Attest delivery tracker
Status: approved; implementation in progress. Update this file as each slice is implemented and verified. Each completed item needs a CDX or ELOG commit, test evidence, and the relevant attest PR link. A checkmark means the stated acceptance evidence exists, not merely that code was written.
| Slice | Status | Completion evidence |
|---|---|---|
| 0. Current product baseline | In progress | ELOG/SSO browser captures, 8 external Playwright journeys including real UI replay, failed API diagnosis, PDF export and narrow result access; fresh SSO UI and API executions verified, broader customer acceptance still open |
| 1. Product contract and install | In progress | Independent Field Notes Flutter sample: app and dashboard builds, setup test, validation, real API case and Patrol UI journey pass with source overrides; registry-only install remains open |
| 2. Setup and authoring | In progress | API cases, guided response handoff and Integration sequence composition and Flutter UI journeys have versioned YAML, preflight validation and real Field Notes runs; Integration setup/check/cleanup phases and registered product actions have guided authoring and passing/failure execution proof; browser review/save, desktop/390 px captures and revision checks verified; the source-distributed authoring skill is validated; packaged skill distribution and broader customer acceptance remain open |
| 3. Running and replay | In progress | ELOG retained UI replay and failed API case inspection verified; independent sample retained a three-checkpoint, two-journey WebM run with individual result rows, external replay/PDF proof, and a failed/blocked journey run; dependent API sequences now have timed request rows with HTTP evidence and external passing/failed/blocked replay proof; fresh SSO API/UI runs pass with browser replay and report proof; ELOG reference refresh remains |
| 4. k6 and UX | In progress | Measured ELOG, Field Notes and fresh generated k6 runs, guided editor and replay scorecards; first read-only HTTP suite can be generated by attest init --performance-url with an optional runtime Bearer token; desktop/390 px captures, revision-checked save, preflight validation and 42 dashboard widget tests; additional suite/login-flow authoring and broader keyboard/accessibility pass remain |
| 5. Cloud control plane | In progress | Isolated authenticated run: real Supabase Auth and PostgreSQL, Field Notes remote worker, 2/2 passing API checks, retained typed result and artifact, two-tenant denial, same-tenant viewer grant/revoke, read after host restart, and deliberate 0/2 failed result. A fresh hosted entitlement run on the current source passed 2/2, retrieved test detail and exact artifact bytes, and proved creator, viewer and revocation behavior. A signed-in Flutter browser run queued, inspected, downloaded evidence and signed out. Customer-hosted deployment, production sign-in, TLS, hosted UI replay and farm acceptance remain open |
| 6. Testing farm | In progress | Two-worker same-host proof: six real API runs split three execution attempts per worker and retained two artifacts per run. Registered-plan claim filtering has focused tests. The durable publication handoff passed in isolated live processes after deleting the executor's spool. Settlement recovery survived a live injected worker failure before publication was queued. Live hosted sharding completed two independent API suites on two workers with four available artifacts and a combined 3/3 result; a deliberate failing shard produced 2 passed / 1 failed without hiding the passing evidence. Clean isolated worker runtimes repeated 3/3 across separate executor and publisher mount namespaces. Unclaimed-shard expiry blocked a late claim and closed a two-shard parent. Live lease reclaim fenced the old process and published only its replacement's 2/2 result. Interrupted-upload retry cleans attempted objects while the worker remains alive. Hosted run details show shard outcomes and both retained detail archives. Separate-host deployment, broader farm dashboard, event streams, quotas and crash-time orphan cleanup remain open |
| 7. AI assistance | In progress | API-key assisted API and UI draft flow: provider adapters, scoped generation, editable YAML, validator and explicit save; local proof includes 176 server tests (1 skipped), 48 dashboard tests, independent Field Notes API/UI route proof, 28 focused ELOG tests (1 skipped), and desktop/390 px browser captures. Product owner confirmed API keys first, OAuth later on 26 September 2026. Actual provider-key proof and ELOG UI/SSO versioned-case adoption remain open. |
| 8. Customer acceptance | Pending | Clean sample plus ELOG/SSO setup, run, replay and report proof |
The local Wrangler rehearsal proves the UI, API, same-origin mutation guard and live WebSocket through a loopback Cloudflare Worker. It is a development preview, not an online release.
Existing baseline
- CDX Attest PR #150 and ELOG Attest PR #374 are open.
- CDX has the CLI, shared runners, dashboard and reports. ELOG has the ELOG/SSO product configuration and adapter. The code suites passed in the previous readiness pass; these do not yet prove the proposed customer path.
- Registry-only
cdx_attest_cliresolution remains a release gate. - The registry release plan records the ten staged package versions, publication order and clean-consumer checks.
- No workflow files are included in the Attest branch diffs. Workflow design requires separate manual review before any such file is written.
Decisions awaiting the product owner
The four focused questions and proposed defaults are in the plan's decision section. The product plan was approved on 24 September 2026. Keep the questions visible for decisions that affect a future slice; browser baseline and honest result presentation can proceed independently.
| Decision | Answer | Date |
|---|---|---|
| First self-service user | Awaiting answer | — |
| Meaning of isolated “sub-OS” runtime | Awaiting answer | — |
| Named set of 13 design principles | Awaiting answer | — |
| Local/self-hosted or hosted first milestone | Awaiting answer; cloud-ready product direction confirmed | — |
Change log
| Date | Repo / commit | Evidence | Result |
|---|---|---|---|
| 2026-09-24 | CDX / planning commit | Plan and tracker review | Approved; implementation started |
| 2026-09-24 | CDX 56c7667d1, 784e5c8e5 | Three external Playwright journeys against ELOG-backed Attest | Passed; baseline captures retained |
| 2026-09-24 | CDX 6aa3f6df6 | 38 Flutter widget tests, 3 external Playwright checks, rebuilt web screenshots | Preparation preview shipped to attest |
| 2026-09-24 | CDX ebf5e81a6 | 4 external Playwright checks including ELOG and SSO run setup | Multi-product baseline passed without starting infrastructure |
| 2026-09-24 | CDX 69c91aa01 | 5 external Playwright checks including retained UI recording | Real player and activity visible |
| 2026-09-24 | CDX 9b8c91efd | 38 dashboard widget tests and 6 external Playwright journeys against rebuilt ELOG-backed Attest | Replay action appears only for an execution with a retained recording |
| 2026-09-24 | CDX ddbd56d55 | Failed-run Full evidence browser capture | Navigation works; case-level API failure detail is missing from the panel |
| 2026-09-24 | CDX 23f0ff006 | 39 dashboard widget tests and real failed API case browser capture | Every retained suite run contributes to failure drilldown |
| 2026-09-24 | CDX 6f1d1c3ec | Decoded query screenshot and 39 dashboard widget tests | HTTP request evidence is easier to inspect |
| 2026-09-24 | CDX 3f68e812c | 4 HTTP evidence contract tests | New captures redact query credentials and URI user information |
| 2026-09-24 | CDX 00b2750b2 | 4 panel widget tests and real 390 px browser click | Overlay panel controls work while the backdrop remains dismissible |
| 2026-09-24 | CDX a151948a9 | 4 button widget tests and browser accessibility locator | Icon actions have individual accessible names |
| 2026-09-24 | CDX 61d1287a3 | 8 external Playwright journeys, PDF response/header and download checks | Report export and narrow run workspace verified |
| 2026-09-24 | CDX / independent consumer proof | Generated field_notes outside either repo; validated, tested and built with source overrides | Registry-only install blocked by unpublished CLI |
| 2026-09-24 | CDX / independent API run | Attest execution QR-20260924-123401-521 has one passing HTTP case, JSON events and run manifest | Independent API execution works; service-only runtime and Flutter app proof were still open at this point |
| 2026-09-24 | CDX / service-only runtime | Attest execution QR-20260924-124733-671, startup tests, generated scaffold tests, static analysis | API products can run without Supabase; existing Supabase startup tests still pass |
| 2026-09-24 | CDX / controller and report preflight | Controller execution QR-20260924-130302-492, invalid-profile validation, 22 CLI scaffold tests and 4 server startup tests | Service-only dashboard execution passes; incomplete report profiles fail before a run |
| 2026-09-24 | ELOG aecdedaf2 | Caliber adapter analysis, dry run and 12 registry tests | ELOG explicitly requires its managed Supabase runtime after CDX made it optional |
| 2026-09-24 | CDX 37bbc5c17 / independent smoke review | One external Playwright journey, desktop/narrow captures, two-page PDF response | Service-only smoke run remains inspectable and exportable after controller restart; formal qualification remains distinct |
| 2026-09-24 | CDX 1b0a16ddb / checked-in Field Notes sample | Flutter app and dashboard web builds, setup test, validation, real one-case API run | Reproducible non-Caliber Flutter consumer works from source; UI case and registry install remain open |
| 2026-09-24 | CDX f12d1d906 / Field Notes Patrol UI | Attest run 2026-09-24T14-31-28-952782Z-notes-ui-patrol-smoke, two screenshots, two traces, WebM and no evidence issues | Independent Flutter browser journey passes; dashboard replay still requires verification |
| 2026-09-24 | CDX f8d71167d / Field Notes smoke replay | 41 dashboard widget tests, external Playwright replay/export/narrow check, screenshots and two-page PDF | Retained independent Flutter UI run is inspectable and exportable after controller restart |
| 2026-09-24 | CDX 334fcdc95, 6827a95d6 / API case foundation | Shared YAML case schema and runner tests, invalid-case preflight exit 64, independent Field Notes execution QR-20260924-145529-919, external Playwright result/PDF/narrow check | Customer can add an API case file without adding a Dart test; guided authoring remains open |
| 2026-09-24 | CDX / API origin preflight | Invalid configured API origin rejected by attest validate with exit 64; five focused runner tests | Broken API base URLs fail before runtime preparation |
| 2026-09-24 | CDX ba242507a, 51e6a60b7, 6dc805538 / guided API authoring | Ten focused server tests, 41 dashboard widget tests, clean client/server analysis, external Playwright review and desktop/390 px captures; case created in browser, validated, then passed in a real two-case Field Notes API run | Customer can create a versioned case without writing Dart; preview is read-only, saves require a current revision, invalid YAML remains repairable |
| 2026-09-24 | CDX d0a631d70, 64e97b7a1, 01620dd6c, 03cf19b26, c66e172c1 / declarative UI | Contract/server validation, real two-journey Patrol run 2026-09-24T16-16-45-533415Z-notes-ui-patrol-smoke, three screenshot/trace checkpoints, WebM, external replay and PDF check | New Flutter journeys run from YAML through one adapter; shared capture works, but the run summary still has one outer Patrol test row |
| 2026-09-24 | CDX 789e3589f / guided UI authoring | 41 client widget tests, client analysis, repeated external Playwright preview, desktop and 390 px captures | Users can create, review, save and repair UI journeys from the dashboard while the controller is idle; preview is read-only |
| 2026-09-24 | CDX b464f3b80 / dashboard-authored sample | Field Notes validation and a real two-journey Patrol run with three checkpoints | The saved guided-editor YAML runs beside the original journey without another Dart test |
| 2026-09-24 | CDX bbe1dad9e / product-owned authoring skill | Skill validator passed; Field Notes attest validate and UI run --dry-run passed | A source checkout can copy a skill that prepares API/UI files and exact run commands; published distribution and independent usage proof remain open |
| 2026-09-24 | CDX dec96a436, b4b2d7d63 / Patrol case results | Server contract and report tests, real Field Notes execution QR-20260924-170715-727 with 2 passed journeys and 3 checkpoints, external Playwright result/replay/PDF check; deliberate failed run 2026-09-24T17-10-32-669860Z-notes-ui-patrol-smoke showed 1 failed and 1 blocked journey | The run summary reports each declarative journey while preserving the shared Patrol recording; the sample YAML was restored after the failure test |
| 2026-09-24 | CDX 760f8dcdd, 01d855b64, f0f85d032; ELOG edc2a4583 / durability and reference API | Append-only Patrol case event test; CLI publication dry run has only the expected private-registry key warning; ELOG CLI tests and SSO API execution QR-20260924-173245-901 | SSO local bootstrap verified and the API suite passed 33 tests with 1 skip; a stale Attest-owned ELOG Supabase project initially occupied port 35421 and was stopped before retry |
| 2026-09-24 | CDX 2f3f28de3 / CLI product identity | CLI and server tests; SSO API execution QR-20260924-175531-523; external Playwright replay check | Standalone CLI executions now retain the selected target; shared underlying suites appear under the right product. The first rerun exposed a missing API checkpoint evidence link. |
| 2026-09-24 | CDX 6f76a0062 / API qualification evidence | Builder and validator tests; SSO API execution QR-20260924-180417-738 with 33 passed and 1 skipped; external Playwright browser capture | Test event log is packaged, checksummed and linked to the checkpoint. All mandatory gates passed; SSO Test replays shows the qualified API scope and explains that no video is available. |
| 2026-09-24 | ELOG edc2a4583 / SSO UI reference | Real execution QR-20260924-173604-700; 6 plan tests, 13 checkpoints and 26 evidence artifacts; external Playwright synchronized replay and PDF checks | UI qualification passed with retained WebM and browser replay. |
| 2026-09-24 | CDX 59cbbed2c through 598ff28b1; ELOG 813a9841d / registry preparation | Authenticated clean probes, six package publish dry runs, CDX CLI tests (22), Field Notes resolution with source overrides, ELOG package resolution and 17 focused tests | Missing CLI and replay player packages identified; new package versions and consumer minimums are staged on attest. attest doctor now names the missing package. Publication and clean registry-only acceptance remain open. |
| 2026-09-25 | CDX fc366557a, e394bf336, bf48adda7 / measured k6 result | 42 dashboard widget tests; official k6 v2.3.0 local binary verified by release SHA-256; real ELOG execution QR-20260924-183417-665; external desktop and 390 px Playwright scorecard checks | 209 HTTP requests and all 7 thresholds passed; the configured performance scope qualified and retained its k6 summary. The dashboard presents observed values, threshold verdicts, and honest unit gaps. |
| 2026-09-25 | CDX 98ef5d991, 4f2c2350f, 669c075b5, bdc016ef2 / k6 workload authoring | 4 focused server tests, clean client/server analysis, dashboard widget test, ELOG web build, live options API GET/preview, external browser desktop/390 px/preview/save captures | Existing k6 suites now expose versioned workload editing. The guided path covers one constant-VU scenario; advanced JSON keeps other executor settings. A second browser save verified unchanged guided values leave the ELOG repository clean. |
| 2026-09-25 | CDX 74fa2f429, 6dc57a8e2, cc37530b7, e99b5769d, cf6f6aa00; ELOG 27bd7ba18 / independent performance proof | ELOG read and mixed profile tests, 3 CLI preflight tests, 42 dashboard widget tests, 5 k6 editor tests, real Field Notes execution QR-20260925-063628-130, external Playwright editor and result captures at desktop/390 px | Edited k6 options drive result metadata and timeout. The non-Caliber sample completed 100 requests and 4/4 thresholds. Its qualification remains blocked by its unapproved environment, now named in the run result. attest validate rejects missing scripts and invalid thresholds before runtime startup; ramping preview checks stages and summarizes their traffic. |
| 2026-09-25 | CDX 1c411471e, 6f6fb1bb7 / generated first performance suite | 24 CLI scaffold tests, clean CLI analysis, fresh /tmp workspace generated from one URL, source overrides resolved, web dashboard built, real k6 execution QR-20260925-065741-806, external Playwright editor and result captures | The first unauthenticated HTTP smoke suite can be initialized without hand-writing registry or k6 files. It completed 100 requests and 4/4 thresholds; formal qualification stayed blocked by unapproved environment and missing Git identity in the temporary checkout. |
| 2026-09-25 | CDX c05920f6b / protected generated performance smoke | 25 scaffold tests including real k6 v2.3.0 against a local Bearer-protected endpoint; clean CLI analysis; generated files, k6 summary and process output scanned for the fixture token | A customer can supply a token at runtime without editing or storing a credential in the generated workspace. The protected 10-second smoke passed and retained outputs did not contain the token. |
| 2026-09-25 | CDX 8174bda7a, d7787b6b5 / dependent integration sequence | 10 focused API/sequence tests, clean server analysis, Field Notes attest validate and real action execution QR-20260925-071400-215 with a session followed by a protected read; retained traces checked for a Bearer value | Versioned API cases can pass an extracted JSON string to later requests. Sequence references are preflighted, execution stops on failure, and runtime/extracted values are masked from HTTP evidence even under ordinary field names. Guided composition remains open. |
| 2026-09-25 | CDX fa3a93eb1, 7a66e1271, 6048bd566, 0238cb38c / guided Integration authoring | 45 dashboard widget tests, focused server tests, clean client/server analysis, rebuilt Field Notes web client, external desktop/390 px composer and API handoff captures; real controller preview/save returned HTTP 200; unchanged sequence YAML remained unchanged after browser save; attest validate and execution QR-20260925-075450-079 passed | Users can order existing cases, review and save a versioned sequence in the dashboard. The API case editor can author response extraction and runtime header/query use for an Integration suite. Reused runtime names that would fail execution are rejected during review. |
| 2026-09-25 | CDX 735194259, c522604b4 / per-request Integration results | 23 focused runner/report tests, clean server analysis, Field Notes attest validate, real execution QR-20260925-080152-646 with 2/2 requests passed; deliberate failure execution QR-20260925-080238-741 showed 1 passed, 1 failed and 1 blocked; passing and failed/blocked external browser captures after controller restart; retained traces checked for a Bearer value | Each sequence request is a timed result row with its matching redacted HTTP exchange. Failed and unrun steps remain distinct. The deliberate failure edit was restored after the test; the duplicated wrapper diagnosis found in browser review was removed. These are smoke results, not formal qualification. |
| 2026-09-25 | CDX e53cea969, a2aa62f5d, af12121cf, 39856ae85, 44e21aea5 / Integration lifecycle | 27 focused runner/report tests, 46 dashboard widget tests, clean server/client analysis, Field Notes attest validate, real passing execution QR-20260925-081652-290 (3/3) and deliberate failed execution QR-20260925-081713-153 (setup passed, check failed, cleanup passed); retained traces scanned for raw Bearer values; desktop, narrow, passed and failed external browser captures; browser preview/save HTTP 200 with unchanged YAML bytes | Optional setup and cleanup phases work with existing case YAML. Cleanup runs after setup or check failure, phase outcomes appear in replay, and the guided editor lets users order all three phases. The deliberate failure edit was restored. |
| 2026-09-25 | CDX f2124ebb6, 7e6bc6a45, 58531d31e / registered product actions | 31 focused sequence/report tests, 46 dashboard widget tests, clean server/client/sample analysis, real Field Notes execution QR-20260925-083421-262 (4/4), deliberate action failure QR-20260925-083843-721 (action failed, setup/check blocked, cleanup attempted and failed for missing credential), undeclared-action attest validate exit 64, retained run files scanned for raw Bearer values; desktop, narrow, passed and failed external browser captures; browser preview/save HTTP 200 with unchanged YAML bytes | Product-owned catalog entries and Dart callbacks can be mixed with API cases in all sequence phases. Action failures retain safe codes instead of exception text; in-memory output can feed later requests. The deliberate failure edits were restored. These are smoke results, not formal qualification. |
| 2026-09-25 | CDX b82251931, ed05e217e / durable cloud reads | Contracts payload round-trip and legacy default, 5 focused server tests, 3 Supabase composition tests, clean analysis across contract/server/runtime packages; a new control-plane instance reconstructed remote selection and suite names; a second tenant could not queue with another identity or read that tenant's run, list or manifest | The local cloud-control boundary survives controller restart and enforces tenant identity on remote reads. Hosted deployment, project membership, multi-worker behavior and sharding remain unproven. |
| 2026-09-25 | CDX de7a44c8e / project membership | 6 focused server tests, 5 storage schema tests, clean server/storage analysis; in-memory control-plane test denies an unlisted user, permits viewer reads but not queue, permits operator queue; disposable PostgreSQL 17 executed the generated project/member/trigger SQL, auto-granted creator owner, and preserved tenant/creator after a conflicting insert | Service-role reads now require tenant and project membership. New project ownership is created at insert time. Hosted identity authentication, membership management and deployed acceptance remain open. |
| 2026-09-25 | CDX 44cefc57b / report authorization proof | Focused remote-control test and analysis passed with a retained report fixture; cross-tenant and unlisted users returned no artifact without an object-store read, while a project viewer retrieved the report bytes | Report retrieval now has a direct regression check for the membership boundary. Hosted evidence retrieval remains unproven. |
| 2026-09-25 | CDX f5e8addec, ELOG 1b37b177c / member management and consumer compatibility | Focused CDX server, storage and runtime tests passed; ELOG's source-linked controller tests passed; disposable PostgreSQL 17 executed the generated role-change function and verified administrator limits, owner transfer, revocation and final-owner protection. The first database probe caught a current_role name collision, which was corrected before the successful probe. | The durable control plane supports member changes with atomic database rules. ELOG now calls the identity-scoped API. The loopback dashboard remains single-identity; a real authenticated hosted entry point is still required. |
| 2026-09-25 | CDX 7219db13c, 80fa9ce04 / hosted HTTP and Auth boundary | Focused server/runtime analysis and tests passed. Live HTTP tests proved missing/invalid token rejection, server-owned queue identity despite spoofed JSON, project mismatch and run-policy denial; a copied Field Notes registry accepted a declared remote API run and rejected changed URL/suite/digest; fake Supabase Auth accepted app metadata and ignored user metadata. | A stateless HTTP transport and pluggable verified-identity boundary are available in source. Deployment, real sign-in, actual worker execution, TLS and two-tenant hosted evidence checks remain open. |
| 2026-09-25 | CDX 5d6350d1f, 7c1aa042b / executable cloud host | Clean runtime analysis and focused Auth tests; attest_cloud_host --help ran; the command started with the Field Notes registry and a placeholder Supabase URL, returned live HTTP 401 without a token and 503 on an Auth outage, then was stopped | A deployment can start the authenticated API without custom composition code. The smoke used no real Supabase database or worker, so hosted execution and two-tenant proof remain open. |
| 2026-09-25 | CDX 46c037dbb / cloud host shutdown | Clean binary analysis; manual source-host launch with Field Notes registry followed by SIGINT exited with code 0 | Container stop can close the HTTP listener and Supabase client cleanly. A full hosted run remains open. |
| 2026-09-25 | CDX eb638cc7e, 0d74fa370, ELOG 2284388b5 / remote worker composition | CLI scaffold suite passed, Field Notes attest validate reported two bindings, Field Notes setup tests and worker analysis passed, ELOG worker analysis passed | A customer sample now shares a manifest adapter between local and remote modes; ELOG's worker closes on SIGTERM. Neither worker has yet completed a run against a deployed Attest store. |
| 2026-09-25 | CDX ea9a35e35, 514b97823, 241bb5f89, cda5b24d3, 8ff10e06f / isolated authenticated cloud run | Dedicated Supabase Auth/PostgREST/Storage/PostgreSQL stack; real Field Notes API and worker; successful run field-notes-cloud-1790331167558 with one suite, 2/2 checks, one available execution artifact and receipt; tenant B denied known run and artifact; viewer grant/read/queue-denial/revoke verified; retained counts after host restart; deliberate API outage run field-notes-cloud-1790331247232 showed 0/2 failed with distinct artifact | Local end-to-end cloud-control path is proven. Three real-run defects were fixed. Customer-hosted deployment, TLS, UI replay, farm behavior and publication remain open. |
| 2026-09-25 | CDX b7d615aa7 / retained hosted test details | Focused hosted HTTP and publisher tests passed; server/contracts analysis clean; real Field Notes execution QR-20260925-102422-328 archived two passing test rows with HTTP evidence. Broader startup test file timed out under flutter test while launching fake services. | Worker output now includes a typed per-suite detail archive and the authenticated host can serve it after publication. The Flutter dashboard still needs the hosted connection and replay presentation. |
| 2026-09-25 | CDX b13c95ce9, ELOG 8d52278db / exact hosted artifacts | Focused CDX HTTP/control-plane and ELOG controller tests passed; analysis clean. Tests covered private inventory, two recordings by exact ID, hidden storage keys, cross-tenant and unlisted denial, viewer access and immediate revocation. | Hosted clients can identify and request a specific retained artifact. The ELOG source-linked control-plane stub remains compatible. |
| 2026-09-25 | CDX be17818a1 / artifact archive paths | Contract, publisher, HTTP and storage-schema tests passed; analysis clean. Legacy artifact JSON defaults to an empty path and new publications retain normalized relative paths. The generated ALTER TABLE ran twice against disposable PostgreSQL 17; an existing row remained readable and a new row stored its path. | Hosted replay has a stable path for matching an artifact to its source evidence. The additive column migration is proven locally; a full customer database upgrade remains open. |
| 2026-09-25 | CDX 845e53276, 69012c94e, 3bc046bce / hosted options and Flutter workspace | Trusted run-options and selection queue HTTP/policy tests passed; hosted client tests and desktop/390 px widget tests passed; client analysis and Flutter web build succeeded. External Playwright opened the compiled fixture app at 1440 and 390 pixels, expanded a failed test, saw no page errors or horizontal overflow, and captured desktop and narrow views. | A signed-in embedding app can supply a token provider and use a server-owned catalog to queue runs, inspect test details, and preview/download retained artifacts. Captures use a fixture API; membership-filtered discovery, live browser sign-in, customer TLS deployment and hosted UI replay still need acceptance. |
| 2026-09-25 | CDX e20ac2231, ELOG f8f5b2790 / hosted discovery and creation gate | Focused hosted HTTP, durable control-plane, Auth verifier and ELOG source-linked tests passed; server and runtime analysis clean. An authenticated nonmember sees no registered project, an unentitled user cannot create one or write its project row, a trusted app_metadata claim allows first-run creation, and viewer grant/revocation changes discovery immediately. | The hosted catalog no longer exposes another member's project or suite definitions. Existing members remain usable. The prior isolated cloud run predates the creation claim, so a fresh live Auth/worker/browser acceptance run is needed on this revision. |
| 2026-09-25 | Current CDX source after e20ac2231 / hosted entitlement acceptance | Disposable Supabase proof: real Auth users, entitled first queue, unentitled HTTP 403, registered Field Notes worker 2/2 pass, one retained detail archive, two available artifacts, exact artifact bytes, viewer grant/read/queue-denial/revoke, and PostgreSQL run/artifact/receipt rows. | The current source Auth, API and worker path is proven locally. Real signed-in Flutter browser use and deployed TLS remain open. |
| 2026-09-25 | CDX 2ee8324ba, e2486de6a / signed-in hosted browser | External Playwright proof: built Flutter web app signed in through real Supabase Auth, queued Field Notes from the UI, observed 2/2 pass, rendered test and artifact rows, downloaded 1,276-byte execution.json on 390 px, signed out, and showed no horizontal overflow or page errors. Client widget tests, sample analyzer and web build passed. | Generic Attest token callback works with a real browser Auth session and a same-origin API proxy. Customer TLS, production identity integration, hosted UI replay and farm behavior remain open. |
| 2026-09-25 | CDX 7eec45eb3 / registered-plan worker claims | Focused durable worker and registry policy tests passed; server and sample worker analysis clean. An incompatible worker left a queued elog.qualification job untouched; a matching worker executed and published it. Live two-worker proof completed six Field Notes runs, three per worker, with 2/2 checks and two available artifacts each. | Same-host multi-worker claims and registry-derived claim filtering work. The local spool publication handoff remains unsafe across separate worker machines; sharding and recovery remain open. |
| 2026-09-25 | CDX 301f6942d / durable publication handoff | Publisher isolation proof: staged four artifacts and an integrity-checked manifest, deleted the source spool, and published through another instance. A subsequent live Supabase/Field Notes run executed and published in separate processes after spool deletion: 2/2 passing checks, two available artifacts, and no local path in publication work. All 154 server tests passed with one opt-in k6 skip; changed files analyze clean. | The publication seam no longer requires a shared worker filesystem for new work. Separate-host deployment, partial-upload cleanup and crash recovery remain open. |
| 2026-09-25 | CDX 9f64c830b, 9437ec117, 67eed0b37 / settled-work recovery | Live injected-crash proof: execution settled in PostgreSQL, orchestration commit failed, spool was deleted, a new worker recovered the decision, and a third published 2/2 with two artifacts. Unit tests cover execution, publication and terminal-failure settlement gaps without duplicate provider or publisher calls. The final code passed a skewed-clock recovery test and its four-argument PostgreSQL RPC compiled in a fresh stack. | A restart no longer strands these settled Attest jobs. Concurrent recovery, separate-host deployment, lease-loss recovery and unreferenced upload cleanup remain open. |
| 2026-09-25 | CDX 193c60ac4 / bounded shard model | Shard contract and planner: deterministic capacity-bounded partitions preserve explicit dependency groups. Aggregate blocks a partial pass when a shard, expected suite or durable evidence is missing; tests cover failure and wire round trips. Contracts 18/18 and server 162 tests passed with one optional k6 skip; focused analysis clean. | The hosted path still dispatches one whole-selection work item. Persisted shard identity, parallel dispatch, evidence association and live aggregate acceptance remain open. |
| 2026-09-25 | CDX 2e984b5b0 / immutable shard persistence | Disposable PostgreSQL and adapter proof: two planned shards and two reports survived baseline reapplication; assignment/report rewrites were rejected; repeated identical writes passed; a changed report was rejected. The first live adapter run exposed reversed row order, which was fixed before acceptance. Server 162 tests and storage 6 tests passed with one optional k6 skip; analysis clean. | Shard records are durable, but hosted queueing and aggregate status are not yet wired to them. |
| 2026-09-25 | CDX e695b6a39 / trusted shard dispatch | Dispatcher contract: two selected suites queued as separate leased work and publication items; the parent stayed open through the first three settlements and closed only after both reports. A failing shard closed the parent failed. Server 165 tests passed with one optional k6 skip; changed files analyzed clean. | The hosted queue does not yet derive independent groups from trusted registry metadata; no live sharded worker run or shard dashboard proof yet. |
| 2026-09-25 | CDX 327764a73, d8e8f4420 / registry-derived shard plans | Optional explicit suite groups round-trip through the catalog; the registry source derives and validates a bounded two-shard Field Notes read plan and passes it to hosted queueing. Contracts 18/18 and server 165 tests passed with one optional k6 skip; changed files analyzed clean. | At this point Field Notes remote approved only API, so its API/load pair was not a live hosted case. Shard dashboard and live two-worker failure acceptance remained open. |
| 2026-09-25 | CDX 8720389c9, 1c05fa093, d70e4a710 / live hosted shards and read model | Disposable Auth/PostgreSQL/Storage proof: two API suites on two workers, each first-attempt success after fixing an execution-ID collision, two exact passing shard reports, four available artifacts and combined hosted 3/3. Server 166 tests passed with one optional k6 skip. | Visual shard UI, a deliberately failed or missing shard, and separate-host acceptance remain open. |
| 2026-09-25 | CDX fdfc4a9b7 / failed shard read model | Deliberate live failure: one API shard passed and the other failed its 401 expectation; the parent failed, both workers published evidence, and hosted read showed both suites, exact shard outcomes and 2 passed / 1 failed. Full server suite 166 tests passed with one optional k6 skip. | Missing-shard, separate-host and visual shard dashboard acceptance remain open. |
| 2026-09-25 | CDX f756276ea / single-work outcome consistency | Hosted failed API proof: one deliberately failing private-access suite published evidence; PostgreSQL and HTTP both showed failed with 0 passed / 1 failed and the retained suite. Server 167 tests passed with one optional k6 skip. | Hosted reporting now agrees with the published test outcome; visual result polish and broader customer acceptance remain open. |
| 2026-09-25 | CDX 89f4158e1, 9f8da1f98 / shard result UI | Authenticated browser captures: the hosted response combined both detail archives; desktop and 390 px Chromium views showed two shard outcomes, receipts and all three test checks without horizontal overflow. Client 51 tests and server 167 tests passed, one optional k6 skip; Impeccable detector clean. | Broader farm dashboard, deliberate missing-shard state and separate-host acceptance remain open. |
| 2026-09-25 | CDX 79fb76dae, e06d17d8f / shard decision trace and terminal failure | A queued history event now references the first real shard work item. A focused two-shard test made one provider fail terminally before evidence; the other published, the failed shard recorded zero-artifact blocked, and both domain and durable parent closed only after the other shard settled. Focused server test file passed 12/12. | An unclaimed shard has no deadline or terminal policy; separate-host proof and broader farm operations remain open. |
| 2026-09-25 | CDX acfa4f3ef / unclaimed shard deadline | Live disposable proof: a worker arriving after the stored deadline claimed no work; the cloud sweep recorded two blocked reports, two typed failures, and a blocked parent. Server 169 tests passed with one optional k6 skip; durable 10, schema 4, runtime 4 passed. | Claimed lease-loss recovery and separate-host farm acceptance remain open. |
| 2026-09-25 | CDX 8929c144e / reclaimed lease fencing | Paused-process PostgreSQL proof: claim generation 2 completed and published 2/2; the displaced process exited on lease loss and its attempt was marked failed. Server 170 tests passed with one optional k6 skip; analysis clean. | Separate-host acceptance, partial-upload recovery and cleanup remain open. |
| 2026-09-25 | CDX 3538282a7 / interrupted-upload cleanup | Injected upload failure and real Supabase delete: partial objects were removed before retry, a previously receipted object was preserved, and two artifacts then published. Server 172 tests passed with one optional k6 skip; Supabase storage 6 passed. | Crash-time orphan sweeping and separate-host acceptance remain open. |
| 2026-09-25 | CDX 5332897cb; ELOG 3a0775652, d42a13556 / registry release staging | Ten-package release order: clean registry probes identified missing CLI/replay packages and absent new versions; all ten archive dry runs assembled, CDX and ELOG source graphs resolved, and ELOG Attest server passed 38 tests with two skips. | Product-owner review and manual registry publication remain required; registry-only customer acceptance is unproven. |
| 2026-09-25 | CDX 22f257a48 / simpler Flutter journey authoring | The guided editor suggests an ID from the journey name, defaults the actor to App user, assigns checkpoint IDs, and describes actions in plain language. A widget test preserves a customer-entered ID and checks the generated YAML; all 47 dashboard widget tests and focused analysis passed. | UI journey authoring is easier, but customer acceptance and the broader keyboard/accessibility pass remain open. |
| 2026-09-25 | CDX a35204f2f, 72aa1a953 / hosted artifact byte ranges | Authenticated artifact responses now handle one HTTP byte range with 206, Content-Range and Accept-Ranges; an invalid range returns 416. Four focused hosted HTTP tests and changed-file analysis passed. The full server run had 170 passes, one optional skip and three startup fixture failures: notes-api readiness or test timeout. | The host still downloads the full object before slicing it; object-store partial reads, hosted browser replay acceptance and the unrelated startup fixture failures remain open. |
| 2026-09-25 | CDX f7d6e2c4e / startup fixture correction | Under flutter test, Platform.resolvedExecutable points to flutter_tester; the startup fixture now resolves the Dart SDK executable before launching its fake services. All four startup tests passed, then the full server suite passed with 173 tests and one optional k6 skip. | The earlier startup fixture failures are resolved. Hosted object-store partial reads and customer replay acceptance remain open. |