Track TopoMT Python matrix evidence before release

Recorded source checkpoint — Adopt the required Python 3.11–3.14 contract under uibcdf/molsyssuite#51. Current published checkpoint 3fddc2234008ddc8fd6661886443f7dd49818f0c declares >=3.11,<3.15 and calls immutable policy-v1.5.3. Source/full-CI, contributor guidance, applicable recipes and installed-candidate gates are aligned; the measured source snapshot used routine Python 3.13.

How — Source CI 37105640084 at e1d2fee16f1666e8c2b673b229243d8bf422eae0 provides the recorded evidence. Python 3.14 ordinary installation/import and interpreter/architecture checks pass on both platforms. Linux full tests at e1d2fee report 2,003 passes, seven failures, 85 skips and five xfails: four DFND hash mismatches, two fpocket parity failures (missing binary and nested ImportError keyword-argument error), and one packaging assertion. The packaging guard is corrected and locally verified at 3fddc22 using the public requirement parser. Scientific failures remain with the component team; no passing matrix or public 3.14 qualification is claimed. Older cells are not certified by those 3.14 results.

Why — Requirement, tested source and public delivery are different claims. Scientific feasibility remains pending, so TopoMT is not promoted in the suite qualification table. The badge stays unchanged. Existing scientific assertions and older-minor dependency routes are preserved; the new 3.14 lanes use reviewed exact MolSysSuite providers rather than bypassing Requires-Python.

PR/recovery controls — Main retains strict protection with 10 checks including the new 3.14 route; previous checks and administrator direct-push bypass are preserved. Historical three-minor matrices cannot clear four-minor debt. Brief probe 37109925282 passes, retains 57 historical skipped commits without a passing four-minor watermark and decides full recovery is due. Heavy jobs were intentionally omitted by probe mode; this is not an executed recovery-suite claim.

Record — devguide/pending_proposals/track_python_matrix_evidence.md; central machine receipts are devguide/rollouts/python314_remaining_source_adoption.json in uibcdf/molsyssuite.

Acceptance still open — Resolve or explicitly track component-owned qualification blockers, produce applicable immutable public artifacts and independent public clean-install evidence, and align support claims before admission. No public upload is performed by this governance change. MolSysMT/MolSysViewer code and scientific suites are not changed or launched; they are dependency source consumers in the new lane.

Current policy deployment — Published commit a0b8eb5cd3115cb7d134ce368583db5fb5b7d3ad adopts immutable policy-v1.5.4 and the synchronized guide. Policy 37122891754 passes on exact source a0b8eb5cd3115cb7d134ce368583db5fb5b7d3ad with Python 3.14. Routine development is now 3.14; full matrices retain every supported minor. Central receipts: devguide/rollouts/policy154_routine_adoption.json under uibcdf/molsyssuite#39. The earlier 1.5.3 scientific source evidence above remains scoped to its recorded commits and closures; public delivery and remaining scientific qualification stay open.

MolSysSuite adoption checkpoint (2026-09-22): the caller now pins the current immutable release policy-v1.4.5 in commit 8ff065a, superseding the historical policy-v1.1.4 acceptance-text target. Hosted run 35723875941 reached the current policy but failed Check Python lint and imports; GH Run Receptor preserved the failure, and native logs measured 603 Ruff findings under the repository configuration. The pin is current, but adoption is not green. This issue remains the local migration owner and is linked from uibcdf/molsyssuite#34.

Checkpoint (2026-09-24): commit 0fd4cd5 updates the immutable caller to policy-v1.4.10, enables the full MOLI Ruff baseline in .github/workflows/ruff.yaml, and prepares pinned suite sources plus a Python 3.13 environment for the CI matrix. The offline MolSysSuite checker and both local Ruff commands pass; hosted policy run 35990325527 and hosted Ruff run 35990324990 both pass (inspected with gh-run-receptor).

The full local pytest-receptor run on the migration and an isolated run of the previous commit bc0736b produced identical results: 672 passed, 82 failed, 61 skipped, 5 xfailed, in the same nine failure groups. The failures include 73 MolSysViewer addon contract assertions, four DFND snapshot hashes, two missing fpocket executable cases, one missing scikit-image, one missing alphaspace2, and one argument-digestion case. These are preexisting blockers in this environment, not regressions from the Ruff cleanup. Hosted CI run 35990324963 is pending; the issue remains open until the full matrix is green and these blockers are triaged or repaired.

The historical policy-v1.1.4 acceptance target in the opening body is superseded by the published policy-v1.4.10 caller. The original checkout's local topomt/_version.py modification was preserved.

Seguimiento de 0fd4cd5: la CI completa 35990324963 terminó con 6/6 trabajos de pruebas fallidos. Todos llegaron a pytest; Ubuntu registró 81 fallos y 673 aciertos por trabajo, macOS 82 fallos y 672 aciertos. gh-run-receptor agrupó las seis ejecuciones bajo la ausencia de alphaspace2, pero el resumen de pytest muestra también las incompatibilidades con MolSysViewer, las referencias DFND y la ausencia de fpocket. La ejecución local del commit anterior y la del nuevo commit ya habían coincidido en 82 fallos y nueve grupos.

Los gates hospedados de Ruff y política sí pasaron. Mantengo este issue abierto para los problemas de producto; a petición de la coordinación de la suite, se difiere su reparación mientras avanzan los demás miembros. Evidencia duradera: uibcdf/molsyssuite@dbd4653, devguide/pending_proposals/track_ongoing_guide_and_policy_caller_adoption.md.

The full CI matrix remains red on TopoMT 015cb48 with the same cause as the preceding run on ff20727: six test jobs report ModuleNotFoundError: No module named 'alphaspace2'. GH Run Receptor confirmed the matching root cause. The shared MolSysSuite policy and Ruff workflows pass on 015cb48; this comment records the independent dependency gap under this existing CI-adoption issue.

The 2026-10-01 DFND/Topography audit distinguishes current focused local evidence from the release matrix. Its DFND/Topography selection passes 312 tests with two skips at 7bd47faba6179a6e02444331e442acd251572564; that does not close this issue.

The last inspected full scientific CI, run 36785275320, is failed at older source 5f2c0f43c8b0ec4b231641e0df58d1474f651785. gh-run-receptor reports six failed test jobs, one skipped job and one successful job, and identifies FileNotFoundError: fpocket as one bounded cause across the failed group. Its evidence does not establish absence of other failures. This is newer diagnostic evidence than the AlphaSpace2-only cause in the original issue body.

Passing Ruff/policy or a dedicated optional-executable workflow must not substitute for a green full scientific/installed-package matrix. #58 remains the CI-route owner; native development can proceed with explicitly bounded local evidence. The audit is recorded in devguide/DFND/audit_topography_2026_10_01.md.

The #60 input-context delivery is published at
5d7d084129a532b379494f541070504f7e2011a2. Local joint native/Topography checks
pass 420 tests with seven existing skips on Linux Python 3.13.14.
This is not a complete installed Python/OS or optional-engine claim.

An additional feature/atom-label/provider/addon selection reports
249 passed, three failed, six skipped, 205.70 s. All three failures occur
before fpocket execution: the active editable DepDigest reports
0.10.1+15.g78a9106 and its check_dependency signature lacks kind and
executable options. TopoMT declares depdigest>=0.12.0. Executing the exact
published-baseline fpocket runner from aef4b34 reproduces
TypeError: check_dependency() got an unexpected keyword argument 'kind'.
The fpocket runner is unchanged by this delivery. The current environment must
meet the declared minimum before claiming installed fpocket validation; no
fallback or weakened requirement was introduced.

GH Run Receptor now preserves terminal FAIL for baseline CI
36849438075
at aef4b34: one cell fails installing the pinned MolSysViewer revision,
five fail in tests, one governance job passes and one backlog job is skipped.
Bounded causes include the dependency-declaration test and the funnel-motif
test. Native macOS 3.11 log inspection confirms the dependency test compares
bare package names against requirement strings (depdigest>=0.12.0) and
the funnel assertion reports False is True. These are observations requiring
their own diagnosis; the input-context delivery does not relabel them green.

Checkpoint:
devguide/topography_input_context.md.

CI 36874210900, at 3da03db196c3a679af54152a868024eae187aada, failed all six scientific matrix jobs during collection with ImportError: cannot import name 'QuantityRecord' from 'pyunitwizard'. GH Run Receptor identified the failures; bounded native failed logs established this import cause.

The first notebook's new reference test requires the provisional QuantityRecord codec, while the controlled PyUnitWizard source was still 342babd. Local testing used a newer editable source and missed the mismatch. This regression belongs to the notebook addition; it is not a native DFND calculation failure.

Commit 14091ae advances the controlled source to the canonical QuantityRecord feature commit 23554a7aca31cba144ef248b9771d4668815dd1c. Its package sources match the locally exercised revision. An isolated install of that exact source passes the selected 84-test scientific/reporting battery with Pytest Receptor. The notebooks document the provisional codec/development-source requirement; no serialization fallback or test skip was added.

This removes the identified collection mismatch locally. A fresh full matrix is still required before any unqualified supported-matrix claim. Record: devguide/pending_proposals/track_python_matrix_evidence.md.

Subsequent matrix evidence at 88e2164: CI run
https://github.com/uibcdf/topomt/actions/runs/36887103616 failed. GH Run Receptor
and bounded native failed-log inspection distinguish five Conda setup failures
(ENOENT during setup-micromamba shell setup) from the macOS/Python 3.11 pytest
failure in test_funnel_motif_detects_steady_narrowing.

These are separate from the fixed QuantityRecord collection regression.
devguide/pending_proposals/track_python_matrix_evidence.md now retains this
evidence. The local CASTp milestone in b564a85 has passing focused evidence but
does not establish a green repository matrix or a platform support claim.

Completed Ubuntu/Python 3.13 job evidence for CI run 36922512390 at b564a85:
7 failed, 984 passed, 85 skipped, 5 xfailed. The bounded job log identifies four
DFND raw-characterization hash differences, two fpocket comparisons invoking
an unavailable executable, and a dependency-contract assertion finding
depdigest absent from installed metadata.

Other cells were still running at inspection, so this job report is not a final
matrix report. None of the named failing selectors is a CASTp regression.
Evidence is retained in devguide/pending_proposals/track_python_matrix_evidence.md.
The subsequent CASTp radius-profile commit be91657 has separate local evidence:
125 tests passed, with no new skips or expected failures in that selector.

Additional bounded matrix evidence at source 9516716: in CI run 36937886645, the completed Ubuntu/Python 3.13 job 110622317159 reports 7 failed, 1,590 passed, 85 skipped, 5 xfailed. Other cells were still running at inspection.

The seven failures are the same named categories already recorded here: four test_dfnd_raw_characterization_hash assertions, two test_fpocket4_native_recovers_wrapper_pockets_ignoring_order comparisons (1GG0/1N57) requiring the unavailable fpocket executable, and test_pyproject_declares_core_dependencies_and_extras missing depdigest from installed metadata. No CASTp selector failure appears in that completed job.

GH Run Receptor identified failed test jobs but the pending report did not resolve causes; the native completed-job log API supplied the bounded evidence after gh run view --log-failed declined access until whole-run completion. Raw logs remain local.

Maintained report. This evidence does not certify the remaining cells, the subsequent 387f9b4 changes or a passing full support matrix.

Completed matrix evidence at 387f9b4: run 36942562766 fails all six test cells. GH Run Receptor's bounded causes name dependency metadata and DFND morphometrics; native failed logs expose more failures.

In both Ubuntu/macOS Python 3.11 and 3.12, 72 modern CASTp void setup errors and six geometry-policy failures reach n_bonds, then NumPy empty-array dimensionality (all the input arrays must have same number of dimensions). This matches the already corrected MolSysMT #283 boundary. That provider issue is closed with fix 91f157eeb; TopoMT CI still pins 3bcfaf4, which predates the correction. Adopt and validate the corrected controlled source here; do not fork or suppress bond-query failures downstream.

Python 3.13 does not report that CASTp error: Ubuntu has 1,968 passes/seven wider failures; macOS has 1,967 passes/eight failures. DFND, fpocket and dependency metadata still prevent a green matrix. The local #87 forty-system membership comparison has Python 3.13 evidence and does not certify those other cells.

This corrects any inference that the preceding CI run has only the seven failures seen in Ubuntu/Python 3.13. The owning developer-guide record retains the exact matrix/source distinction. No dependency or CI configuration change is included in the bounded membership audit.

docs(dfnd): define the system study and benchmark area [skip ci]

Authorized incremental source-workspace checkpoint: protocol c85549f and evidence/checker 7515faee8b644c6621eccfd35a4237d7392cd38f are pushed with the maintainer-requested [skip ci]. Remote main was verified equal to that head. No hosted CI result or support/installed qualification is claimed.

Affected scope is the independent tetrahedron boundary reference, shared explicit-unit convex-intersection verifier, current raw/public correspondence and BenchLab/reporting docs. Final local evidence: 127 selected tests pass on Python 3.14.10 (11.31 s), repository Ruff passes (556 formatted files), mypy passes for the three affected Python files, and complete Sphinx HTML builds with the same twelve existing warnings. Six download files, figure bytes and navigation are checked. The prior shell certificate replays identically and 161 historical scientific files retain hashes.

Actual runtime source intentionally retains the untouched pre-existing human topomt/_version.py edit (SHA-256 d76f39e77bb54929c2d4a02f55ec04d61534ec09df776d2fe89ef731626a88b5). Its complete source fingerprint is recorded in the new native correspondence; this is not a clean installed-candidate claim. Archived scope/resources/results: devguide/archive/tetrahedron_boundary_connectivity.md.

Deferred evidence: complete suite, Python/support matrices, clean installed/public candidate and hosted exact-head gates remain under this issue and the existing checkpoint recovery route. Recover them at an authorized validation checkpoint; skipped incremental pushes and bounded reference tests do not clear that backlog. Physical refinement remains open under #151; general native scientific/public gates remain #75/#76/#60.

docs(resources): record Zeo++ study candidate [skip ci]

feat(governance): keep findings and discussion actions current [skip ci]

Authorized incremental source checkpoint: b7f4e85e8f04017abf0ca90fc4d9af761f3529a0 is pushed with the maintainer-requested [skip ci], and remote main was verified equal. This is governance/tooling/record work, not scientific-runtime, installed or support-matrix qualification.

Affected local evidence: 31 selected lifecycle/report/public-length-admission tests pass (1.47 s), repository Ruff/format passes (558 files), namespace-aware mypy passes for three Python files, 33 Markdown records/local links and three form sources validate. All 161 prior scientific files retain hashes. Public HTML has the same twelve existing warnings and verified unchanged build source. Live finding/card/index/issue and complete 42-Discussion action routing were independently read back; nine published source blobs match.

devguide/archive/comparative_findings_lifecycle.md records scope/resources/results. Actual runtime source retains the untouched human version-file edit (SHA-256 d76f39e77bb54929c2d4a02f55ec04d61534ec09df776d2fe89ef731626a88b5). No hosted CI success, complete suite, clean installed candidate, full support matrix or newly observed form UI is claimed. Recover the applicable full/exact-head gates at the existing authorized validation checkpoint; this skipped push clears none of that backlog. Ackredit runtime adoption remains #157 and native scientific limits remain their existing owner issues.

feat(attribution): record scientific methods and providers [skip ci]

Qualification handoff from #157 / bcf833f: automatic scientific attribution is now implemented in source, with original/native roles, detached result references, cached reuse, public citation readers and version-2 bundles preserving version-1 scientific identity.

Bounded source evidence: 409 passed in the broad selection and 137 passed in final attribution/persistence/report controls (overlapping selections), real editable Ackredit 0.12.1+10.g4beb850 at clean source 4beb850c8ba814e6103b4d303128a1a36eb57610, Python 3.14.10. See the checkpoint. No rebuilt/installed TopoMT candidate or public package was qualified.

Remaining gates owned here include the full repository suite, actual Ackredit-present coverage across Python 3.11–3.14, original-engine/version receiving, same-file installed/public artifact checks and applicable remote CI. Optional-provider skips do not establish real-provider coverage; ensure the chosen receiving/CI inputs actually provide Ackredit. The provider-documented portable floor is 0.9.0, but older-provider receiving and a public TopoMT extra were not qualified here. Actual scikit-image meshing and scikit-learn neighbor kernels were unavailable in this source environment; explicit-neighbor and unselected-backend controls passed.

Delivery intentionally used the maintainer-authorized [skip ci] route. No skipped workflow is represented as a successful matrix/checkpoint.

Fresh consumer evidence from OpenCASTp's2026-11-10 halfway qualification (uibcdf/opencastp#13): the source-wide run again emits SyntaxWarning at topomt/_private/atom_label.py:11 for the invalid escape \ in format_to_regex's normal triple-quoted docstring. Python3.14.9 strict compile reproduces SyntaxError without importing TopoMT or executing molecular code: warnings.simplefilter('error', SyntaxWarning); compile(path.read_text(), str(path), 'exec'). Inspected source commit0f82d5adfae73ae24846f713ab5bf62782a4f1ad, fileSHA2564d5bbb74781d8bab3d038b47ae0ab8d8f14f644a55c2731baf6144d8dc9d863c. A raw docstring or escaped backslash can preserve the regex example. This is source hygiene/strict-warning compatibility, not an observed numerical mismatch. Existing checkout changes were preserved; no sibling source edit or new duplicate issue.

Additional source-identity evidence from the bounded BenchLab precision delivery: a clean git archive of 834bdc105b5d06727325cb91aa16edce67e0078d imports TopoMT from that temporary source copy and passes strict Sphinx/local navigation (86 HTML pages, 7,471 targets, zero warnings). Its committed topomt/_version.py SHA-256 is c63a97ae89338cff1d0a778054f8bfbcaea51ea974b281d062073f2eba5b2f1a, while the observed cached package version is 1.0.0+70.g303bd33.dirty. These are retained separately in devguide/evidence/volume_documentation_receiving_v1.json.

A cached version string alone therefore does not identify the current scientific source; exact-candidate version/import checks remain necessary in this issue's qualification route. The human worktree version file was preserved unchanged. This is source documentation/identity evidence, not an installed package, public release version or passing support matrix; no artifact was uploaded.

Source checkpoint for the new #178 straight-tube reader and #179 typed-query gateway fix: scientific runtime source b07a20e; final reader/source d62d4f3. The selected tube/reference/physical/public/query/book/report/finding checks pass 106 tests with three strict pending scientific targets (#180 twice, #181 once); the final presentation/report guards also pass. Ruff passes globally; scoped mypy passes four changed source modules. Exact full-precision input, independent references, raw/public observations and real Ackredit attribution are retained.

Both final notebooks run in separate fresh kernels inside the checkout and from the exact 362-file ZIP extracted in /tmp with clean committed TopoMT imports. Strict clean-source Sphinx passes without warnings (96 pages / 8,975 local links), and actual Chrome review led to semantic HTML tables and expandable inputs. Portable receipt v2 · Documentation receipt v2.

Incremental commits use [skip ci]; the final Pages checkpoint is 42947682435. Full-suite, installed-package, interpreter/platform and release qualification remain deferred under this issue; neither these focused checks nor skipped CI qualify them. No scientific operator/default change or original-provider execution is delivered.

Naming migration #182 rechecks the four known whole-record characterization failures. A clean git archive of pre-migration source 87af3c8 reproduces all four mismatches. The migrated worktree yields exactly the same complete canonical raw JSON SHA-256 for all four controls, and all 24 scientific DFND source modules are byte-identical. The two renamed PDB inputs retain their exact original SHA-256 and generator coordinates/radii. Evidence: devguide/evidence/straight_tube_raw_preservation_v1.json (published with #182).

Thus these are existing stale characterization fingerprints, not a geometry or solver change caused by renaming. Their disposition remains owned here; #182 does not replace expected hashes or claim full-suite/matrix qualification.

Additional bounded #182 receiving: 67 focused catalog/centerline/lineage/tube/reference/notebook/book/governance checks pass, with the same three explicitly pending tube targets under #180/#181. Seven affected viewer tests also pass. The expanded 149-item selector and a later 58-item viewer group hit our explicit 120 CPU-second ceiling before completion; they are incomplete evidence, not an assertion of algorithm/test correctness. Inputs/source geometry are byte-identical before and after naming, and the four known stale raw fingerprints reproduce identically on clean pre-migration 87af3c8.

We retain the bounded results and resource-stop logs, rather than increasing parallelism or treating partial progress as a passing full GUI/source suite. Broader GUI/full-suite/runtime qualification remains here. This does not block the independently verified naming/resource migration or establish an installed/support matrix.

Bounded #183 update: wide-tube independent/native receiving and clean committed-source notebooks passed inside/outside the checkout (65 selected tests passed; five scientific targets remain strict expected failures under #180/#181/#184). This is not full-suite/GUI/installed/platform qualification. Source 3ad9515 also fixes new BenchLab quantity-aware serialization (#185); the first new wide attempt is retained and the corrected capture round-trips attribution exactly. Historical saved bibliography created with the old encoder is not retroactively qualified: review/reconstruct those metadata from original retained identities/offline bibliographic declarations before treating it as portable citation evidence, without silently rewriting frozen bytes. Startup Pint-cache and pre-existing atom-label SyntaxWarnings remain visible in clean-source reader outputs; actual imported module/source and producer versions are reported separately. Scope/receipts: devguide/evidence/wide_tube_local_validation_v1.json, wide_tube_portable_v1.json, and devguide/archive/benchlab_metadata_encoding.md. Original scientific operators were unchanged.

Bounded API source checkpoint — 2026-11-11

Delivered #171 → #174 → #77 in four authorized direct commits, all [skip ci], through 66e9c53. The exact executable/public-doc candidate is 3d72725991c192103501b2f7c538e345fb04e09e; the final commit changes only devguide records/navigation/evidence.

Selected 16-module Python 3.14.10 source panel: 209 passed, six existing strict xfails, 57.528 s process wall, peak RSS 523,956 KiB. The retained xfails belong to Resolve bounded physical boundary receiving on the frozen straight tube #180/Reconcile tube local DFND candidates with an independently empty all-atom carrier #181/Resolve wide-tube physical boundary ordering within the numerical margin #184; no scientific finding is resolved by this API batch. Real Ackredit guards and all six BenchLab public controls are included.
Five changed source/consumer files pass scoped mypy 2.4.1 for Python 3.11 and 3.14 static targets. Three public Topography source doctests and two final reporting/index guards pass. Remaining native internal typing is independently open under Qualify bounded typing for native DFND component, data and mesh objects #189; CASTp typing remains Correct existing typing errors in classical and modern CASTp geometry #86.
Repository Ruff lint and format pass (612 files). A fresh clean Git archive builds with strict Sphinx, one worker: 108 pages / 10,475 checked local links, 18.57 s wall, peak RSS 309,944 KiB. Notebook execution stays off. The existing Python invalid-escape SyntaxWarning in atom_label.py:11 remains separate from Sphinx's zero-warning result.
All 385 non-cache BenchLab files and six ZIPs retain inventory and SHA-256 bytes. The pre-existing local _version.py edit is preserved unstaged at SHA-256 d76f39e77bb54929c2d4a02f55ec04d61534ec09df776d2fe89ef731626a88b5. The committed/archive version identity is recorded separately; Adopt the Python distribution policy and guarded noarch publication route #78 retains its distribution/version issue.
One CPU/numerical thread, 6 GiB address-space and 120 CPU-second ceilings; memory availability stayed above 55 million KiB with zero memory PSI and no sampled active swap traffic. No Slurm allocation or exclusive reservation was claimed.

Durable source/environment/limits receipt. #171/#174/#77 are closed with archived reports and current summaries; #189 owns the additional native evidence formerly appended to #77.

Deferred evidence remains owned here and under #58/#78: no new full suite, remote full-CI checkpoint, installed distribution, Python/OS matrix, release/package or website deployment was executed. The next authorized full-CI/installed checkpoint must inspect its exact source/artifact and dependency closure; this local selection does not clear the previously reported full-suite backlog. BenchLab's separate scientific joint review remains #187.

Source checkpoint a32f162 publishes the bounded A–F reference-validator continuation. Relevant selections pass 312 distinct tests (309 source controls plus 29 receiving/governance checks, with overlap), repository Ruff/format and mypy on 38 modules targeting Python 3.11 syntax. Runtime receiving uses Python 3.14 in the existing development environment.

Strict local Sphinx passes with notebook execution off, verified report anchors and exact capsule download bytes. Fresh extracted saved-proof checking succeeds without TopoMT/NumPy/SciPy/quantity/Ackredit/producer modules. All 355 original frozen files remain unchanged. The capsule, source proofs, resources and unresolved original positive-volume precision are versioned.

Commits carry [skip ci] under the maintainer checkpoint route. These checks do not qualify the full suite, remote CI, wheel/Conda installation, supported Python/platform matrix or hosted publication. Those delivery/support gates remain open here; native policy and original-hull precision keep #181/#76 ownership. No release/support claim is added.

Navigation Menu

Repository navigation

Track TopoMT Python matrix evidence before release #16

7 remaining items

Bounded API source checkpoint — 2026-11-11

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Track TopoMT Python matrix evidence before release #16

Description

Activity

LMMV commented on Sep 22, 2026

LMMV commented on Sep 24, 2026

LMMV commented on Sep 24, 2026

LMMV commented on Sep 27, 2026

dprada commented on Oct 1, 2026

dprada commented on Oct 1, 2026

dprada commented on Oct 1, 2026

dprada commented on Oct 1, 2026

dprada commented on Oct 1, 2026

dprada commented on Oct 1, 2026

dprada commented on Oct 2, 2026

7 remaining items

LMMV commented on Nov 9, 2026

LMMV commented on Nov 9, 2026

LMMV commented on Nov 9, 2026

dprada commented on Nov 10, 2026

LMMV commented on Nov 11, 2026

LMMV commented on Nov 11, 2026

LMMV commented on Nov 11, 2026

LMMV commented on Nov 11, 2026

LMMV commented on Nov 11, 2026

LMMV commented on Nov 11, 2026

Bounded API source checkpoint — 2026-11-11

LMMV commented on Nov 15, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions