Repository navigation
feat(OMN-20290): a delegated code edit node runs an edit-verify loop over a worktree, each model turn one onex delegate run - #3186
Merged
Conversation
…over a worktree, each model turn one onex delegate run
…nifest covers large repositories, and checks run in the worktree's own environment
…ine-number prefixes and says where its first line is, and check output loses its colour codes
|
| Rule | Description |
|---|---|
| ARCH-TOPIC-001 | No hardcoded Kafka topic strings in handler code |
| ARCH-DI-001 | No event_bus=None bypass |
| ARCH-DI-002 | No direct Handler instantiation outside workflow_runner/adapters |
| ARCH-DI-003 | No reinvented DI containers |
| ARCH-TOPIC-002 | contract.yaml topics follow `onex.{cmd |
ARCH-DI-002: Direct Handler instantiation (warning)\n\n```\nsrc/omnimarket/cli/cli_code_edit.py:246: result = HandlerDelegatedCodeEditOrchestrator(ports).run(request)
src/omnimarket/delegation/rubric/tool_use_record.py:27: verdict = HandlerDelegationRubricCheck().handle(request)\n```\n\nHandler classes should be wired only in workflow_runner.py or adapter modules. Direct instantiation elsewhere bypasses the contract system.\n\n
Static architectural lint — no model inference (OMN-14176).
✅ Hostile Reviewer — PASSEDBlocking findings (critical/error): 0 Gate semantics
Powered by omniintelligence.review_pairing.cli_review — node-based adversarial review via HandlerLlmCliSubprocess (OMN-8524) |
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| loop_run_id = request.correlation_id | ||
| self._ports.claim_loop_receipt(loop_run_id) | ||
| state = _State() | ||
| status = EnumCodeEditStatus.INFRA_ERROR |
| class ProtocolDelegatedCodeEditPorts(Protocol): | ||
| def claim_loop_receipt(self, loop_run_id: str) -> None: | ||
| """RAISE LoopReceiptExistsError when a receipt for this id exists.""" | ||
| ... |
|
|
||
| def write_loop_receipt( | ||
| self, loop_run_id: str, payload: dict[str, object] | ||
| ) -> None: ... |
| self, request: ModelDelegatedCodeEditRequest | ||
| ) -> tuple[tuple[str, int], ...]: | ||
| """Tracked and untracked files of the worktree, with their line counts.""" | ||
| ... |
|
|
||
| def read_file(self, request: ModelDelegatedCodeEditRequest, path: str) -> str: | ||
| """RAISE WorkspacePathError for a path outside the worktree or not a file.""" | ||
| ... |
| self, request: ModelDelegatedCodeEditRequest, path: str, content: str | ||
| ) -> None: | ||
| """RAISE WorkspacePathError for a path outside the worktree.""" | ||
| ... |
|
|
||
| def run_check( | ||
| self, request: ModelDelegatedCodeEditRequest, check: ModelDeclaredCheck | ||
| ) -> ModelCheckResult: ... |
| turn: int, | ||
| ) -> ModelTurnReply: | ||
| """One ``onex delegate`` run; the reply's actions are already parsed.""" | ||
| ... |
|
|
||
| def diff(self, request: ModelDelegatedCodeEditRequest) -> str: | ||
| """The worktree's diff against its HEAD, untracked files included.""" | ||
| ... |
| transcript: dict[str, object], | ||
| ) -> dict[str, object]: | ||
| """The tool_use rubric verdict record of this run's transcript.""" | ||
| ... |
onexbot-occ-writer Bot
added a commit
to OmniNode-ai/onex_change_control
that referenced
this pull request
Oct 1, 2026
…nimarket#3186 The declared checks were executed for real in the OmniNode-ai/omnimarket checkout at the PR head by the product-repo OCC receipt runner, and the results written append-only into this companion. Producer-minted PENDING receipts are superseded by a net-new record; no existing receipt was edited. Ticket: OMN-16859 (AC3b) Onex-Workflow: OCC Receipt Runner Onex-Run: https://github.com/OmniNode-ai/omnimarket/actions/runs/36864577685
jonahgabriel
enabled auto-merge (squash)
October 1, 2026 13:04
jonahgabriel
pushed a commit
to OmniNode-ai/onex_change_control
that referenced
this pull request
Oct 1, 2026
) * evidence(OMN-20290): author OCC companion for OmniNode-ai/omnimarket#3186 OCC companion by node_pr_lifecycle_fix_effect (OMN-13317 F1 / OMN-13990 / OMN-14285). Product PR head df3aa834dbce191434ec5b7807270d9c3156f1bb. * evidence(OMN-20290): self-bind OCC#12143 + rebind contract_sha256 * evidence(OMN-16859): executed test_passes receipts for OmniNode-ai/omnimarket#3186 The declared checks were executed for real in the OmniNode-ai/omnimarket checkout at the PR head by the product-repo OCC receipt runner, and the results written append-only into this companion. Producer-minted PENDING receipts are superseded by a net-new record; no existing receipt was edited. Ticket: OMN-16859 (AC3b) Onex-Workflow: OCC Receipt Runner Onex-Run: https://github.com/OmniNode-ai/omnimarket/actions/runs/36864577685 --------- Co-authored-by: omnimarket-bot <bot@omninode.ai> Co-authored-by: onexbot-occ-writer[bot] <307849072+onexbot-occ-writer[bot]@users.noreply.github.com>
The CI imperative-contract-guard failed on PR 3186 because loop_ports.py imported score() from tool_use_score.py, making that module live due to its sqlite3 usage. We moved score() and RECORD_SCHEMA unchanged to a new I/O-free module tool_use_record.py. Both tool_use_score.py and loop_ports.py now import score from the new module. Local guard run exits 0, 2311 tests pass, and ruff and mypy --strict are clean.
The merge integrates the dev commit 96d3415 which adds a required engine argument to _request and records measured.engine in score(). The conflict in tool_use_score.py is resolved by keeping score() in the I/O-free tool_use_record.py module, which now records the engine as dev does. The recorded_calls_request function takes an optional engine argument and passes it through, while the code edit loop records the model of the last turn that named one as the engine. Verification confirms ruff and mypy strict are clean on 4001 source files, 2599 tests pass with 2 skipped, and the imperative contract guard exits 0.
Release Identity Gate failed on PR head 890d916 because the pyproject version 0.4.273 matched the latest published tag v0.4.273. This commit bumps project.version in pyproject.toml and the omnimarket entry in uv.lock to 0.4.274. scripts/check_release_identity.py --base origin/dev now reports "OK: version 0.4.274 is ahead of latest published 0.4.273."
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
OMN-20290 (epic OMN-20287). This PR follows the operator ruling of 2026-10-01 (RULING row 2026-10-01T11:28:57Z): internal agentic code drafting moves off crush, a second non-canonical script, and onto the delegation workflow customers get.
What this adds
node_delegated_code_edit_orchestrator(ORCHESTRATOR, ports pattern like the delegated test loop). The input is one task plus a git worktree. The output is a verified diff, a loop receipt and a tool_use rubric verdict.onex delegate --task-type code_generation --response-contract <schema>run, so every model call is routed by the routing contract, gated, receipted and projected.view(paged withoffset),ls,grep,write,edit,run_checkandfinish. The tool names match crush's, so the same rubric scores both engines.*stays inside one segment and**crosses segments. Paths that are absolute, contain.., point into.git, or resolve through a symlink are refused.run_checkruns only declared checks, as argv with no shell. There is no shell tool.finishruns every check.accepted, onno_progress(the same check fingerprints over the same diff twice), onbudget_exhaustedat the turn cap, ondelegate_failed(two failed runs in a row), or oninfra_error.onex code-edit run, registered underonex.cli, plus the real ports (omnimarket.delegated_code_edit.loop_ports).--delegate-in-processkeeps them on this machine instead.--caller-laneand--ticketare stamped on every turn.recorded_calls_requestindelegation/rubric/tool_use_transcript.py, so a recorder that already holds typed calls can be scored without a session store.Model before build
omnibase_infraformal/delegated_code_edit_loop/has its own PR. TLC ran on .201 (TLC 2.19, 2 runners, 3 turns).Lab proof (.201 dev lane, deployed-lane delegate turns, worktrees on .201)
Every turn below has a
delegation_eventsrow on the dev-lane projection withcaller_lane=deleg-nodes-fix,ticket_id=OMN-20290andterminal_ok=t. That is 18 rows across the three loops. First attempts went to bothlocal-coder(.201) andlocal-omnipc2-chat(.202) under the routing contract.2d3d4d6408719e89df08d355The
no_progressrun stopped as designed. It also showed a real gap. A view was cut at 6,000 characters, so the model never saw line 145 and its edits failed on text it had guessed. Commit df3aa83 added view paging withoffset, recovery when an edit'sold_stringcarries copied line-number prefixes, and a hint naming the line where the first line of a failed edit occurs. The rerun was then accepted in 3 turns.Local checks
tests/nodes/node_delegated_code_edit_orchestrator,tests/unit/delegated_code_edit,tests/unit/cli/test_cli_code_edit.py.tests/test_market_node_runtime_dogfood.py(11 passed). The node metadata dependency check passes.node_runtime_sweep --import-checkis clean.mypy --strictpasses on the new modules, and ruff is clean.Not in this PR
delegation_step.pycode (OMN-20292) and remote-lane local mode (OMN-20293).lab_crush.pystays until the replay proves the node.Evidence-Ticket: OMN-20290
Evidence-Source: OCC#12143