ship in a new immutable version. Native Smart App Control compatibility on the owner's PC remains unresolved; the tested Docker option is available. Public-image arithmetic/workflow acceptance does not replace private-study or researcher usability validation.

Acceptance workflow · Operation and limits

Cytellect 0.1.0-local.15

Source: 9c35a730cfcc7b0e811a81bfe36c51398a99358f

This prerelease includes the current image workspace, versioned measurements, common statistical comparisons/associations and editable SVG/PDF output. LLM proposals remain off by default; production paid model calls are not enabled.

The actual ZIP passed fresh and repeated Windows setup with pinned Python 3.14.8 and Fiji, 26 installed browser cases (zero failed, skipped or flaky), independent generated-pixel statistics and bundle replay, and launcher shutdown. The new /workspace cases cover registered BBBC007 original images through correction/output and an explicit independent-unit comparison. Every archive payload was checked before publication.

• SHA-256: 2484d2eaf067ecbc986cdf7251e413c184ea7f50db709d429d5498894df72b44
• Size: 13,775,657 bytes
• Acceptance: https://github.com/genellect/cytellect/actions/runs/37349387343

Compatibility: the intended developer PC's Smart App Control blocks an unsigned SciPy extension. That native-PC limitation remains unresolved; these Windows CI results do not establish acceptance on that PC. Docker full-stack acceptance passed separately. No OS protection was disabled.

Public-image arithmetic and software acceptance do not establish detection accuracy on a private experiment, biological independence, researcher usability or journal acceptance. Private-study and human-use evaluation remain separate gates.

Setup reuses matching managed Fiji rather than downloading another copy for each application update. An unrelated Fiji installation is not silently altered or adopted. Research data and credentials are not included in the archive.

Cytellect 0.1.0-local.11

Windows用の研究開発プレビューです。解析計画から領域の確認、定量、統計、編集可能な図の保存までを、同じブラウザ画面で進めます。解析はPC内で実行します。

この版で追加したこと

• 測りたい量から解析を選び、実際のチャンネル・測定指標へ引き継ぐ計画ガイド。
• 任意の標識を1〜3チャンネル登録し、手動領域・整数ラベル画像・確認した核染色からの核検出で面積と輝度を測定。
• 画像の対応を確認する一括登録、重複を防ぐ再試行、修正した代表視野の領域を保持する全視野解析。
• 独立反復が不明でも作れる視野別の分布図と、実験単位・対応関係を明示する群間比較。
• 採用した測定条件、領域、除外、Methods、SVG/PDF、測定表を対応づけた保存と再実行。

既存のNCL・GFP解析も保持しています。任意の細胞構造の自動検出や、細胞全体の境界推定、LLMによる手法選定は対象外です。

配布物の確認

ソース：c21f9057dd130b87ba55b113d40a0396dcb79e74（PR #13）。ZIP：5,823,606 bytes、SHA-256：

ba5835dd55ed9ad6df9698c6200c715c8084ef6d3990b3764f98c6e6d440819b

Windows受入試験の2回目で、新規・再セットアップ、実Fijiと公開画像の処理を含むブラウザ操作9件（失敗・skip・flaky各0）、独立した数式による測定・統計の照合、保存パッケージの再実行、終了を確認しました。実際のZIPの138保存項目とmanifestも別途検査しました。2種類の受入JSONを添付しています。

初回は起動画面の確認1件が失敗しました（8件成功）。同じソース・同じ試験条件の再実行は通過しましたが、初回失敗の原因は未特定です。この履歴を修正済みとは扱っていません。

利用前に確認すること

• ZIPを展開し、Cytellect Setup.cmdを開きます。初回のみ固定したPython・Fiji等を取得します。
• 別の手元PCでは、Smart App Controlが仮想環境の未署名Pythonランチャーを拒否し、導入できませんでした。CIの成功は各PCの実行制御への適合を保証しません。保護設定を無効化せず、管理者に配布物を確認してください。署名済み配布版は未提供です。
• 研究画像と結果はPC内に保存します。保存期限は最後の明示操作から24時間。終了中の物理削除は次回起動時に行うため、必要な成果物を先に書き出してください。
• 公開画像と数値参照による検証は、ご自身の実験の生物学的妥当性や使いやすさの証明ではありません。非公開実画像による比較と、人による利用評価は未完了です。

任意の練習画像として Cytellect-BBBC013-practice-v1.zip を添付しています。BBBC013のDRAQ/FKHR-EGFP画像6枚と出典を含みます。公開BMPの画素値を保持したTIFFで、装置の元形式ではありません。DRAQをDAPIと呼ばず、3視野を独立した生物学的反復と仮定しないでください。画像提供：Ilya Ravkin／BBBC、CC BY 3.0。

本番サイトからの取得確認（2026-10-04）

PR #14は4件の必須検査を通過し、main ad898312dfef21508b02e33a088838e844778ca3へ反映しました。Vercelの自動本番デプロイ dpl_AmyRBWg3dxam939sLWsEn3emtD5n が同じSHAでREADYとなり、本番サイトを実ブラウザの1440px・390pxで確認しました。

ダウンロードボタンから取得したZIPは上記の5,823,606 bytes・SHA-256と一致しました。公開画像の領域と値の連動、計画メモの保存、更新した公開画像クイックスタートの本文も確認しています。確認経路に横はみ出し、Cytellectのコンソールエラー、研究用APIへの通信はありませんでした。検証コードのセレクター誤りによる途中停止2回は、検証コードを修正して完了した記録と分けて保持しています。

これは公開サイトと配布物の同一性の確認です。ZIP内の解析ソースは引き続き c21f9057dd130b87ba55b113d40a0396dcb79e74 であり、非公開実画像・人による操作評価・このPCの導入未完了という範囲は変わりません。

公開サイトのmain ad898312… も、run 37144236252でPython・Web・Windows・実Fiji/ブラウザの4検査を通過しました。

Cytellect 0.1.0-local.9

Cytellect 0.1.0-local.9 — Windows preview

Windows向け研究用プレビューです。解析と保存はPC内で行い、操作画面はブラウザに統一しています。

今回の更新は、測定値の確認と解析記録、保存期限の処理を改善します。

• 弱い信号を確認する任意の表示を追加しました。背景のばらつきに対する補正輝度の比と欠測理由を表示し、利用者が閾値を設定できます。初期値は未設定で、自動除外は行いません（信号QC 1.0.0）。
• GFP閾値の感度解析で、実際に使用した閾値・選別方法を各行へ正しく記録します。選別の数値計算や統計式は変わりません（統計プロトコル1.2.1）。旧1.2.0の行別選別情報を再利用する場合は再計算してください。
• 保存期限を過ぎた実行中ジョブについても、実行権の更新と結果の確定を拒否し、処理停止後に削除します。PCやアプリを閉じている間の物理削除は、次回起動時に再開します。
• 画素値を維持するBio-FormatsでのCZI変換手順、核の結合・分割の検証、既定レシピの整合確認、CIの秘密情報・依存ライセンス記録を追加しました。

Windowsへの分離インストールと実Fijiによる公開GFP画像の解析を確認済みです。350核の測定後に任意QC閾値を変更しても、測定値・対象選別・全マスク画素が変わらず、原画像を含めない2つのZIPの整合性を確認しました。インストール済みコードで科学計算114検査、図版22検査が成功し、日英・89/183mm・3種類の12図を生成・目視確認しました。両検査群は4ケースが重複するため、件数を合算しません。

ZIPを展開し、Cytellect Setup.cmdを開いて導入します。専用のPython/Fijiを使用し、既存のシステムPythonやFijiを変更しません。研究画像はこのPC内で処理します。

項目	値
ファイル	Cytellect-0.1.0-local.9-windows-x64.zip
サイズ	5,704,533 bytes
SHA-256	bdb202b1b85b91f6d1a67aa6d4da1149b5f87d591fb91189945b718dfe5fd720
解析ソース	d810a0f82ee0e76aa2e264e9477b6f805c2c6157

非公開の実画像による妥当性検証と研究者による利用評価は引き続き必要です。自動核検出の上限は各辺2048 pxかつ合計2,700,000画素です。公開サイトのサンプル表示と、将来のクラウド解析サービスは別の提供段階です。

Cytellect 0.1.0-local.8

Cytellect 0.1.0-local.8

Windows x86-64 preview for the browser workspace, local FastAPI service and pinned Fiji worker. Extract the archive and open Cytellect Setup.cmd. Setup provisions isolated dependencies; existing system Python and Fiji are not changed. Routine use opens the same browser interface intended for later hosted delivery.

This update preserves reviewed scientific state across editing, re-detection, sensitivity analysis and reproducibility exports. It adds explicit nucleoplasmic areas and failed nucleolar states, native region-definition comparisons, source-linked model/sensitivity tables, and reproducible numerical CSV exports. Incomplete OME acquisitions are rejected rather than filled with missing-channel zeros. Failed nucleolar processing keeps valid nuclear measurements but blocks scientific acceptance until corrected or explicitly excluded with a reason.

Source: 8b87f21ae9ce5449b2a327c368477249983c7a9d (PR #4).

Archive: Cytellect-0.1.0-local.8-windows-x64.zip — 5,701,069 bytes.

SHA-256: 5c49c0b64eac78ce6a89a9aa8db5cf529c132044315ed0eb306a3c6e86ac5cc5.

Research-use preview: published images and independent arithmetic checks do not establish suitability for an untested private experiment. NCL-defined nucleolar regions require biological review. Editable 89/183 mm SVG/PDF presets address figure formatting, not journal acceptance. Local work expires 24 hours after explicit use; physical cleanup while the application is off resumes on the next launch. Windows mode trusts the OS user and does not add a worker egress firewall. The public Vercel site remains a cleared-image viewer and download page; private analysis runs locally.

The archive includes redistribution-cleared public demonstration derivatives and their notices. Private research images, raw study inputs, credentials and developer runtimes are excluded. Do not attach unpublished research to GitHub issues.

Verified package

• Exact-source main CI: https://github.com/genellect/cytellect/actions/runs/37050026779 (all four required checks passed).
• Windows installation verified 106 manifest entries, Python 3.12.15, Tk and actual pinned Fiji. Existing private installation root and Python dependency cache were reused; the new Fiji runtime was downloaded. No system Python/Fiji or shortcuts were changed by this acceptance run.
• Installed arithmetic, input, ROI, statistical reference and replay suite: 97 passed, zero skipped.
• Installed figure suite: 22 passed, zero skipped; 12 English/Japanese 89/183 mm figures passed vector/font/source checks and visual review of rendered PDFs.
• Real published GFP browser workflow: 350 nuclei, registration through review/export/reload passed. Original images excluded; all 17 export manifest hashes verified. Desktop/mobile preview and table checks, no external requests or application errors, and clean launcher shutdown passed.
• Public NCL reference and broader browser/native-definition checks are documented in the repository. Private experimental suitability and researcher acceptance remain open.

Cytellect 0.1.0-local.6 — Windows preview

Cytellect Windows preview

IF画像の核・核小体検出、領域修正、GFP・NCL定量、統計、編集可能な図表を、ローカルのブラウザで扱う開発プレビューです。画像と結果は利用するPC内に保存します。公開Webサイトは解析サーバーへの研究画像アップロードを受け付けません。

起動

1. Cytellect-0.1.0-local.6-windows-x64.zip をダウンロードして展開します。
2. Cytellect Setup.cmd を開きます。初回はインターネットから固定版のPython・Fiji・依存環境を専用フォルダーに導入します。
3. 作成されたCytellectショートカットから起動し、ブラウザで操作します。既存のPython・FijiやシステムPATHは変更しません。

Windows x64向けです。保存期限は最後の明示的な操作から24時間です。終了中に期限を迎えたファイルは、次回起動時に削除します。必要な成果物は書き出してください。

Source and integrity

• Source: 81047dc8001d4ecb456eab1aceafec278d9533e8
• Archive: 5,683,514 bytes
• SHA-256: 2cb9fbf64a5ad91931e39a20c8ee2df6d9cc62a118a76f1d938d75f580780668
• The archive contains source/static UI, a per-file integrity manifest and third-party notices. Cleared public demo derivatives and their attribution are included. It contains no private research data, credentials or developer runtime.

Verified scope

• Required Python, Web, Windows lifecycle and actual Fiji/browser CI passed on the exact source commit: run 37039980341.
• Installed Python 3.12.15, Tk and actual pinned Fiji passed the known-pixel installation check (9 nuclei, 18 nucleolar candidates). This is an installation/arithmetic check, not biological accuracy evidence.
• The installed public BBBC013 GFP workflow produced 350 nuclei, kept absent NCL measurements missing, completed review/export and reproduced the source identity in a hash-verified package without original images. Desktop/mobile browser checks found no application errors or external research requests; shutdown released the service and worker.
• Independent measurement/statistical references, real published NCL/GFP comparisons, and 89/183 mm editable SVG/PDF rendering are documented in validation. Installed figure/GUI tests were performed on f1af4aba215c4b0524e49598d976a1f59290a5b8; all 56 non-Web package files are byte-identical in this release. Only the Web QC labels changed afterward.

Remaining limits

This is not validation of an individual experiment or a claim of journal acceptance. Private raw-image evaluation, legacy reconciliation and researcher PoC remain open. Automatic nuclear detection has the documented standard-2g capacity limit; original pixels are not silently resized. NCL-defined nucleolar regions require review because NCL redistribution can change their boundaries.

The local edition trusts the operating-system user and is not a LAN service. Native Windows workers do not have OS-enforced network isolation. Later cloud delivery will reuse the scientific implementation and requires its own privacy, retention and host acceptance.

Cytellect 0.1.0-local.10 — Windows preview

Cytellect Windows preview 0.1.0-local.10

This release corrects exported Methods that could describe a non-DAPI nuclear stain as DAPI. The historical internal role dapi does not establish the acquired dye. Methods template 1.0.0 now describes the nuclear-stain channel and asks researchers to confirm dye identity from acquisition records. The auxiliary detector's Japanese label is also stain-neutral. Earlier exported Methods should be reviewed before manuscript reuse.

Detection parameters, masks, pixel measurements, GFP selection, statistical calculations, figure rendering and pinned dependencies are unchanged. Methods text, its manifest hash and the ZIP/source identity change. No new biological accuracy claim is made.

Source: f8a5a16080d94634fb708de5e11cf054a4618a7d (PR #8).
Archive: Cytellect-0.1.0-local.10-windows-x64.zip, 5,704,801 bytes.
SHA-256: d94e64386500a7d51b11e58db69ef60ecbab469c7a745a13f06e14783c79437e.

Extract the ZIP and open Cytellect Setup.cmd. Setup provisions dedicated pinned Python, scientific dependencies and Fiji; existing Python/Fiji installations are preserved. This preview is unsigned. Routine analysis runs through the local browser workspace. Research files stay on the chosen local disk and expire 24 hours after explicit activity; cleanup while the application is closed resumes at its next launch.

Acceptance: exact-source required CI, package installation, the installed public BBBC013 GFP browser/export workflow, both actual export Methods, unchanged QC before/after measurement values and mask pixels, and clean shutdown. Previous local.9 scientific/figure evidence remains historical evidence for unchanged code, not newly executed local.10 tests.

Private original-image validation and researcher usability acceptance remain open. Vercel provides published samples and this download; a hosted private analysis service is not yet offered.