Replace static model governance with evidenced resources

kilo commented

2026-10-29 01:09:08 +00:00

Member

Summary

Implements D-084: removes the static model catalog, workflow model policy and family-track registry, their runtime loaders and distribution artifacts.
Builds runtime candidate views from explicit source/model grants, discovery and configured resources; enforces exact model/effort identity, source-scoped telemetry, per-resource authorization, compatibility, capacity and scarcity, and rechecks admission. Unknown quality and bounds remain unknown. Scoped administrator context attestations are time-bounded, exact-runtime/model/effort-bound and labeled as administrator-attested, not independently verified.
Uses explicit runtime-inventory v2 decision provenance and preserves historical catalog decisions and pinned MI v1 receipts. Missing MI publication does not revoke independently supported local capability; conflict, withdrawal, stale and untrusted evidence remain fail-closed.
Retains legacy user state, fails closed on old track grants and offers explicit UI migration; refuses legacy profile/file inputs actionably. Updates CLI, REST, MCP, gateway, installed and documentation contracts.

Validation

uv sync --only-dev; make check: 3,090 tests passed, 18 skipped; basedpyright 0 errors/0 warnings.
Standalone uv run basedpyright, git diff --check, staged whitespace.
make package-check with a wheel built from pre-change develop 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e: synthetic user state and legacy track policy retained and fail-closed, wheel/sdist and clean installation free of all three removed artifacts.
Docker image build and offline installed-artifact smoke passed; temporary image removed. Synthetic mobile UI was exercised with no horizontal overflow at 390px.
Deterministic synthetic source execution and admission, exact effort pins, source-scoped quota, MI cases, migration and revocation race are covered. These do not claim live-provider acceptance.

Review And Integration

Independent review #5 APPROVE with no findings for HEAD cd9a74fbd9ce378b59ccc5668aa3e07e16c27bd5 and base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e, in fresh native pr-reviewer session ses_e7a1424b2ffep2G2PGXBweaq81. Dispatch resolved GPT-6.1 Sol/openai/high; backend runtime identity independently unobserved. Terra Medium was owner-requested for implementation; actual reasoning effort is unverified. Earlier review findings #1–4 were remediated and re-reviewed.

Merged to develop as 516c155963c3cd06cb9c3dd74323fe4ddaf1b6ed; integrated tree exactly matches the approved and validated HEAD. No live inference, release or deployment occurred.

Refs #234

## Summary - Implements D-084: removes the static model catalog, workflow model policy and family-track registry, their runtime loaders and distribution artifacts. - Builds runtime candidate views from explicit source/model grants, discovery and configured resources; enforces exact model/effort identity, source-scoped telemetry, per-resource authorization, compatibility, capacity and scarcity, and rechecks admission. Unknown quality and bounds remain unknown. Scoped administrator context attestations are time-bounded, exact-runtime/model/effort-bound and labeled as administrator-attested, not independently verified. - Uses explicit runtime-inventory v2 decision provenance and preserves historical catalog decisions and pinned MI v1 receipts. Missing MI publication does not revoke independently supported local capability; conflict, withdrawal, stale and untrusted evidence remain fail-closed. - Retains legacy user state, fails closed on old track grants and offers explicit UI migration; refuses legacy profile/file inputs actionably. Updates CLI, REST, MCP, gateway, installed and documentation contracts. ## Validation - `uv sync --only-dev`; `make check`: 3,090 tests passed, 18 skipped; basedpyright 0 errors/0 warnings. - Standalone `uv run basedpyright`, `git diff --check`, staged whitespace. - `make package-check` with a wheel built from pre-change develop `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`: synthetic user state and legacy track policy retained and fail-closed, wheel/sdist and clean installation free of all three removed artifacts. - Docker image build and offline installed-artifact smoke passed; temporary image removed. Synthetic mobile UI was exercised with no horizontal overflow at 390px. - Deterministic synthetic source execution and admission, exact effort pins, source-scoped quota, MI cases, migration and revocation race are covered. These do not claim live-provider acceptance. ## Review And Integration Independent review #5 APPROVE with no findings for HEAD `cd9a74fbd9ce378b59ccc5668aa3e07e16c27bd5` and base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`, in fresh native `pr-reviewer` session `ses_e7a1424b2ffep2G2PGXBweaq81`. Dispatch resolved GPT-6.1 Sol/openai/high; backend runtime identity independently unobserved. Terra Medium was owner-requested for implementation; actual reasoning effort is unverified. Earlier review findings #1–4 were remediated and re-reviewed. Merged to `develop` as `516c155963c3cd06cb9c3dd74323fe4ddaf1b6ed`; integrated tree exactly matches the approved and validated HEAD. No live inference, release or deployment occurred. Refs #234

kilo added 1 commit

2026-10-29 01:09:08 +00:00

Replace static model governance with evidenced resources (#234 )

ci / check (pull_request) Failing after 4s

Details

5ae0afe617

kilo referenced this pull request

2026-10-29 01:09:29 +00:00

Replace legacy static model governance with evidenced execution resources #234

kilo commented

2026-10-29 01:09:58 +00:00

Author

Member

Ordinary whole-PR review #1 reserved (default maximum 10; no exception): issue #234, PR #235, HEAD 5ae0afe6175965c3d8a89477d32dad2264054b27, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Owner-requested reviewer: fresh native user-visible pr-reviewer, GPT-6.1 Sol high; model catalog lists GPT-6.1 Sol provider openai, variant high. Verdict and actual session/runtime provenance pending. Source is frozen for review.

Ordinary whole-PR review #1 reserved (default maximum 10; no exception): issue #234, PR #235, HEAD `5ae0afe6175965c3d8a89477d32dad2264054b27`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Owner-requested reviewer: fresh native user-visible `pr-reviewer`, GPT-6.1 Sol high; model catalog lists `GPT-6.1 Sol` provider `openai`, variant `high`. Verdict and actual session/runtime provenance pending. Source is frozen for review.

kilo commented

2026-10-29 01:12:52 +00:00

Author

Member

Review #1 native dispatch: Agent Manager request am-1793236223131-qtskpt created visible local session ses_e7acabb3dffebARlXWwI6D7lfz on the existing frozen checkout. The dispatch resolved GPT-6.1 Sol provider openai, reasoning variant high, requested agent pr-reviewer. Native result and independent runtime identity evidence remain pending; dispatch metadata alone is not runtime verification.

Review #1 native dispatch: Agent Manager request `am-1793236223131-qtskpt` created visible local session `ses_e7acabb3dffebARlXWwI6D7lfz` on the existing frozen checkout. The dispatch resolved `GPT-6.1 Sol` provider `openai`, reasoning variant `high`, requested agent `pr-reviewer`. Native result and independent runtime identity evidence remain pending; dispatch metadata alone is not runtime verification.

kilo commented

2026-10-29 01:21:47 +00:00

Author

Member

Ordinary review #1 outcome: REQUEST_CHANGES at HEAD 5ae0afe6175965c3d8a89477d32dad2264054b27, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Native reviewer session ses_e7acabb3dffebARlXWwI6D7lfz; Agent Manager dispatch resolved GPT-6.1 Sol/openai/high, but backend runtime identity is not independently observed. Findings: P1 server-local provider eligibility is incorrectly applied to unrelated paired-worker sources (control_api.py:969-978); P2 dispatch can inject another source's effort after resource-local selection (routing_core.py:2801-2807, gateway_coordinator.py:2060-2091); P2 Sources UI cannot grant/migrate models or configure required context evidence (control_api.py:1358-1370, server_ui.py:1147-1159). UX impact material. Read-only reviewer inspected complete diff and contracts, verified refs/currentness and diff whitespace; it did not run tests. Review #1 is consumed; remediation and a fresh whole-PR review remain required.

Ordinary review #1 outcome: REQUEST_CHANGES at HEAD `5ae0afe6175965c3d8a89477d32dad2264054b27`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Native reviewer session `ses_e7acabb3dffebARlXWwI6D7lfz`; Agent Manager dispatch resolved GPT-6.1 Sol/openai/high, but backend runtime identity is not independently observed. Findings: P1 server-local provider eligibility is incorrectly applied to unrelated paired-worker sources (`control_api.py:969-978`); P2 dispatch can inject another source's effort after resource-local selection (`routing_core.py:2801-2807`, `gateway_coordinator.py:2060-2091`); P2 Sources UI cannot grant/migrate models or configure required context evidence (`control_api.py:1358-1370`, `server_ui.py:1147-1159`). UX impact material. Read-only reviewer inspected complete diff and contracts, verified refs/currentness and diff whitespace; it did not run tests. Review #1 is consumed; remediation and a fresh whole-PR review remain required.

adrian.tkacz added 1 commit

2026-10-29 02:22:49 +00:00

Scope worker telemetry and preserve target effort (#234 )

ci / check (pull_request) Failing after 4s

Details

5f79eb9d9f

kilo commented

2026-10-29 02:23:11 +00:00

Author

Member

Review #1 remediation pushed as 5f79eb9d9f550b941a7be80dedf4582ebddd4862. Runtime resource routing now uses each resource's own explicitly bound worker telemetry, never unrelated server-local account snapshots or eligibility; source health and strict non-paid assurance remain enforced. Exact selected-target effort is carried through dispatch/audit and preserved by target-derived pins; unverified effort is not inherited from another source. Sources UI now provides explicit grants, legacy-policy confirmation, context attestation/removal, adoption controls and capacity scope management, preserving dormant evidence. Browser checked synthetic desktop/mobile flow, including 390px viewport with no horizontal overflow. Scoped quota simulation rejects ambiguous multi-resource overrides and provider-only replenishment observations are actionably refused in runtime mode.

Post-remediation validation: make check PASS (3,077 tests, 18 skipped; basedpyright 0 errors/0 warnings); standalone uv run basedpyright, git diff --check, staged whitespace and baseline-wheel installed upgrade/package checks PASS. Focused 183-test source/worker/UI suite and synthetic source effort/telemetry isolation tests passed. No live provider execution is claimed. Approval is invalidated by the new HEAD; fresh whole-PR review required.

Review #1 remediation pushed as `5f79eb9d9f550b941a7be80dedf4582ebddd4862`. Runtime resource routing now uses each resource's own explicitly bound worker telemetry, never unrelated server-local account snapshots or eligibility; source health and strict non-paid assurance remain enforced. Exact selected-target effort is carried through dispatch/audit and preserved by target-derived pins; unverified effort is not inherited from another source. Sources UI now provides explicit grants, legacy-policy confirmation, context attestation/removal, adoption controls and capacity scope management, preserving dormant evidence. Browser checked synthetic desktop/mobile flow, including 390px viewport with no horizontal overflow. Scoped quota simulation rejects ambiguous multi-resource overrides and provider-only replenishment observations are actionably refused in runtime mode. Post-remediation validation: `make check` PASS (3,077 tests, 18 skipped; basedpyright 0 errors/0 warnings); standalone `uv run basedpyright`, `git diff --check`, staged whitespace and baseline-wheel installed upgrade/package checks PASS. Focused 183-test source/worker/UI suite and synthetic source effort/telemetry isolation tests passed. No live provider execution is claimed. Approval is invalidated by the new HEAD; fresh whole-PR review required.

kilo commented

2026-10-29 02:23:23 +00:00

Author

Member

Ordinary whole-PR review #2 reserved (2 of default maximum 10): issue #234, PR #235, HEAD 5f79eb9d9f550b941a7be80dedf4582ebddd4862, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Owner-requested reviewer remains fresh native user-visible pr-reviewer, GPT-6.1 Sol high; exact model catalog currently offers GPT-6.1 Sol provider openai, variant high. Verdict and actual execution provenance pending. Source frozen for the full review.

Ordinary whole-PR review #2 reserved (2 of default maximum 10): issue #234, PR #235, HEAD `5f79eb9d9f550b941a7be80dedf4582ebddd4862`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Owner-requested reviewer remains fresh native user-visible `pr-reviewer`, GPT-6.1 Sol high; exact model catalog currently offers `GPT-6.1 Sol` provider `openai`, variant `high`. Verdict and actual execution provenance pending. Source frozen for the full review.

kilo commented

2026-10-29 02:27:09 +00:00

Author

Member

Review #2 native dispatch: Agent Manager request am-1793240616382-h6qieb created fresh visible local session ses_e7a87b212ffeaQQLpnc3bsuDYm on the frozen checkout. Dispatch resolved GPT-6.1 Sol provider openai, variant high, with requested pr-reviewer agent. Native verdict and independent runtime identity evidence remain pending; dispatch metadata is not runtime verification.

Review #2 native dispatch: Agent Manager request `am-1793240616382-h6qieb` created fresh visible local session `ses_e7a87b212ffeaQQLpnc3bsuDYm` on the frozen checkout. Dispatch resolved GPT-6.1 Sol provider `openai`, variant `high`, with requested `pr-reviewer` agent. Native verdict and independent runtime identity evidence remain pending; dispatch metadata is not runtime verification.

kilo commented

2026-10-29 02:32:53 +00:00

Author

Member

Ordinary review #2 outcome: REQUEST_CHANGES at HEAD 5f79eb9d9f550b941a7be80dedf4582ebddd4862, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Native session ses_e7a87b212ffeaQQLpnc3bsuDYm; dispatch resolved GPT-6.1 Sol/openai/high; backend runtime identity independently unobserved. Findings: P1 executable recommendation cannot round-trip an evidenced effort through sr-pin, so producer rejects discovered-model targets (routing_core.py:934-938, gateway_coordinator.py:1642-1646, executable_client.py:42-45); P2 retained preparation rejects runtime route schema 2 (executable_client.py:35-38). Both break the supported select-to-execution path and are material UX regressions. Reviewer inspected complete diff/current refs and passed diff whitespace; its focused test execution was denied by tool permissions, so it claims no test pass. Review #2 consumed; scoped remediation and a fresh full review remain required.

Ordinary review #2 outcome: REQUEST_CHANGES at HEAD `5f79eb9d9f550b941a7be80dedf4582ebddd4862`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Native session `ses_e7a87b212ffeaQQLpnc3bsuDYm`; dispatch resolved GPT-6.1 Sol/openai/high; backend runtime identity independently unobserved. Findings: P1 executable recommendation cannot round-trip an evidenced effort through `sr-pin`, so producer rejects discovered-model targets (`routing_core.py:934-938`, `gateway_coordinator.py:1642-1646`, `executable_client.py:42-45`); P2 retained preparation rejects runtime route schema 2 (`executable_client.py:35-38`). Both break the supported select-to-execution path and are material UX regressions. Reviewer inspected complete diff/current refs and passed diff whitespace; its focused test execution was denied by tool permissions, so it claims no test pass. Review #2 consumed; scoped remediation and a fresh full review remain required.

adrian.tkacz added 1 commit

2026-10-29 02:57:25 +00:00

Preserve executable pin effort and versioned provenance (#234 )

ci / check (pull_request) Failing after 4s

Details

6a17ec8aba

kilo commented

2026-10-29 02:57:50 +00:00

Author

Member

Review #2 remediation pushed as 6a17ec8abafc384903ed0f03964e3e7140e128e1. Runtime executable recommendations now use version-2 envelope/route/retained requirements and an explicit ~effort pin suffix; historical version-1 inputs remain supported without inferring effort from variant. Producer, parser, consumer and exact admission validate the same resource/model/effort and reject inconsistent or unsupported versions. Synthetic offline tests exercise discovered-source select → prepare → admission → dispatch, explicit none versus absent effort, changed/unverified effort refusal, and separately configured resources without inherited effort.

Validation: make check PASS (3,080 tests, 18 skipped; basedpyright 0/0); separate uv run basedpyright, git diff --check, staged whitespace, and prior-develop-wheel installed upgrade/package checks PASS. Focused executable/routing/kernel/gateway suite passed 212 tests. No live provider execution is claimed. Review #2 is invalidated by this HEAD change; a fresh complete review is required.

Review #2 remediation pushed as `6a17ec8abafc384903ed0f03964e3e7140e128e1`. Runtime executable recommendations now use version-2 envelope/route/retained requirements and an explicit `~effort` pin suffix; historical version-1 inputs remain supported without inferring effort from variant. Producer, parser, consumer and exact admission validate the same resource/model/effort and reject inconsistent or unsupported versions. Synthetic offline tests exercise discovered-source select → prepare → admission → dispatch, explicit `none` versus absent effort, changed/unverified effort refusal, and separately configured resources without inherited effort. Validation: `make check` PASS (3,080 tests, 18 skipped; basedpyright 0/0); separate `uv run basedpyright`, `git diff --check`, staged whitespace, and prior-develop-wheel installed upgrade/package checks PASS. Focused executable/routing/kernel/gateway suite passed 212 tests. No live provider execution is claimed. Review #2 is invalidated by this HEAD change; a fresh complete review is required.

kilo commented

2026-10-29 02:58:01 +00:00

Author

Member

Ordinary whole-PR review #3 reserved (3 of default maximum 10): issue #234, PR #235, HEAD 6a17ec8abafc384903ed0f03964e3e7140e128e1, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Owner-requested reviewer: fresh native user-visible pr-reviewer, GPT-6.1 Sol high. Native verdict and independent runtime identity evidence pending. Source is frozen for review.

Ordinary whole-PR review #3 reserved (3 of default maximum 10): issue #234, PR #235, HEAD `6a17ec8abafc384903ed0f03964e3e7140e128e1`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Owner-requested reviewer: fresh native user-visible `pr-reviewer`, GPT-6.1 Sol high. Native verdict and independent runtime identity evidence pending. Source is frozen for review.

kilo commented

2026-10-29 03:01:57 +00:00

Author

Member

Review #3 native dispatch: Agent Manager request am-1793242708565-c3ppgn created fresh visible local session ses_e7a67c586ffeA0NB0h84K7q65Y on the frozen checkout. Dispatch resolved GPT-6.1 Sol provider openai, variant high, with requested pr-reviewer agent. Native verdict and independent runtime identity evidence remain pending; dispatch metadata is not runtime verification.

Review #3 native dispatch: Agent Manager request `am-1793242708565-c3ppgn` created fresh visible local session `ses_e7a67c586ffeA0NB0h84K7q65Y` on the frozen checkout. Dispatch resolved GPT-6.1 Sol provider `openai`, variant `high`, with requested `pr-reviewer` agent. Native verdict and independent runtime identity evidence remain pending; dispatch metadata is not runtime verification.

kilo commented

2026-10-29 03:07:40 +00:00

Author

Member

Ordinary review #3 outcome: REQUEST_CHANGES at HEAD 6a17ec8abafc384903ed0f03964e3e7140e128e1, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Native session ses_e7a67c586ffeA0NB0h84K7q65Y; dispatch resolved GPT-6.1 Sol/openai/high; backend runtime identity independently unobserved. Finding P2: logical effort resolution considers the global catalog before authorization, so a blocked discovered source with an evidenced effort can make an authorized unqualified configured resource reject its advertised logical model (gateway_coordinator.py:1729-1736, _resolve_logical_effort); the client gets an unsatisfiable effort requirement and must discover a resource pin. Reviewer inspected complete frozen diff/current refs; test and whitespace execution were denied by reviewer permissions, so no test pass is claimed. Review #3 consumed; scoped remediation and a fresh full review remain required.

Ordinary review #3 outcome: REQUEST_CHANGES at HEAD `6a17ec8abafc384903ed0f03964e3e7140e128e1`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Native session `ses_e7a67c586ffeA0NB0h84K7q65Y`; dispatch resolved GPT-6.1 Sol/openai/high; backend runtime identity independently unobserved. Finding P2: logical effort resolution considers the global catalog before authorization, so a blocked discovered source with an evidenced effort can make an authorized unqualified configured resource reject its advertised logical model (`gateway_coordinator.py:1729-1736`, `_resolve_logical_effort`); the client gets an unsatisfiable effort requirement and must discover a resource pin. Reviewer inspected complete frozen diff/current refs; test and whitespace execution were denied by reviewer permissions, so no test pass is claimed. Review #3 consumed; scoped remediation and a fresh full review remain required.

adrian.tkacz added 1 commit

2026-10-29 03:33:47 +00:00

Scope logical model resolution to authorized resources (#234 )

ci / check (pull_request) Failing after 4s

Details

38285e04ca

kilo commented

2026-10-29 03:34:06 +00:00

Author

Member

Review #3 remediation pushed as 38285e04caf7211b71c0f77806c8f5540159a10b: logical-model provider and effort resolution now uses the client's current authorized, adapter-backed resource identities and source-specific effort evidence, rather than the global model catalog. A blocked source cannot impose an effort or create a cross-provider ambiguity for an authorized unqualified resource. Synthetic HTTP tests cover advertised metadata, omitted-effort execution, explicit high/none refusal, and blocked-provider non-disclosure; denied plan-lane logical names now correctly return not-found without revealing authorization state.

Validation: make check PASS (3,082 tests, 18 skipped; basedpyright 0/0); standalone uv run basedpyright, git diff --check, staged whitespace and prior-develop-wheel installed upgrade/package checks PASS. Focused 231-test gateway/routing/executable suite and source end-to-end tests passed. No live-provider acceptance is claimed. Review #3 is invalidated by the HEAD change; fresh complete review is required.

Review #3 remediation pushed as `38285e04caf7211b71c0f77806c8f5540159a10b`: logical-model provider and effort resolution now uses the client's current authorized, adapter-backed resource identities and source-specific effort evidence, rather than the global model catalog. A blocked source cannot impose an effort or create a cross-provider ambiguity for an authorized unqualified resource. Synthetic HTTP tests cover advertised metadata, omitted-effort execution, explicit high/none refusal, and blocked-provider non-disclosure; denied plan-lane logical names now correctly return not-found without revealing authorization state. Validation: `make check` PASS (3,082 tests, 18 skipped; basedpyright 0/0); standalone `uv run basedpyright`, `git diff --check`, staged whitespace and prior-develop-wheel installed upgrade/package checks PASS. Focused 231-test gateway/routing/executable suite and source end-to-end tests passed. No live-provider acceptance is claimed. Review #3 is invalidated by the HEAD change; fresh complete review is required.

kilo commented

2026-10-29 03:34:16 +00:00

Author

Member

Ordinary whole-PR review #4 reserved (4 of default maximum 10): issue #234, PR #235, HEAD 38285e04caf7211b71c0f77806c8f5540159a10b, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Owner-requested reviewer: fresh native user-visible pr-reviewer, GPT-6.1 Sol high. Native verdict and independent runtime identity evidence pending. Source frozen for review.

Ordinary whole-PR review #4 reserved (4 of default maximum 10): issue #234, PR #235, HEAD `38285e04caf7211b71c0f77806c8f5540159a10b`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Owner-requested reviewer: fresh native user-visible `pr-reviewer`, GPT-6.1 Sol high. Native verdict and independent runtime identity evidence pending. Source frozen for review.

kilo commented

2026-10-29 03:38:14 +00:00

Author

Member

Review #4 native dispatch: Agent Manager request am-1793244881551-rpcel2 created fresh visible local session ses_e7a469d4fffeadz2uuG6CLKFXt on the frozen checkout. Dispatch resolved GPT-6.1 Sol provider openai, variant high, with requested pr-reviewer agent. Native verdict and independent runtime identity evidence remain pending; dispatch metadata is not runtime verification.

Review #4 native dispatch: Agent Manager request `am-1793244881551-rpcel2` created fresh visible local session `ses_e7a469d4fffeadz2uuG6CLKFXt` on the frozen checkout. Dispatch resolved GPT-6.1 Sol provider `openai`, variant `high`, with requested `pr-reviewer` agent. Native verdict and independent runtime identity evidence remain pending; dispatch metadata is not runtime verification.

kilo commented

2026-10-29 03:43:58 +00:00

Author

Member

Ordinary review #4 outcome: REQUEST_CHANGES at HEAD 38285e04caf7211b71c0f77806c8f5540159a10b, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Native session ses_e7a469d4fffeadz2uuG6CLKFXt; dispatch resolved GPT-6.1 Sol/openai/high; backend runtime identity independently unobserved. Findings: P2 composed machine select/simulate omits live worker continuation capability (control_api.py:982-997), incorrectly refusing supported tools; P2 absent MI publication still introduces MI-derived channel restrictions (routing_core.py:3050-3051), blocking independent local evidence; P2 numeric runtime version matching aliases distinct supported prerelease builds (server_config.py:338-340, execution_sources.py:658-668, worker_codex_adapter.py:165,436-442,1576), allowing an old context attestation after runtime change. Reviewer read complete diff/current refs; attempted final Git checks were denied by reviewer permissions and no reviewer-run tests were claimed. Material UX impact. Review #4 consumed; remediation and fresh complete review required.

Ordinary review #4 outcome: REQUEST_CHANGES at HEAD `38285e04caf7211b71c0f77806c8f5540159a10b`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Native session `ses_e7a469d4fffeadz2uuG6CLKFXt`; dispatch resolved GPT-6.1 Sol/openai/high; backend runtime identity independently unobserved. Findings: P2 composed machine select/simulate omits live worker continuation capability (`control_api.py:982-997`), incorrectly refusing supported tools; P2 absent MI publication still introduces MI-derived channel restrictions (`routing_core.py:3050-3051`), blocking independent local evidence; P2 numeric runtime version matching aliases distinct supported prerelease builds (`server_config.py:338-340`, `execution_sources.py:658-668`, `worker_codex_adapter.py:165,436-442,1576`), allowing an old context attestation after runtime change. Reviewer read complete diff/current refs; attempted final Git checks were denied by reviewer permissions and no reviewer-run tests were claimed. Material UX impact. Review #4 consumed; remediation and fresh complete review required.

adrian.tkacz added 1 commit

2026-10-29 04:28:21 +00:00

Preserve live tool capability and exact runtime evidence (#234 )

ci / check (pull_request) Failing after 4s

Details

cd9a74fbd9

kilo commented

2026-10-29 04:28:46 +00:00

Author

Member

Review #4 remediation pushed as cd9a74fbd9ce378b59ccc5668aa3e07e16c27bd5. Authenticated select/simulate now receive the same fail-closed live continuation-capability set as gateway routing; tests cover protocol-v3 eligibility, protocol-v2 exclusion, missing and failed sources. Missing MI publication no longer introduces MI-derived channel constraints, while independently supplied constraints and adverse/current MI evidence remain enforced. Codex and ZCode now preserve complete bounded runtime version identity separately from numeric compatibility, including prerelease/build suffixes; worker inventory schema 2 marks exact identity, while accepted legacy schema 1 cannot authorize a context attestation. Tests verify prerelease upgrades invalidate old evidence and audit retains the exact identity.

Validation: make check PASS (3,090 tests, 18 skipped; basedpyright 0/0); standalone uv run basedpyright, git diff --check, staged whitespace and prior-develop-wheel installed upgrade/package checks PASS. Focused runtime/MI/tool-capability suites passed; no live-provider execution is claimed. Review #4 requested changes and is consumed; the new HEAD requires a fresh whole-PR review.

Review #4 remediation pushed as `cd9a74fbd9ce378b59ccc5668aa3e07e16c27bd5`. Authenticated select/simulate now receive the same fail-closed live continuation-capability set as gateway routing; tests cover protocol-v3 eligibility, protocol-v2 exclusion, missing and failed sources. Missing MI publication no longer introduces MI-derived channel constraints, while independently supplied constraints and adverse/current MI evidence remain enforced. Codex and ZCode now preserve complete bounded runtime version identity separately from numeric compatibility, including prerelease/build suffixes; worker inventory schema 2 marks exact identity, while accepted legacy schema 1 cannot authorize a context attestation. Tests verify prerelease upgrades invalidate old evidence and audit retains the exact identity. Validation: `make check` PASS (3,090 tests, 18 skipped; basedpyright 0/0); standalone `uv run basedpyright`, `git diff --check`, staged whitespace and prior-develop-wheel installed upgrade/package checks PASS. Focused runtime/MI/tool-capability suites passed; no live-provider execution is claimed. Review #4 requested changes and is consumed; the new HEAD requires a fresh whole-PR review.

kilo commented

2026-10-29 04:29:20 +00:00

Author

Member

Ordinary whole-PR review #5 reserved (5 of default maximum 10): issue #234, PR #235, HEAD cd9a74fbd9ce378b59ccc5668aa3e07e16c27bd5, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Owner-requested reviewer: fresh native user-visible pr-reviewer, GPT-6.1 Sol high. Native verdict and independent runtime identity evidence pending. Source frozen for review.

Ordinary whole-PR review #5 reserved (5 of default maximum 10): issue #234, PR #235, HEAD `cd9a74fbd9ce378b59ccc5668aa3e07e16c27bd5`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Owner-requested reviewer: fresh native user-visible `pr-reviewer`, GPT-6.1 Sol high. Native verdict and independent runtime identity evidence pending. Source frozen for review.

kilo commented

2026-10-29 04:33:23 +00:00

Author

Member

Review #5 native dispatch: Agent Manager request am-1793248189228-6gpnwr created fresh visible local session ses_e7a1424b2ffep2G2PGXBweaq81 on the frozen checkout. Dispatch resolved GPT-6.1 Sol provider openai, variant high, with requested pr-reviewer agent. Native verdict and independent runtime identity evidence remain pending; dispatch metadata is not runtime verification.

Review #5 native dispatch: Agent Manager request `am-1793248189228-6gpnwr` created fresh visible local session `ses_e7a1424b2ffep2G2PGXBweaq81` on the frozen checkout. Dispatch resolved GPT-6.1 Sol provider `openai`, variant `high`, with requested `pr-reviewer` agent. Native verdict and independent runtime identity evidence remain pending; dispatch metadata is not runtime verification.

kilo commented

2026-10-29 04:39:18 +00:00

Author

Member

Ordinary whole-PR review #5: APPROVE, empty findings, at exact HEAD cd9a74fbd9ce378b59ccc5668aa3e07e16c27bd5, base/merge-base 2002d348a8d6d5f3d1b742a93ba3adf944d4b24e. Fresh native read-only reviewer session ses_e7a1424b2ffep2G2PGXBweaq81; dispatch resolved GPT-6.1 Sol/openai/high. Backend runtime identity was not independently observable and is not claimed runtime-verified. Reviewer inspected the complete PR, source contracts, tests and package/upgrade changes, verified exact currentness and git diff --check, and identified no material security or UX finding. It did not run tests; orchestrator full make check passed 3,090 tests with 18 skips, and package/installed-upgrade and type checks passed. Residual risks: synthetic tests are not live-provider acceptance; source context attestations are administrator assertions, not independently verified provider evidence. Approval applies only to the exact HEAD/base pair.

Ordinary whole-PR review #5: **APPROVE**, empty findings, at exact HEAD `cd9a74fbd9ce378b59ccc5668aa3e07e16c27bd5`, base/merge-base `2002d348a8d6d5f3d1b742a93ba3adf944d4b24e`. Fresh native read-only reviewer session `ses_e7a1424b2ffep2G2PGXBweaq81`; dispatch resolved GPT-6.1 Sol/openai/high. Backend runtime identity was not independently observable and is not claimed runtime-verified. Reviewer inspected the complete PR, source contracts, tests and package/upgrade changes, verified exact currentness and `git diff --check`, and identified no material security or UX finding. It did not run tests; orchestrator full `make check` passed 3,090 tests with 18 skips, and package/installed-upgrade and type checks passed. Residual risks: synthetic tests are not live-provider acceptance; source context attestations are administrator assertions, not independently verified provider evidence. Approval applies only to the exact HEAD/base pair.

kilo merged commit 516c155963 into develop

2026-10-29 04:42:08 +00:00

kilo referenced this pull request from a commit

2026-10-29 04:42:08 +00:00

Merge pull request 'Replace static model governance with evidenced resources' (#235) from issue-234-evidenced-resources into develop

kilo referenced this pull request

2026-10-29 04:43:02 +00:00

Replace legacy static model governance with evidenced execution resources #234

Rows
Columns

Replace static model governance with evidenced resources #235

Summary

Validation

Review And Integration