release(contract): publish immutable temporal-analysis contract for Orgmetra consumers

Cross-repository release gap

Fresh Orgmetra protected-truth review found that ContextualWisdomLab/Orgmetra still documents tepp_adapter as orgmetra.tepp.v1 pinned directly to repository commit 40adac9a26a8af85147ffa2795fb548ea243e0e5.

Fresh TEPP authority is protected main@a243f18da4a4ca8a8d068c39922537f1f8ed6ad0, while the GitHub Release inventory is currently empty. The old commit pin is therefore neither current owner truth nor an immutable published consumer release under the CWL released-contract rule.

Orgmetra issue #431 owns the consumer-side integration correction. This issue owns the TEPP-side prerequisite: publish the smallest supported immutable contract needed by Orgmetra workforce_validation/longitudinal-analysis consumers rather than making Orgmetra depend on mutable main, a PR head, or copied TEPP source/schema.

Owner boundary

TEPP remains authoritative for its temporal/text/scientific analysis semantics and implementation. Orgmetra remains authoritative for HRIS/HCM employment truth, Workforce Validation study design/interpretation, purpose-bound authorization and downstream employment decisions.

A TEPP result/receipt is scientific evidence only. It must not become an employment-decision, validity/fairness or lifecycle authority merely because execution succeeded.

Required released projection

First inventory the existing TEPP public API/package surfaces and select the narrowest stable projection actually required by Orgmetra. Do not invent a second Orgmetra-specific implementation if an existing owner contract can be versioned and released.

The released contract must, where applicable to the chosen TEPP surface, bind:

explicit contract/package/schema identity and semantic version;
exact analysis method/model/configuration identity rather than a floating alias;
input/evidence provenance coordinates sufficient for downstream reconstruction without copying HR person-level truth into the shared contract;
time/window/cohort/multilevel or multiple-membership coordinates required to interpret the result correctly;
exact output schema and scientific-result locator/digest;
explicit failure/non-convergence/not-verifiable semantics rather than coercing absence/failure to a numeric result;
algorithm/runtime/backend identity where result parity materially depends on CPU/GPU/backend behavior; and
compatibility/deprecation/rollback semantics for consumers.

Keep raw HR person-level data, credentials, hidden provider state, unnecessary free text and Orgmetra application-table identifiers out of the shared projection unless a separately justified purpose-bound contract requires them.

TDD / scientific acceptance

Before release, add owner-side contract tests that prove at least:

deterministic canonical serialization/versioning for the released projection;
malformed, floating, unknown or incompatible method/config/schema coordinates fail closed;
scientific failure/non-convergence/insufficient-evidence states cannot deserialize as successful estimates;
time, grouping, multilevel/cross-classified/multiple-membership semantics required by the method are not silently discarded;
no mutable branch/PR/source path is needed by a consumer to reconstruct the released evidence;
CPU/GPU parity evidence is retained where the released method exposes multiple backends, with declared tolerances and no skipped-backend false GREEN; and
consumer fixtures allow Orgmetra to reject a mismatched study/result coordinate and preserve not_verifiable.

Synthetic data may prove mechanism and true-parameter recovery where scientifically appropriate, but any buyer/scientific claim that requires real data must use right-cleared evidence and preserve sampling/error denominators.

Release acceptance

Do not close at source merge. Publish an immutable GitHub/package release and record:

exact protected-main source head;
version/tag/artifact identity and digest;
SBOM and provenance;
reproducibility and compatibility evidence;
deprecation/rollback semantics;
exact-head test, statement/branch/rustdoc/docstring, security and scientific evidence required by the TEPP release gate; and
zero unresolved valid owner findings under then-live governance.

After release, Orgmetra must pin that exact supported release through its ACL and run its own consumer-conformance tests. Until then, the old 40adac9... commit pin remains design/history evidence only, not production contract authority.

Refs ContextualWisdomLab/Orgmetra#431. No source copy, cross-service SQL, force-push, destructive rebase, self-approval, gate weakening or synthetic release evidence.

test(api): reject non-verifiable successful analysis results (#638)

test(api): align release validation-state RED with existing wire field (

fix(api): fail closed on non-validated success summaries (#638)

docs(changelog): record fail-closed release validation state (#638)

docs(contract): define temporal-analysis success/failure boundary (#638)

Fresh owner review found a concrete release-contract ambiguity on protected main@a243f18da4a4ca8a8d068c39922537f1f8ed6ad0: tepp_api::AnalysisResultSummary.validation_status accepts any non-empty string, so a wire payload with run_state = succeeded can carry validation_status = "not_verifiable", "non_converged", or another non-success value while still retaining a result artifact/digest. This violates #638 acceptance item 3 and must be fixed in TEPP, not reinterpreted in Orgmetra.

Draft owner PR #639 now carries the bounded repair from the protected base:

source-level RED 59ce5c971990ecbfd89d4c95a7dbb0dfe06ddf31;
RED fixture aligned to the existing wire surface 4cdf663022560230653f24b8369e87ec5d0b6e84;
causal fix efd4e9e613ad2bcad036d1a0d271ef3a5a0ec300, admitting only canonical validated in successful summaries while not_verifiable / non_converged / insufficient_evidence remain artifact-free terminal failures;
changelog 5f0e592dde9ba71fe9e11fd1c8e0d81701ccea26;
owner-contract doctoring/current head faaac1ac6bed2a4f54c8d2e0c89473e60f06c218.

#639 deliberately reuses the existing AnalysisRun* contract instead of creating an Orgmetra-specific projection. It does not close #638: immutable package/tag/artifact identity, SBOM/provenance, reproducibility/compatibility, rollback, exact-head test/security/scientific evidence, and independent review are still required before Orgmetra #431 can pin a supported release. At creation the PR is Draft/open/mergeable; no hosted current-head workflow run had been admitted yet, so no GREEN is claimed.

test(api): reject unknown provider success labels (#638)

TRACEABILITY update: #639 advanced ordinary-forward to exact ce9f948f426d3bb322d859fac8ef9ad8c995d0a4 after extending the contract to reject an unknown non-empty provider label (provider_custom_success), proving the success-state rule is closed (validated only) rather than a denylist of known failures. Base remains protected main@a243f18da4a4ca8a8d068c39922537f1f8ed6ad0, compare is ahead 6 / behind 0 with merge base equal to protected main. Current-head Rust Foundation 35578670108, Security 35578670198, CodeQL 35578670267, Semgrep 35578670260, and Documentation 35578670435 are admitted but queued; Rust Foundation jobs currently have no runner and no steps. No hosted GREEN or independent approval is claimed.

test(api): prove request serialization is clock-independent (#638)

fix(api): separate request wire determinism from live cutoff admission (

docs(changelog): record clock-independent analysis request serializat…

docs(contract): separate replay serialization from live cutoff admiss…

#639 has now advanced three #638 prerequisites without widening the consumer boundary.

Successful results fail closed unless AnalysisResultSummary.validation_status == "validated"; non-converged / insufficient-evidence / unverifiable outcomes stay artifact-free terminal failures.
Source RED 4a9fc85ca4b2f1a82fbd595013b13504f2b09059 → causal fix f0acd5d8b002c8512caf7c4300f680eb400efcc7 separates canonical request serialization from live cutoff admission. A syntactically valid future cutoff can be serialized as immutable request evidence without consulting Timestamp::now(), while live from_json still refuses the not-yet-available cutoff.
Fresh follow-up found that RFC 3339 offset aliases still allowed multiple valid byte identities for one cutoff instant. Source RED 4da3adb0a919a50ad57d93fdcaa780ff73c2ab20 requires canonical UTC 2026-08-01T00:00:00Z to remain valid while equivalent 2026-08-01T09:00:00+09:00 fails both serialization and live decoding. Causal fix 5c6c059f5e934826167d98a5ba7c2435b4eb1e35 parses through KnowledgeCutoff and requires exact equality with KnowledgeCutoff::to_rfc3339().

Current doctoring/exact head is d52ce4f88527dacbf8275fee4faa02a62420f3fa; protected base remains main@a243f18da4a4ca8a8d068c39922537f1f8ed6ad0, ahead 14 / behind 0. Current exact-head Rust Foundation 35581623907, Semgrep 35581623798, Security 35581623913, CodeQL 35581623903, and Documentation 35581624361 are non-terminal; all four Rust jobs have no assigned runner and steps=[]. No current-head review exists, so no hosted GREEN or merge authority is claimed.

Scope remains narrow. #638 is still open for full method/configuration and provenance coordinates; time/window/cohort; multilevel/cross-classified/multiple-membership structure; uncertainty/backend identity where material; consumer mismatch fixtures; exact-head scientific/security gates and independent review; protected merge; immutable package/tag/artifact digest; SBOM/provenance; reproducibility/compatibility; rollback/deprecation; and the actual TEPP release. Orgmetra #431 must continue to fail closed rather than consume this Draft head.

test(api): require canonical UTC cutoff identity in analysis requests (…

fix(api): canonicalize request cutoff byte identity (#638)

#646 owner-path follow-up on canonical #605: #645 count/min/max provenance was still non-unique, so a same-shaped cohort could reuse an owner-derived design classification without an exact support mismatch signal. Source RED 7a71fef17e5873669261a200b68e4e4012cdeca4 requires order-insensitive but multiplicity-sensitive exact support identity and distinguishes same design/count/window with a different opaque member coordinate. Production lineage adds sha2 (4c27462e...), canonical owner digest 76ac7441..., public digest-version export 2f34c69e..., explicit version/topology fixtures 35e55783... / 477bea00..., changelog c7740c8f..., and reachable-topology canonicalization/current head 539c5ee7b7f85e22dce4c6f7d2ba0a9222318925.

tepp.membership_observation_support.v1 hashes the sorted observation multiset and, for each (MemberId, EventTime), sorted active role/group plus exact binary64 weight bits with explicit framing. Reordering cannot change identity; duplicate observations remain multiplicity-sensitive; changing the opaque support or active group topology changes identity.

Boundary remains strict: SHA-256 is identity binding, not reconstruction. #638/#639 still need a released, privacy-appropriate, interpretable observation-support/time/window/cohort projection and must verify it against this owner digest after #605 lands normally.

Membership owner update from #605/#649, exact source head e425c7034da3f505c8dff49fc19d3ffe80adf40d.

#646's owner SHA-256 closed exact support/topology identity but was intentionally not reconstruction evidence. #649 now adds the missing Membership-owned reconstructable longitudinal support projection: tepp.membership_observation_support_projection.v1 carries projection-local member/group ordinals, canonical UTC event times, stable role names, exact binary64 membership-weight bits, declared design version/name, and the exact #646 source-support digest. Raw MemberId/GroupId UUID strings are not serialized; this is purpose-scoped privacy reduction, not anonymity.

RED ba3d8a674e88dedb73fd5b225a71d5446a88e659 → initial owner projection b7d3dad74b64e3560627566312b3d806282eee0b → hardening/export/fixture/changelog through 55841295677be6905fb555c4e2cef259c4e891a6. Fresh review then found parsed bytes could relabel a nested topology as another valid design name. RED d8e70c2bbe62c69fd3873379ea7a2baf37510538 → canonical Membership design-signal reuse 1716b2f31278a4a727f24160b4dd21ba42407797 → topology-bound parser fix b6adb39cc80a5c117874355e2248cd1a20beadfc → current doctoring/coverage e425c7034da3f505c8dff49fc19d3ffe80adf40d.

Consumer boundary remains strict: #639 must not copy this mutable source or reproduce ordinal/digest/design mapping. After #605 lands on protected main, #639 can bind the owner-issued support projection and add consumer mismatch fixtures. This closes the Membership-side reconstruction coordinates for observation-specific time, repeated-member/group linkage, role, topology, weights, design and support identity; Analysis still owns broader run/window/cohort semantics and must not infer HR/person truth from local ordinals.

Current-head gates are fresh but non-terminal: Rust Foundation 35642670231, Security 35642670233, Semgrep 35642670345, CodeQL 35642670212, Documentation 35642670208. No hosted GREEN or release authority is claimed.

Follow-up Membership owner coordinate gap #650 is now implemented on #605 exact 2eca15330e52e585ed4044fc152b88fac29ca2ec.

#649's canonical support wire is intentionally sorted/privacy-reduced, so it could not by itself tell Analysis which of its own documents should be associated with which local member ordinal without consumer-side raw-ID sorting. RED 0355549c187043768e69c0d4a667b18e3a27e672 captures that DDD leak. e2e8eb2668c18d2e65d47ef8f885ec37e6f6de78 adds Membership-issued non-wire input coordinates (member_ordinal, EventTime) parallel to the admitted observation slice; ce4d068a1381debb6a8e9d8503e44a025dd76b45 exports the contract; 2eca153... doctoring is current.

After #605 protected landing, #639 can zip its own document IDs with input_coordinates() and serialize the owner support wire without reproducing UUID ordering. Duplicate observation coordinates remain duplicate; canonical support bytes remain input-order insensitive. This is still mutable owner source today, not release authority.

Membership owner hardening #651 is now source-complete on #605 exact 3177e98c40135368f81644bf5354baaf80216a46.

Fresh review found tepp.membership_observation_support_projection.v1 could deserialize states canonical MembershipNetwork can never admit: duplicate simultaneous (role, group) edges with different valid weights, and same-role weights whose exact represented binary64 sum exceeds 1.0 (including 1.0 + min_subnormal, which ordinary f64 accumulation can round back to unity). RED 2774e1c327f971dcba370dfec28dc74b81f49c99 → shared owner-internal exact budget predicate fff822d40cd8ad28f6ed8aa3c17b63beafbcbf15 → parser repair 8a73734d3f3f36dfa8f141bacea6c08171cca938 → doctoring 3177e98c40135368f81644bf5354baaf80216a46.

This keeps reconstruction inside the owner aggregate's reachable state space without duplicating numerical/share-budget semantics in Analysis. #639 must still wait for protected-main landing; no mutable source consumption is authorized.

Membership owner handoff update from #605/#652: the current owner branch now distinguishes two evidence identities. support_sha256 remains the exact opaque-source support/topology fingerprint and is intentionally not recomputable after UUID redaction; new tepp.membership_observation_support_projection_digest.v1 is recomputed from the complete privacy-reduced serialized projection, including the source digest, local member/group ordinals, canonical event times, roles and exact binary64 weights. RED 028a0660eeea76c6ce0c16ecfd7b179accb99a9d → causal repair 4e5b64e663868d63a77d9dbed2baedd7cae27621 → export/coverage/changelog through #605 exact 32871b52e801bd501d3ff1f4fee8cd12f17530b3.

For #638 acceptance, Analysis should not reproduce either digest algorithm or treat the recomputable wire digest as authentication. After #605 lands/release authority exists, #639 should consume the owner projection, bind its own document IDs through input_coordinates(), retain the canonical support wire, and add mismatch fixtures. Release provenance/signature remains separate from this deterministic content-binding contract.

Membership owner-path update from canonical #605: issue #653 found that tepp.membership_observation_support_projection.v1 could represent the same local (member_ordinal, EventTime) twice with different individually valid assignment topology. That state is impossible under deterministic canonical MembershipNetwork resolution even though duplicate observation multiplicity itself is valid.

#605 lineage: RED 0bee645841bc9bb5029e4d82ec0083663d291395 → causal parser repair 8cd8a6db62c5ce75a56cd342302bb3c9466704e4 → CHANGELOG/current exact head ef50885bc21e7e1247a22e4ce1e1c59d8efbd331. Repeated identical coordinates remain valid/multiplicity-sensitive; contradictory repeats fail closed.

This is still mutable Draft owner work. #638/#639 must not copy the parser or consume this head. Wait for normal #605 protected-main landing, then bind the released/landed Membership projection and add consumer mismatch fixtures.

Fresh #639 review found a separate deterministic-byte-identity defect in the digest-bound topic-lineage artifact: Uuid::parse_str accepted alternate textual aliases for predecessor/successor document IDs even though the producer emits canonical Uuid::to_string() text. That allowed semantically identical document coordinates to mint different valid artifact bytes/SHA-256 identities.

Owner lineage is now #654 RED 153d6f62ecd0d056ba5152626a43588517714095 → causal validation repair d87f7f5295e476f016352df2e8b07840c5b42c24 → CHANGELOG 35c674e098d16c41720996e78579de0eb5ef547e → release-contract doctoring/current #639 head 3343f0ca7a9f62ef84338b496b3f9817dc0c09cd.

The fix keeps UUID identity, edge topology, estimator arithmetic and schema v2 unchanged; it only requires stored predecessor/successor text to equal the lowercase hyphenated representation returned by Uuid::to_string(). #639 is ahead 53 / behind 0 from protected main@a243f18d...; current exact-head workflows are non-terminal, so no hosted GREEN/release claim is made. Membership/time-window integration and the other #638 release gates remain open.

Fresh #639 owner update closes another deterministic configuration-identity defect required by this release contract. Issue #656 established that the zero-capable relation_strength and ridge controls could serialize IEEE-754 -0.0 and 0.0 into distinct digest-bound method-configuration bytes even though both represent the same zero-strength scientific policy.

Lineage is RED 16273447913c20c49e41e7f88324932314f41183 → causal wire repair 86455d8e1ad4b45ea683a0e57ce6ea18a89c1794 → CHANGELOG 57271e66ba078db5c274963e390133049a92832b → release-contract doctoring/current #639 head 38aac65c1c93e1a2da1c1761598b6ef0447e5d8d. The parser rejects sign-negative zero only at the retained v1 configuration serialization boundary; estimator arithmetic and positive numerical domains are unchanged.

#655 also landed ordinary-forward on the same Draft branch before #656: RED c160d0b987c0308b399dd4a9f5cee74bae3a09b6 → canonical edge-order repair 9419edef834e9353b1fe0838367c5203d34155f8 → CHANGELOG a54288c99426a645022c4f85e69ed4925c70c82b → contract doctoring 8c1f7f477056365652e7038c2a4f1b2f28fd573c.

#639 is now ahead 61 / behind 0 from protected main@a243f18d…; its current exact-head workflows are fresh but non-terminal. This remains mutable Draft evidence, not release authority. Membership projection integration, mismatch fixtures, uncertainty quantities/parity, broader provenance, recovery evidence, exact-head GREEN + independent review, protected merge, and immutable package/tag/SBOM/provenance/reproducibility/rollback remain open before Orgmetra can pin TEPP.

Fresh #639 review found another deterministic release-identity defect at the outer artifact boundary: TopicLineageArtifact::from_json accepted byte-distinct JSON (for example trailing newline or pretty printing) even though the result SHA-256 is over compact owner serialization. #657 now owns that finding. Canonical #639 lineage is RED cf637dd9826a92ce361212f43ffe23f104ae9c09 → parser repair 468b702b322e5ebdebab13b581fa7f4a72d7b806 → CHANGELOG b6772ab596c691aa2f068b2fcfd5600d2be261d6 → release-contract doctoring ab7f3afeab7fe2251f69da82dfeaa8e76cd1904f → rustfmt-only current head 49f599c5b132a5d4a1a458645d61e068d12085d3. The parser now requires the supplied v2 bytes to equal owner canonical serialization exactly; it does not normalize aliases. This closes one #638 canonical-serialization prerequisite only. Membership projection, uncertainty quantities/parity, broader provenance, recovery/rolling-origin, compatibility/rollback, protected-head GREEN/review, and immutable release remain open.

New owner-bound provenance gap #658: topic-lineage execution currently accepts a detached snapshot_id string beside an already-built ReferenceTopicInput, so matching the request ID does not prove the estimator population came from that owner snapshot. ReferenceTopicInput::new checks document membership in a supplied CorpusSnapshot but does not retain an owner-issued snapshot/content provenance coordinate. Do not patch this with a free digest. Correct order is Evidence #527 owner settlement/released consumer form as applicable → typed Analysis/Evidence ACL or provenance-bearing admitted input → mismatch RED (input from B while claiming A fails before fitting) + result identity drift on owner source evidence drift + cutoff-safe future exclusion → #639 projection binding. #658 is broader-provenance acceptance under #638, not estimator arithmetic.

Release-contract update from #639, with #660 fixture correction recorded explicitly. #659 has a valid source-level causal RED: duplicate deterministic seeds (079cb2f51ee2f6c6bcd928bcb083ad2837b40d36) → parser repair 5f3e0fc429d1ba9f1cb5199a4c512c598681c12a → CHANGELOG 97e95f8fb140c59858a3598dc2d8a99dadb435d8 → doctoring d19f4bf6f1c31a187ea2d12e6b105f59c54646e0. #660 independently verifies that wire seed 0 aliases seed 1 numerically because the CPU reference RNG starts from seed.max(1); production repair 6c9a86d5105c0137f9da48c8c7d6f9a9e836d03f rejects zero at the unreleased wire boundary, followed by CHANGELOG 9a638c9e... and doctoring 7f3813f0.... The initial #660 fixture f1753a96... was confounded by selected_seed = 1 against [0,11] and is not claimed as RED; corrected regression d7989f9ce7ea2f3a2daa6300f5aa77dadea592a1 binds selected seed zero. Distinct nonzero seed order is preserved; no normalization or estimator arithmetic change. #658 Evidence-owned source provenance and #605 Membership projection remain open prerequisites; no release authority is claimed.

#639 owner-path update: #661 closes one additional digest-bound artifact identity alias without changing estimator arithmetic. Causal source RED f4237a790c93f0b2a6370729d61e83b44031fbd5 proves objective=+0.0 and -0.0 are numerically equal but byte-distinct; minimal Analysis validation repair 82510f0a75f92623b2d49aaa743b5d5d39ccc639 refuses only sign-negative zero. CHANGELOG c1effb2da5097a9317fb6115f00204608764164a and release-contract doctoring 65a42cd2f834e3db0376b2f303484bed62432072 are ordinary-forward on #639.

This does not change #638’s remaining causal frontier: Evidence-owned source-provenance binding (#658/#527), Membership-owned reconstructable longitudinal support (#605), uncertainty quantities/backend parity, mismatch/recovery/rolling-origin evidence, compatibility/rollback, exact-head gates/review, and actual immutable release remain open. #639 stays Draft; no predecessor evidence or consumer activation is claimed.

#662 now has a source-level runtime-causal RED on #639 exact fd6e2f475c2b07dec2674eb91bd69e2fd33616fb. The positive fixture is an otherwise-valid tepp.trsl_topic_lineage.v2 payload with coordinate-bound diagonal_laplace_uncertainty; current owner parsing rejects it only because v2 still has no quantity projection. The earlier compile-time prototype fb8e817e... is retained in history but is not the causal RED claim.

The intended GREEN boundary is narrower than joint posterior work: topic_measurement already owns ReferenceTopicModel::document_coordinate_variances; Analysis should project those existing quantities only after obtaining read-only modeled document IDs from ReferenceTopicInput. Rows must be UUID-canonical, complete for every evidence document, K-1 coordinates each, with coordinate j = topic j / reference topic K-1, finite strictly-positive variance, and digest binding. Calling build_joint_coordinate_precision merely to recover IDs would recompute a stronger uncertainty object and is explicitly out of scope.

The negative malformed/reordered/detached fixtures are present but not individually claimed as predecessor REDs yet because unknown-field rejection masks them. Before #662 can close, the producer must use a real converged reference fit and a fit-A/artifact-B mismatch must demonstrate that changing the uncertainty projection changes the terminal result identity. #638 remains open through #658 provenance, #605 Membership reconstruction, backend parity/calibration/recovery, compatibility/rollback, exact-head GREEN/review, and immutable release.

#639 has ordinary-forwarded #662 through current source head ce919d0c1c4c25b4fa9e30b1ddc43dfed6244e94: fit-owned document_coordinate_variances are now projected with canonical document UUIDs and explicit ALR numerator/reference coordinates, included in artifact hashing, with an actual converged-fit fixture and fit-A/fit-B terminal-result mismatch coverage. Exact-head hosted gates are still queued, so #662/#638 remain open.

Fresh review then opened #663. The remaining v2 topic coordinates are still bare integer topic indexes: sequence-edge topic_index and uncertainty ALR indexes are not bound to an owner-derived fitted topic basis even though the numerical owner retains topic_term_probabilities. #663 owns a deterministic topic-basis identity/mismatch contract; no LLM topic naming or post-hoc topic reordering is allowed. If basis identity needs immutable vocabulary/source coordinates, it stays dependent on #658 rather than minting a detached caller digest.

Fresh scientific gap from #662 review: #664 tracks the missing fitted ALR location beside released diagonal-Laplace variance. Source progress on #639 current head c582997602daa1f9e05f6c8a3be248f50c8e9c82 adds only the numerical-owner prerequisite tepp.fitted_document_coordinate_summary.v1: admitted document + explicit ALR numerator/reference + fitted location + existing positive diagonal variance, with actual-reference-fit positive coverage and malformed-state refusal. The v2 artifact wire is intentionally unchanged. Release-level projection remains dependent on #663 fitted-topic basis binding and #658/#527 source/vocabulary provenance; no joint covariance, calibration, or source authority is claimed.

Navigation Menu

Repository navigation

release(contract): publish immutable temporal-analysis contract for Orgmetra consumers #638

Cross-repository release gap

Owner boundary

Required released projection

TDD / scientific acceptance

Release acceptance

42 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

release(contract): publish immutable temporal-analysis contract for Orgmetra consumers #638

Description

Cross-repository release gap

Owner boundary

Required released projection

TDD / scientific acceptance

Release acceptance

Activity

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

42 remaining items

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 21, 2026

seonghobae commented on Sep 22, 2026

seonghobae commented on Sep 22, 2026

seonghobae commented on Sep 22, 2026

seonghobae commented on Sep 22, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions