We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
e2e-sweep cleanup: noetl-tools v3.1.0 + noetl-server v3.0.1 Home (Last refreshed + #49 umbrella row + server/tools ecosystem version cells + sessions-log preview), Releases (two timeline rows + Last refreshed), Sessions-Log (new dated entry), Umbrella-Rust-Server-Port (Recent-activity row + Next-concrete-steps item 4). tools v3.1.0 (tools#47): when:true bool + |tojson fallback + UndefinedBehavior::Chainable. server v3.0.1 (server#171): 64MB result-store body limit + pipeline command/spec stash. All 7 e2e sweep playbooks PASS on Rust-only kind. Tracks noetl/ai-meta#49.
2026-06-08: noetl-server v2.64.0 e2e validation — 25/27 PASS Sessions-Log: new entry for step-level set: + ctx shims session. Umbrella-49: new Recent activity row + updated R5 fixture rig status. Home: Last refreshed with e2e validation summary. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
#76 closes — noetl-server v2.62.0 sequential-mode iterator dispatch Sessions-Log: new session entry. Releases: v2.62.0 row. Home: fix #70 drift (was still Active, now Recently closed); add #76 to Recently closed; update server version to v2.62.0; tools to v2.24.1. Umbrella-Rust-Server-Port: Recent activity row for #76.
docs(wiki): #49 — `{{ execution_id }}` multi-statement pg finding resolved on v2.57.1 Re-ran the open finding from the #54 sweep against the fresh Rust-only kind stack (noetl-server-rust:v2.57.1 + noetl-worker-rust:v5.15.0-tools223): - tests/k66b/pg_exec_id_multistatement (query: alias) — both single-statement (`SELECT {{ execution_id }} AS exec_id_one`) and multi-statement (`CREATE TEMP TABLE + DELETE WHERE eid = + INSERT VALUES + SELECT`) render `{{ execution_id }}` correctly; the wire result returns the expected execution_id in every position. - tests/k66c/pg_command_alias_repro (command: alias — the wiki-described failing shape) — exactly the `loop_with_pagination`-start-step CREATE+DELETE+INSERT+SELECT pattern; `{{ execution_id }}` renders correctly, INSERT succeeds, verifying SELECT returns the row. The "`WHERE execution_id = ;` syntax error at command-build time" fingerprint no longer reproduces. Likely fixed via the template defer-render series in server#71/#73. Umbrella-49 page: strikethrough applied to "Next concrete steps" item 4's "Open server-side finding" + resolution note; "Porting gap noted: artifact tool kind" also updated to reflect tools v2.20.0 + #64 closure (was already shipped, was a stale note). Home: Last refreshed prefix updated to lead with the housekeeping finding. Sessions-Log: new dated entry. Refs noetl/ai-meta#49
Wiki: Phase D R5 R7 ships — Replay engine port COMPLETE (server#157, v2.57.0; closes server#148) Server#157 merged → release-please tagged v2.57.0 → sub-issue server#148 auto-closed at 17:53:40Z via PR body's `Closes noetl/server#148` keyword. All seven Phase D R5 rounds shipped today (v2.51.0 → v2.57.0): R1 endpoint scaffold + execution projection (v2.51.0) R2 stages + frames + commands projections (v2.52.0) R3 loops + business_objects projections (v2.53.0) R4 typed Checksum + projection_checksums (v2.54.0) R5 snapshot seed + base_state + upcaster digest (v2.55.0) R6 payload resolver (v2.56.0) R7 cross-server parity harness against Python (v2.57.0) The Replay engine port — Python's ~1236-LoC `noetl/server/api/replay/service.py` → Rust — is complete with: - 564/0/0 unit tests covering every projection + helper + fallback path. - 8-test cross-server parity harness with verbatim-extracted Python regen for fixture refresh. R7 is test-only (no runtime changes); no kind-val required. release-please bumped to v2.57.0 from the `feat(replay):` commit prefix. Touches the four pages that drift together per wiki-maintenance.md Rule 0a (Home + Sessions-Log + Releases + Umbrella-Rust-Server-Port). Closes noetl/server#148 Refs noetl/ai-meta#49 Phase D R5
Wiki: Phase D R5 R6 ships — payload resolver (server#156, v2.56.0, kind-val GREEN) Server#156 merged + image v2.56.0 built + loaded into kind + rolled deployment + replay endpoint re-probed against two executions: 1. fanout_reduce execution 322023958058635264 — identical shape as v2.55.0 (no result.reference events; back-compat preserved). 2. Execution 640422512395813188 — execution.payload_refs populates with 3 entries carrying real SHA-256 hex digests (d0de6b8...). Live resolver path confirmed end-to-end. Every event's result.reference JSON now gets parsed into a typed PayloadSummary and appended to the relevant projection's payload-refs list. PayloadSummary + PayloadRefEntry types mirror Python's dict shapes; extract_payload_ref + payload_summary mirror Python's helpers with three-tier fallback chain. ReplayExecutionState.payload_refs + ReplayFrameState.output_ref/output_ref_summary + ReplayBusinessObjectState.payload_refs/last_payload_ref all populated. Only R7 (cross-server parity harness against Python) remains in the Phase D R5 umbrella. Touches the four pages that drift together per wiki-maintenance.md Rule 0a (Home + Sessions-Log + Releases + Umbrella-Rust-Server-Port). Refs noetl/ai-meta#49 Phase D R5
Wiki: Phase D R5 R5 ships — snapshot seed + base_state + upcaster digest (server#155, v2.55.0, kind-val GREEN) Server#155 merged + image v2.55.0 built + loaded into kind + rolled deployment + replay endpoint re-probed against the prior fanout_reduce execution. Wire-shape back-compat verified: new fields `replay_snapshot` + `upcaster_registry_digest` both `skip_serializing_if Option::is_none` — default folds produce R1-R4-identical JSON. The replay fold can now start from a prior fold's output and continue from there. ReplaySnapshotSeed mirrors Python's frozen dataclass; ReplaySnapshotInfo is the output subset; ReplayFoldOptions carries the optional inputs; new fold_replay_state_with_options entry point + 5-arg fold_replay_state as a back-compat shim. Snapshot-storage backend deferred to a downstream sub-issue. Touches the four pages that drift together per wiki-maintenance.md Rule 0a (Home + Sessions-Log + Releases + Umbrella-Rust-Server-Port). Refs noetl/ai-meta#49 Phase D R5
Wiki: Phase D R5 R4 ships — typed Checksum + projection_checksums (server#154, v2.54.0, kind-val GREEN) Server#154 merged + image v2.54.0 built + loaded into kind + rolled deployment + replay endpoint re-probed against the prior fanout_reduce execution. Every fold now produces a typed `Checksum` over the full state + a 6-entry `projection_checksums` map (`execution` / `stage` / `frame` / `command` / `business_object` / `loop`). Per user direction this session: the hash function is the *type* of the checksum (not a sibling field) — `ChecksumType` enum gates future types (BLAKE3, SHA-512, …) without a wire-format break. Kind-validated: - top-level checksum 41265876487f32350fc60c5039358456ded76598b99e7a0833ac4a17ceaae426 - 6 projection_checksums entries all populated with the same typed {type: "sha256", value: "<64-char hex>"} shape Touches the four pages that drift together per wiki-maintenance.md Rule 0a (Home + Sessions-Log + Releases + Umbrella-Rust-Server-Port). Refs noetl/ai-meta#49 Phase D R5
Wiki: Phase D R5 R3 ships — loops + business_objects projections (server#153, v2.53.0, kind-val GREEN) Server#153 merged + image v2.53.0 built + loaded into kind + rolled deployment + replay endpoint re-probed against the prior fanout_reduce execution. All five per-projection maps (stages / frames / commands / loops / business_objects) are now typed BTreeMaps with deterministic key ordering — ready for R4's typed `Checksum` + `projection_checksums` bundle. Loops + business_objects maps stay empty for the fanout_reduce fixture (the v10 control-flow shape doesn't emit `loop.*` or business-object events — expected per the PR body). Fold correctness for those projections is verified through the unit-test layer (`fold_populates_loop_with_counters_and_completion`, `fold_populates_business_object_through_lifecycle`). Touches the four pages that drift together per wiki-maintenance.md Rule 0a (Home + Sessions-Log + Releases + Umbrella-Rust-Server-Port). Refs noetl/ai-meta#49 Phase D R5
Wiki: replace "canonical checksum" with the typed-Checksum shape for R4 Per user direction (writing-style banned-word rule + future- proofing): the Phase D R5 R4 design lands a TYPED checksum shape rather than the flat Python pair (`checksum_algorithm` + `checksum`). Reasons: 1. The hash function in use is the *type* of the checksum, not a separate sibling field. 2. We may add more checksum types in the future (BLAKE3 for speed, SHA-512 for compliance contexts, ...) — the typed shape slots them in without a wire-format break. Proposed Rust types (final names land in the R4 PR): enum ChecksumType { Sha256, /* future: Blake3, Sha512, ... */ } struct Checksum { type: ChecksumType, value: String /* hex */ } // in ReplayState: checksum: Option<Checksum> projection_checksums: BTreeMap<String, Checksum> Underlying hash stays SHA-256 — same algorithm Python uses today (`_canonical_checksum` in `service.py`, `hashlib.sha256`). Parity test in R7 asserts byte-for-byte match on the hex string. Touches the four pages that drift together per wiki-maintenance.md Rule 0a (Home + Sessions-Log + Releases + Umbrella-Rust-Server-Port) — every R4 row updated. Refs noetl/ai-meta#49 Phase D R5 R4 Refs noetl/server#148
Wiki: Phase D R5 R2 ships — stages + frames + commands projections (server#152, v2.52.0, kind-val GREEN) Server#152 merged + image v2.52.0 built + loaded into kind + rolled deployment + replay endpoint re-probed against the prior fanout_reduce execution. Commands map populates with 4 entries carrying worker_id + issued_event_id + last_event_id; stages + frames stay empty (the v10 control-flow shape doesn't emit stage.* / frame.* events — expected per the PR body). Touches the four pages that drift together per wiki-maintenance.md Rule 0a: - Home.md — Last refreshed + #49 row last-update flipped to 2026-06-07 + server-cell version v2.48.0 → v2.52.0 with the R5 R2 + R5 R1 + R4 follow-ups recent-activity sweep. - Sessions-Log.md — new entry prepended at the top. - Releases.md — v2.52.0 row prepended at the top of the timeline table. - Umbrella-Rust-Server-Port.md — Recent-activity row + new "Phase D R5" sub-table tracking R1+R2 ✅ / R3 next / R4-R7 pending. Refs noetl/ai-meta#49 Phase D R5
wiki: server#149 merged — Phase D R5 R1 Replay endpoint scaffold (v2.51.0) Sessions-Log: new dated entry at top documenting the Phase D R5 opening + R1 ship (server@f77dead, v2.51.0). Includes a Phase D R5 status table showing R1 merged + R2-R7 in queue. Home: Last-refreshed cell prefix updated with the new headline + server pointer bump. Releases.md: prepended row for noetl-server v2.51.0 + Last refreshed cell updated. Umbrella-Rust-Server-Port: Recent-activity new row at top for Phase D R5 R1 with the full 7-round decomposition cited. Refs noetl/ai-meta#49 Phase D R5 Refs noetl/server#148 noetl/server#149 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
wiki: server#147 merged — status query bug fixed + kind-val GREEN (v2.50.1) Sessions-Log: new dated entry at top documenting the read-side status-query fix (server@d26abf8) + kind-val evidence (prior execution flipped RUNNING→COMPLETED on same DB data; fresh fanout_reduce execution reached COMPLETED in ~600ms wall). Home: Last-refreshed cell prefix updated with the new headline + server pointer bump. Releases.md: prepended row for noetl-server v2.50.1 + Last refreshed cell updated. Umbrella-Rust-Server-Port: Recent-activity new row at top for the Phase D R4 read-side fix. Refs noetl/ai-meta#49 Phase D R4 Refs noetl/server#146 noetl/server#147 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
wiki: Phase D R4 — fanout_reduce kind-val GREEN on Rust-only stack Sessions-Log: new dated entry at top documenting the end-to-end green-state evidence. Built noetl-server-rust:v2.50.0 from server@499b079 via podman + Dockerfile, loaded into kind, rolled the deployment, ran the fanout_reduce_phase6 fixture from e2e@5da36ea. Direct DB query against noetl.event confirms all three barrier assertions pass. Home: Last-refreshed cell prefix updated with the green-state headline + event-log evidence + the surfaced status-query bug (noetl/server#146). Umbrella-Rust-Server-Port: Recent-activity new row at top documenting the green-state run. Phase D status table updated to mark slices 1-3 as GREEN at the orchestrator AND kind-val levels. Refs noetl/ai-meta#49 Phase D R4 Refs noetl/server#146 (read-side status-query bug surfaced during the run) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
wiki: e2e#32 merged — Phase D R4 fanout_reduce kind-val rig ready Sessions-Log: new dated entry at top documenting e2e#32 (e2e@5da36ea) — durable fixture + script for the canonical fanout_reduce shape. Includes a Phase D R4 status table showing slices 1-3 all shipped at the artefact level; run-on-kind is the remaining housekeeping step (needs a fresh v2.50.0+ server image loaded into kind). Home: Last-refreshed cell prefix updated with the new headline + e2e pointer bump + Phase D R4 status. Umbrella-Rust-Server-Port: Recent-activity new row at top for Phase D R4 slice 3. Phase D status table extended with the slice 3 row + the slice 1/2 kind-val column updated to note the rig is in place. Refs noetl/ai-meta#49 Phase D R4 Refs noetl/e2e#31 noetl/e2e#32 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
wiki: server#145 merged — Phase D R4 slice 2 closes barrier follow-up (v2.50.0) Sessions-Log: new dated entry at top documenting server#145 (server@499b079, v2.50.0) closing the gap exposed by slice 1's ignored test. Includes a Phase D R4 status table showing slice 1 + 2 shipped at the orchestrator level; remaining work is kind-validation (slice 3) + possibly context-merging (slice 4, deferred until kind-val signals a need). Home: Last-refreshed cell prefix updated with the new headline + server pointer bump. Releases.md: prepended row for noetl-server v2.50.0; Last refreshed cell updated. Umbrella-Rust-Server-Port: Recent-activity new row at top for Phase D R4 slice 2; Phase D status table extended with the slice 2 row. Refs noetl/ai-meta#49 Phase D R4 Refs noetl/server#144 noetl/server#145 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
wiki: cli#56 + server#143 merged — executor 0.4.1 + Phase D R4 fan-in barrier Sessions-Log: new dated entry at top documenting the two PR merges (cli@77be8be v4.10.0 + server@be37e5c v2.49.0) + the worker PR status (still draft until the noetl-executor 0.4.1 publish lands). Home: Last-refreshed cell prefix updated with the new headline + links. Releases.md: prepended rows for noetl-server v2.49.0 (Phase D R4 fan-in barrier) and noetl/cli v4.10.0 (noetl-executor 0.4.1 bridge propagation); Last refreshed cell updated. Umbrella-Container-Tool-Callback: Round-4-follow-up status table updated — cli#56 marked MERGED, worker#60 still blocked on the 0.4.1 publish. Umbrella-Rust-Server-Port: new Recent activity row for Phase D R4 slice 1 (fan-in / reduce barrier). Phase D status table gets a new row marking R4 slice 1 as v2.49.0 shipped with kind-val pending. Refs noetl/ai-meta#43 Refs noetl/ai-meta#49 Phase D R4 Refs noetl/cli#56 Refs noetl/server#142 noetl/server#143 Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
docs(wiki): orchestrator-strand fix landed (server v2.27.2) + e2e sweep #54 - Releases: v2.27.2 row + Last-refreshed. - Home: server cell -> v2.27.2 (strand-fix headline); #54 row + Last-refreshed marked landed; Secrets-Wallet version line -> v2.27.2. - Sessions-Log: entry updated in-flight -> landed (pointers 5808486 / f3a49b2). - Umbrella-Rust-Server-Port: Recent-activity row for the e2e sweep + fix. Refs noetl/ai-meta#54
Rust-only regression rig: canonical v10 SQL + http config shapes (tools v2.18.4) - Sessions-Log: new entry for the regression sweep + the 3 config-shape fix classes (tools#24/#25/#26, worker#50/#51). - Releases: tools v2.18.3 + v2.18.4 rows; Last refreshed bumped. - Home: Last refreshed, Sessions + Releases previews, Ecosystem map tools→v2.18.4 / worker→2.18.4-lock, #54 umbrella status. - Umbrella-Rust-Server-Port: Recent-activity row + Next-steps R5 rig item + open loop_with_pagination server-side finding + artifact porting gap. Tracks noetl/ai-meta#54.
dashboard: postgres-tool observability — real SQLSTATE errors (tools v2.18.2 + worker#49) Closed the last credential/iterator follow-up — postgres tool surfaces real SQLSTATE + message instead of opaque "db error". Validated e2e on the Rust-only kind stack. - Home.md: Last refreshed; #54 row; ecosystem map tools v2.18.2 + worker noetl-tools-2.18.2 note; Sessions + Releases previews. - Sessions-Log.md: new 2026-06-05 postgres-observability entry. - Releases.md: tools v2.18.2 row. - Umbrella-Rust-Server-Port.md: R5 Tier 4 observability row.
dashboard: iterator_save_test GREEN — full v10 + credential + iterator-pipeline surface (server v2.19.7) server#73 (v2.19.7) defers task_sequence _prev/_results refs at command-build — iterator_save_test reaches playbook.completed and writes 3 rows to the real demo_noetl DB. Closes the credential + iterator + pipeline chain (server#71, worker#46, worker#48, server#73). - Home.md: Last refreshed; #54 row; ecosystem map + diagram server cell v2.19.7; Sessions + Releases previews. - Sessions-Log.md: new 2026-06-05 iterator_save_test-GREEN entry. - Releases.md: server v2.19.7 row. - Umbrella-Rust-Server-Port.md: R5 Tier 4 iterator_save_test-GREEN row.
dashboard: nested-pipeline credentials resolved (worker v5.11.3) + server#72 finding worker#48 (v5.11.3) closes the credential-path chain — nested task_sequence sub-task aliases now resolve; iterator_save_test's save_item connects to demo_noetl. Last blocker filed: server#72 (server pre-renders task_sequence _prev/_results refs to empty → malformed SQL). - Home.md: Last refreshed; #54 row; ecosystem map + diagram worker cell v5.11.3; Sessions + Releases previews. - Sessions-Log.md: new 2026-06-05 nested-credentials entry. - Releases.md: worker v5.11.3 row. - Umbrella-Rust-Server-Port.md: R5 Tier 4 nested-credentials row.
dashboard: keychain-credential path validated (server v2.19.6 + worker v5.11.2) R5 Tier 4 credential-path widening — 3-bug chain in the keychain subsystem; plain-step credential path proven end-to-end against the real demo_noetl DB. - Home.md: Last refreshed; #54 row credential-path note; ecosystem map + diagram version cells (server v2.19.6, worker v5.11.2); Sessions + Releases previews prepended. - Sessions-Log.md: new 2026-06-05 keychain-credential entry. - Releases.md: server v2.19.6 + worker v5.11.2 rows. - Umbrella-Rust-Server-Port.md: R5 Tier 4 credential-path row.
dashboard: v10 control-flow end-to-end on Rust-only (server v2.19.5 + worker v5.11.1 + tools v2.18.1) Phase F R5 Tier 4 — four v10 fixtures reach playbook.completed on the Rust-only kind stack after 7 fixes across server/worker/tools. - Home.md: Last refreshed 2026-06-05; #54 row updated; ecosystem map + architecture diagram version cells bumped; Sessions + Releases previews prepended. - Sessions-Log.md: new 2026-06-05 entry (7-fix inventory + 5-fixture validation table + pointer bumps). - Releases.md: server v2.19.5, worker v5.11.1, tools v2.18.1 rows. - Umbrella-Rust-Server-Port.md: R5 Tier 4 recent-activity row.
docs(wiki): correct forward-dated entries (2026-06-07 → 2026-06-04) Sweep across three wiki files; 38 occurrences corrected. Sessions-Log.md (14 entries), Releases.md (14 rows), Umbrella-Rust-Server-Port.md (10 rows) all had Phase F R3 + R4 work stamped with 2026-06-07. The introducing commits' author date was 2026-06-04 in every case — a previous agent session had a wrong "today's date" mental model and forward-dated by three days. User reported the visible mismatch on the wiki: https://github.com/noetl/ai-meta/wiki/Sessions-Log#2026-06-07-phase-f-r4-5--kind-validation-n2-shards-r4-complete Verified via `git log -S "<date string>" --date=short -- <file>`: - Phase F R3b-1 entry first appeared in 2170690 (author 2026-06-04). - Phase F R3b-2 entry first appeared in d66224b (author 2026-06-04). - Phase F R4-1 through R4-5 entries all in commits authored 2026-06-04. Future drift prevented by agents/rules/wiki-maintenance.md Rule 0a's new "Date discipline" section (committed in the same change set as the parent ai-meta repo).
docs: Rust-only e2e complete + legacy cleanup — noetl/server v2.19.4 + noetl/tools chain control_flow_workbook runs FULLY end-to-end on the Rust-only stack — first NoETL playbook exercising the complete control-flow surface (workbook → python with result capture → conditional next.arcs → parallel branches → terminal) without touching Python. Five fixes shipped across three repos in one iteration: - noetl/tools#17 (v2.17.1) — PythonTool result-global capture - noetl/tools#18 (v2.18.0) — TaskSequenceTool runtime - noetl/worker#42 — Cargo.toml bump 2.16 → 2.18 - noetl/server#67 (v2.19.4) — orchestrator template context (expose step data at top level + capture call.done) - Kind cluster legacy cleanup (delete Python deployments + services + configmaps + SA) Closes noetl/ai-meta#60 (the umbrella; auto-closed via PR), noetl/tools#15, noetl/tools#16, noetl/server#66. - Releases.md: add v2.19.4 + v2.18.0 + v2.17.1 + worker bump rows (in chronological top-of-timeline order). - Sessions-Log.md: prepend 2026-06-04 entry covering the full iteration including kind cluster cleanup. - Umbrella-Rust-Server-Port.md: add the consolidated row above the prior v2.19.3 row.
docs: noetl/server v2.19.3 — pipeline + failure + workbook fixes shipped together Three orchestrator + parser fixes from the same e2e probe sequence on the Rust-only stack: - noetl/server#61 (Gap C — pipeline flat shape decode) - noetl/server#63 (Gap B — playbook.failed on command.failed) - noetl/server#65 (Gap A — tool.kind:workbook resolution) Closes noetl/ai-meta#57, #58, #59 (and their server sub-issues #60, #62, #64). - Releases.md: add v2.19.3 row consolidating the three fixes. - Sessions-Log.md: prepend session entry covering the probe sequence + fixes + follow-up gaps filed. - Umbrella-Rust-Server-Port.md: add the v2.19.3 row above the v2.19.2 row in chronological recent-activity order. Two noetl-tools follow-ups now tracked separately at noetl/cli#53 (task_sequence runtime) and noetl/cli#54 (PythonTool result capture) — independent rounds for noetl-tools work, not server.
docs: noetl/server v2.19.2 (v10 workload + input alias) + Rust-only direction - Releases.md: add v2.19.2 row for noetl/server#59. - Sessions-Log.md: prepend 2026-06-04 entry covering both the Rust-only standing direction and the v10 playbook compat fixes (input alias + workload merge). - Umbrella-Rust-Server-Port.md: add the v10 compat row above the EE-5 row in chronological recent-activity order. Closes noetl/ai-meta#56 (the umbrella; auto-closed via PR).
docs: noetl/server v2.19.1 (EE-5 lax decode) + Phase F R5 unblocker - Releases.md: add v2.19.1 row for noetl/server#57. - Sessions-Log.md: prepend 2026-06-04 session entry describing the Rust worker → Rust server execution_id wire-shape drift, root cause, fix, validation, and tracking trail. - Umbrella-Rust-Server-Port.md: add EE-5 follow-up row before the R4-5 row (chronological top-of-recent-activity). Closes noetl/ai-meta#55 (the umbrella; auto-closed via PR).
log: Phase F R4-5 shipped — R4 complete R4-5 kind validation landed via noetl/ops#160. Phase F R4 is end-to-end complete across noetl/server v2.13.0 -> v2.19.0 + noetl/ops validation script. noetl/server#48 closes. Only remaining Phase F round: R5 (production cutover; separate ops decision). Phase G (keychain-backed shards) deferred. Sessions-Log + Releases timeline + Umbrella-Rust-Server-Port all updated; Next-concrete-steps reframed around R5 + Phase G.