Creatidy/creatidy-kernel
5
0
Fork 0
Code Issues 8 Pull requests 1 Projects Releases Packages Wiki Activity Actions 1

D1-03b: Promote bounded task execution and quota-free preflight (#37) #38

Merged
adrian.tkacz merged 1 commit from feat/37-durable-task-execution into develop 2026-09-30 09:30:52 +00:00
Conversation 3 Commits 1 Files changed 12 +1666 -368
kilo commented 2026-09-29 23:09:40 +00:00
Member
Copy link

Status And Coordination

REVIEW_INFRASTRUCTURE_BLOCKED, not READY_FOR_OWNER_MERGE. Implementation and exact-head CI are green; the required fresh routed independent review has no eligible exact owner-requested GPT-6.1 Sol High identity. No reviewer substitution, review verdict or merge is claimed.

Refs #37 (Issue A), parent #30 and blocked first-live preflight #34. Separate Issue B: Infrastructure/creatidy-onprem#43. Onprem implementation remains deferred until explicit owner confirmation that this Kernel PR merged.

Base: c8f105e2c0
Frozen HEAD: fcf3f8df7c
Branch: feat/37-durable-task-execution -> develop
Implementation source/test worker: GPT-6.1 Sol High.

Changes

  • Neutral task_execution module, TaskSpec/LiveTaskComponents/checker/collector/run/status/export names and durable operation identifiers; no experimental-state aliases or migrations.
  • Public task preflight/run/status/export CLI, registry-backed task 143 only; old command removed and legacy environment fields rejected without fallback. No arbitrary issue ingestion, workflow DSL, scheduler, model registry or multi-agent framework.
  • One immutable TaskRuntimeConfig parser for CREATIDY_KERNEL_* Router, binding, Codex, Forge and optional source/state configuration. Infrastructure remains separate from TaskSpec authority. Validate origins/API path, absolute paths/version/exact binding; redact credentials and freeze closed operational environments for Codex/schema and verification.
  • Quota-free human/JSON preflight emits schema-1 READY_FOR_LIVE_TASK/BLOCKED, typed blockers, exact_base_sha and explicit no-inference/no-Forge-write/no-execution-approval fields. It checks source/base/allowed paths/frozen structural assumptions, bounded command shapes and executable readiness, generates native Codex schemas and initializes only, constrains Router selection to controller binding, and performs exact Forge repository/base/frozen-issue reads.
  • Preserve explicit approval/trusted-development, controller-owned Allocation/Attempt/candidate/verification, Git-clean/anti-weakening gates, exact accepted subject, AGit delivery and recovery. Logical task identity excludes optional base selector; the original approved pin remains independently durable across admission crashes, and conflicts reject.
  • Read-only preflight shares native SQLite topology validation without writable-store startup; unreconciled sidecars block instead of repair. Credential-shaped Forge SHA echoes cannot enter evidence.
  • Developer Makefile defaults to concise help with original gates unchanged. Installed-package checks cover all four task helps. README documents bounded authority, neutral profile/JSON contract and rollout restrictions.

Validation

  • make check: PASS, 869 tests; locked tooling, formatting/lint, basedpyright 0 errors, inward import contract kept, secret scan no findings, git diff --check clean.
  • make package-check: PASS, wheel/sdist rebuild equivalence, metadata/license, isolated installation and help/offline replay/export.
  • make audit: PASS, no known vulnerabilities; no dependency/lock changes.
  • Focused final preflight suite: 63 passed, including credential-echo rejection. D1 execution/recovery/native adapter regressions pass.
  • Exact-head CI run 2115 / 71: SUCCESS for fcf3f8df7c.
  • Final Git worktree clean, one intended issue commit.

All task Runtime/Router/Forge execution tests used synthetic producer-shaped fixtures. No real Kernel provider/task inference, thread/start or turn/start, Scarcity Router #143 implementation/task PR, deployment or merge occurred.

Review Ledger

max_remediation_cycles: 10
remediation_cycles_used: 0
exact_reviewed_head: none
dispatch_verified: false
independent_review_verdict: NOT_RUN / REVIEW_INFRASTRUCTURE_BLOCKED

Fresh frozen-head dispatch evidence: repository_review + monotone L4, reasoning/coding/tool_use=5, required tool/reasoning mode, exact openai/gpt-6.1-sol/high. Valid selection returned selected=null, no_eligible_candidate (catalog v4, 2026-09-24; profile policy v8). GPT-6 Sol is not an alias. No floors were lowered, no alternatives tried and no reviewer dispatched. Exact model eligibility must be repaired before a fresh independent whole-PR review can proceed.

Required review focus: no engine expansion or durable experimental naming; config/domain authority separation; inference-free/read-only readiness; secret redaction/isolation; controller-owned binding/effort including null != literal none; original pin recovery; all D1 execution/exact-subject invariants; stable onprem contract.

DEFER And Boundaries

Preflight does not attest authenticated model execution, Forge write grants or hostile-worker isolation; it runs no candidate verification campaign or state repair. Existing task semantic correctness/novel weakening/coverage equivalence remain expressly deferred to independent review of the later real task candidate. SOPS and operator Make UX belong only to #43. Both owner merges and fresh #34 readiness through durable operator surfaces remain prerequisites to a separately owner-authorized first live execution. This PR and issue stay open/unmerged.

## Status And Coordination **REVIEW_INFRASTRUCTURE_BLOCKED**, not READY_FOR_OWNER_MERGE. Implementation and exact-head CI are green; the required fresh routed independent review has no eligible exact owner-requested GPT-6.1 Sol High identity. No reviewer substitution, review verdict or merge is claimed. Refs #37 (Issue A), parent #30 and blocked first-live preflight #34. Separate Issue B: [Infrastructure/creatidy-onprem#43](https://forgejo.creatidy.com/Infrastructure/creatidy-onprem/issues/43). Onprem implementation remains deferred until explicit owner confirmation that this Kernel PR merged. Base: c8f105e2c03d96ee99bb9c0cca1791798cccdc4b Frozen HEAD: fcf3f8df7c784e7d42a53be002393206e61cea78 Branch: feat/37-durable-task-execution -> develop Implementation source/test worker: GPT-6.1 Sol High. ## Changes - Neutral task_execution module, TaskSpec/LiveTaskComponents/checker/collector/run/status/export names and durable operation identifiers; no experimental-state aliases or migrations. - Public task preflight/run/status/export CLI, registry-backed task 143 only; old command removed and legacy environment fields rejected without fallback. No arbitrary issue ingestion, workflow DSL, scheduler, model registry or multi-agent framework. - One immutable TaskRuntimeConfig parser for CREATIDY_KERNEL_* Router, binding, Codex, Forge and optional source/state configuration. Infrastructure remains separate from TaskSpec authority. Validate origins/API path, absolute paths/version/exact binding; redact credentials and freeze closed operational environments for Codex/schema and verification. - Quota-free human/JSON preflight emits schema-1 READY_FOR_LIVE_TASK/BLOCKED, typed blockers, exact_base_sha and explicit no-inference/no-Forge-write/no-execution-approval fields. It checks source/base/allowed paths/frozen structural assumptions, bounded command shapes and executable readiness, generates native Codex schemas and initializes only, constrains Router selection to controller binding, and performs exact Forge repository/base/frozen-issue reads. - Preserve explicit approval/trusted-development, controller-owned Allocation/Attempt/candidate/verification, Git-clean/anti-weakening gates, exact accepted subject, AGit delivery and recovery. Logical task identity excludes optional base selector; the original approved pin remains independently durable across admission crashes, and conflicts reject. - Read-only preflight shares native SQLite topology validation without writable-store startup; unreconciled sidecars block instead of repair. Credential-shaped Forge SHA echoes cannot enter evidence. - Developer Makefile defaults to concise help with original gates unchanged. Installed-package checks cover all four task helps. README documents bounded authority, neutral profile/JSON contract and rollout restrictions. ## Validation - make check: PASS, 869 tests; locked tooling, formatting/lint, basedpyright 0 errors, inward import contract kept, secret scan no findings, git diff --check clean. - make package-check: PASS, wheel/sdist rebuild equivalence, metadata/license, isolated installation and help/offline replay/export. - make audit: PASS, no known vulnerabilities; no dependency/lock changes. - Focused final preflight suite: 63 passed, including credential-echo rejection. D1 execution/recovery/native adapter regressions pass. - [Exact-head CI run 2115 / 71](https://forgejo.creatidy.com/Creatidy/creatidy-kernel/actions/runs/71): SUCCESS for fcf3f8df7c784e7d42a53be002393206e61cea78. - Final Git worktree clean, one intended issue commit. All task Runtime/Router/Forge execution tests used synthetic producer-shaped fixtures. **No real Kernel provider/task inference, thread/start or turn/start, Scarcity Router #143 implementation/task PR, deployment or merge occurred.** ## Review Ledger max_remediation_cycles: 10 remediation_cycles_used: 0 exact_reviewed_head: none dispatch_verified: false independent_review_verdict: NOT_RUN / REVIEW_INFRASTRUCTURE_BLOCKED [Fresh frozen-head dispatch evidence](https://forgejo.creatidy.com/Creatidy/creatidy-kernel/pulls/38#issuecomment-13116): repository_review + monotone L4, reasoning/coding/tool_use=5, required tool/reasoning mode, exact openai/gpt-6.1-sol/high. Valid selection returned selected=null, no_eligible_candidate (catalog v4, 2026-09-24; profile policy v8). GPT-6 Sol is not an alias. No floors were lowered, no alternatives tried and no reviewer dispatched. Exact model eligibility must be repaired before a fresh independent whole-PR review can proceed. Required review focus: no engine expansion or durable experimental naming; config/domain authority separation; inference-free/read-only readiness; secret redaction/isolation; controller-owned binding/effort including null != literal none; original pin recovery; all D1 execution/exact-subject invariants; stable onprem contract. ## DEFER And Boundaries Preflight does not attest authenticated model execution, Forge write grants or hostile-worker isolation; it runs no candidate verification campaign or state repair. Existing task semantic correctness/novel weakening/coverage equivalence remain expressly deferred to independent review of the later real task candidate. SOPS and operator Make UX belong only to #43. Both owner merges and fresh #34 readiness through durable operator surfaces remain prerequisites to a separately owner-authorized first live execution. This PR and issue stay open/unmerged.
feat: promote bounded task execution (D1-03b, #37)
All checks were successful
ci / check (pull_request) Successful in 59s
Details
fcf3f8df7c
kilo commented 2026-09-29 23:10:39 +00:00
Author
Member
Copy link

Frozen-Head Review Ledger

Status: REVIEW_INFRASTRUCTURE_BLOCKED (not READY_FOR_OWNER_MERGE).
PR: #38; linked issue #37; base c8f105e2c0; frozen HEAD fcf3f8df7c.
max_remediation_cycles: 10; remediation_cycles_used: 0.

Fresh dispatch boundary at 2026-09-29T23:10:04.197Z: scarcity_select(profile_id=repository_review), monotone L4 / reasoning=5,coding=5,tool_use=5 / requires_tool_use=true,requires_reasoning_mode=true, exact owner identity required_provider=openai,required_model={provider:openai,model:gpt-6.1-sol},required_variant=high. Requirement validation succeeded. Router schema 1, catalog v4 (2026-09-24), profile policy v8, resource policy v1 returned selected=null, degraded=false, reason_codes=[no_eligible_candidate]. No quota snapshots or credentials are reproduced here.

The catalog lacks the exact owner-requested openai/gpt-6.1-sol/high identity; openai/gpt-6-sol/high is a distinct identity and additionally reports unknown tool-use support. No alias, lower floor, alternative candidate, or reviewer substitution was used. No reviewer session was dispatched, no independent verdict exists, dispatch_verified=false, exact reviewed HEAD=none. This is a routing infrastructure blocker, not a substantive review or remediation cycle. Agent Manager offers the requested model, but harness availability cannot replace required Router eligibility. The Router catalog/eligibility must support the exact identity before the required frozen-head review can proceed.

Local gates passed: make check (869 tests, typing/imports/secret scan/diff clean), make package-check, make audit. Exact-head CI is being confirmed separately. Issues/PR stay open; no merge or first-live permission is claimed. Onprem #43 remains deferred until explicit owner confirmation of Kernel merge. No real Kernel task/provider inference, thread/start or turn/start, Scarcity Router #143 changes/task PR, deployment or merge occurred.

## Frozen-Head Review Ledger Status: REVIEW_INFRASTRUCTURE_BLOCKED (not READY_FOR_OWNER_MERGE). PR: #38; linked issue #37; base c8f105e2c03d96ee99bb9c0cca1791798cccdc4b; frozen HEAD fcf3f8df7c784e7d42a53be002393206e61cea78. max_remediation_cycles: 10; remediation_cycles_used: 0. Fresh dispatch boundary at 2026-09-29T23:10:04.197Z: scarcity_select(profile_id=repository_review), monotone L4 / reasoning=5,coding=5,tool_use=5 / requires_tool_use=true,requires_reasoning_mode=true, exact owner identity required_provider=openai,required_model={provider:openai,model:gpt-6.1-sol},required_variant=high. Requirement validation succeeded. Router schema 1, catalog v4 (2026-09-24), profile policy v8, resource policy v1 returned selected=null, degraded=false, reason_codes=[no_eligible_candidate]. No quota snapshots or credentials are reproduced here. The catalog lacks the exact owner-requested openai/gpt-6.1-sol/high identity; openai/gpt-6-sol/high is a distinct identity and additionally reports unknown tool-use support. No alias, lower floor, alternative candidate, or reviewer substitution was used. No reviewer session was dispatched, no independent verdict exists, dispatch_verified=false, exact reviewed HEAD=none. This is a routing infrastructure blocker, not a substantive review or remediation cycle. Agent Manager offers the requested model, but harness availability cannot replace required Router eligibility. The Router catalog/eligibility must support the exact identity before the required frozen-head review can proceed. Local gates passed: make check (869 tests, typing/imports/secret scan/diff clean), make package-check, make audit. Exact-head CI is being confirmed separately. Issues/PR stay open; no merge or first-live permission is claimed. Onprem #43 remains deferred until explicit owner confirmation of Kernel merge. No real Kernel task/provider inference, thread/start or turn/start, Scarcity Router #143 changes/task PR, deployment or merge occurred.
kilo commented 2026-09-29 23:13:02 +00:00
Author
Member
Copy link

Exact-Head CI Confirmed

CI run #2115 / run 71 succeeded for pull_request #38 with exact commit SHA fcf3f8df7c (started 23:09:41Z, completed 23:10:41Z on 2026-09-29). Local full gates are also green, including 869 tests. Final local worktree is clean and contains one intended issue commit.

This resolves the earlier pending/running CI observations, not the independent review blocker. The review ledger remains REVIEW_INFRASTRUCTURE_BLOCKED: no eligible exact owner-required GPT-6.1 Sol High route, no reviewer dispatch/verification/verdict, no reviewed HEAD. No READY_FOR_OWNER_MERGE, integration, onprem implementation or live inference is claimed. Remedy must retain exact identity/floors and perform a fresh frozen-head review; do not alias GPT-6 Sol or waive review.

## Exact-Head CI Confirmed [CI run #2115 / run 71](https://forgejo.creatidy.com/Creatidy/creatidy-kernel/actions/runs/71) succeeded for pull_request #38 with exact commit SHA fcf3f8df7c784e7d42a53be002393206e61cea78 (started 23:09:41Z, completed 23:10:41Z on 2026-09-29). Local full gates are also green, including 869 tests. Final local worktree is clean and contains one intended issue commit. This resolves the earlier pending/running CI observations, not the independent review blocker. The [review ledger](https://forgejo.creatidy.com/Creatidy/creatidy-kernel/pulls/38#issuecomment-13116) remains REVIEW_INFRASTRUCTURE_BLOCKED: no eligible exact owner-required GPT-6.1 Sol High route, no reviewer dispatch/verification/verdict, no reviewed HEAD. No READY_FOR_OWNER_MERGE, integration, onprem implementation or live inference is claimed. Remedy must retain exact identity/floors and perform a fresh frozen-head review; do not alias GPT-6 Sol or waive review.
adrian.tkacz deleted branch feat/37-durable-task-execution 2026-09-30 09:30:53 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
adrian.tkacz
Labels
Clear labels   
bug
Something is not working

  
duplicate
This issue or pull request already exists

  
enhancement
New feature

  
help wanted
Need some help

  
invalid
Something is wrong

  
question
More information is needed

  
wontfix
This won't be fixed

No labels
bug
duplicate
enhancement
help wanted
invalid
question
wontfix
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
Creatidy/creatidy-kernel!38
No description provided.