[erikleblansch]

Scope

Integrates the OpenCode native CPU candidate with main, retaining main's AGENTS.md and Dependabot settings and the reviewed task-holder correction from PR #18. Current Code head 32d7a08 adds only reviewed README/evidence updates after tested executable fd4ff96 and explicitly pins core 7d0f9f0b492c5e692cc9d26f23d95f55f2c1e651 (VOLPAROSSA/volparossa#266).

The latest integration changes only the native core SHA and its exact test expectation, plus current README/evidence text. It preserves older Torch pins/encodings, model, ordinary tools, permissions, resources, deadlines, closed export schemas and workflow behavior. Within the native execution stack, the core adds adaptive complete-prompt prefill checkpoints; native C++, build/conversion, IPC and supervisor/resource code are unchanged. The core also contains the separately checked native/Storage/main source join, not only the prefill change.

Evidence and remaining work

• Independent review and 88 focused Code cases pass with one explicit optional patched-loader fixture skip. Actual pinned core imports/provenance and CPU-overlay checks pass. Core has 116 passing inert native scheduling/contracts cases, including 18 new prefill cases.
• Original hosted Code source run 38823962068 passes 533 unique cases with five explicit optional-fixture skips and no failures. Exact checkout tree 6d87cd093bc76bf94e20662e3dfd9bafab591300 matches fd4ff96. CodeQL 38823957344 passes all three language jobs. These checks do not establish actual model speed or coding completion.
• Original actual 38824828727 PASSES on exact fd4ff96/core7d0f9f0b with qwen3-4b-instruct-2507-v1 and llama_cpp_bf16_v1. Original API, source/workflow records, ZIP digests/CRCs, all 35 packed Code files, complete core archive, native wrappers and input manifest were independently audited. Five real requests produce two reads, an approved edit, an approved bash test and a final response; independent three-case checker and unchanged original tests pass. Task resources, unchanged inputs, graceful core stop, session/project/guest/VM cleanup and observed host route/DNS restoration pass. Task duration 2219455 ms; no general quality, competitive speed, per-ACK latency/count or full-firewall inventory claim.
• Original artifact SHA-256 abf57ddb609c631746d1d9bcd532a17e73c52d5a321f096eaef3e6a713bf3abe; log SHA-256 18098c557bb8a56caa78c2a7823c9007bd71474e85f687595c8339d944943ab4. The legacy hardcoded dirty-source label is not a measured status; actual source guards and independent hashes establish the inputs.
• Original actual 38806073909 completed two reads, an approved edit and test, unchanged original tests and an independent check; whole-trial final checks still failed.
• Original actual 38816006720, exact Code 0031c0e/core 6d65bb12, retains real task status and memory but fails on compute_control_ack_deadline after two reads and an edit. No test command completes. Eight controls were issued/seven acknowledged; no exact native batch duration or underlying cause is claimed.
• The new core starts prefill conservatively and adapts synchronized batches without dropping tokens or counting a real owner pause as decode cost. The ten-second guard remains; a stalled single-token operation still fails closed. Different batching can affect floating-point output.

Publish this accepted Code slice with the unchanged immutable public core7d0f9f0b pin; core main is not an interchangeable dependency. The separate combined core/Storage PR232 remains draft while Image/Cloud consumer acceptance is open. No new binary composition, model run or pin change is needed for this docs-only publication step. This includes the earlier PR #10 work. No native editor UI, general coding quality, or confidential peer execution claim is made.

[erikleblansch]

5c6248b adds explicit owner-level outputConstraint=qwen_tagged_tool_json_v1 selection through runtime configuration, isolated session argv, provider connection negotiation, submitted conversation and exact result binding. No tools means no constraint field; absent option preserves old frames/argv. Unsupported cores fail before runtime startup, without fallback. The model may still answer normally and all existing tool approvals remain required.

76 targeted Node tests and10 inert Python session tests pass (real HTTP/Unix framing, synthetic core/native replies only). Independent review checked compatibility with core PR212's manifest/request/IPC contract. The new source-check workflow includes the Python selector tests. Actual trial pins/provisioning are not yet changed: native source compilation/sanitizers and the joined source checks must pass before the real OpenCode scenario is rerun. No successful tool/action/coding claim is made.

[erikleblansch]

Commit689c7dd wires the explicit native selector all the way into the real OpenCode trial runtime and pins core9d044921f7e684737f9b10868429de42bc5416c9. New native trials require build3/backend2 capability provenance and selector-bound task receipt2 at both guest and outer verification; original Torch argv, receipt1, prompts, permissions and budgets remain unchanged. 83 targeted checks pass (38 native Python with the exact retained core parser,24 smoke Python,8 CI compatibility,13 JS); one separate retained-loader check is explicitly skipped. No local build or model execution.

Core source-only normal/native constraint checks pass; the wider unchanged upstream sanitizer test still records its known quantized-kernel failure. Core Quality38672934132 and the Code source checks gate any new real inference trial. This is a wired candidate, not proof that the earlier invalid JSON/model task failure is resolved.

[erikleblansch]

Original actual trial 38675278393 (Code 6aa6291 / core 9d044921f7e684737f9b10868429de42bc5416c9) is audited and still FAILED, with meaningful partial progress:

• Native build, conversion and provisioning pass with backend v2/build v3 and qwen_tagged_tool_json_v1.
• Seven submitted requests yield one completed function call and six execution_failed errors. The real OpenCode runtime completes one read action; no edit, test, complete native turn or coding acceptance follows.
• No invalid-output or execution-budget error is reported. The only retained core event is result_observed; the six failures' causes and full request ordering are not established.
• Runtime/project, provisioning-group, guest/private-data, VM/scratch and owned CI-host cleanup complete. Scoped normalized route/DNS observations match, not raw IPv6 route bytes.
• All 33 bundled Code sources, the core archive and six native wrappers match the exact pins. No raw model output exported.

Artifact 11946278747 ZIP SHA-256: 487bee78bb4857c6a5b48df7d69df21ceeb608ded9d9264122668480f62afa74. Original log ZIP: 5ff8a4055bf966f192f80b366cbeb8428d77793c6ee6525b3109eecf292af36e. README/evidence updated at 70c72fe. No retry dispatched; the next narrow change closes the pre-supervisor/post-report diagnostic gap without changing the model, prompt, budgets, retry policy or acceptance.

[erikleblansch]

The closed task-boundary diagnostic is now verified on exact core 5607f1177dc230a71414e7468c221b950f43bca6: Quality 38678859785 passes all 3 new diagnostic Rust cases and 13 output-constraint cases (workspace: 3561 distinct passes, 7 ignored). Namespace skips remain skips. Original log ZIP SHA-256: 60e43c4a0a63a250b4af601245d9e41772135149e47663a730a511866347d1b9.

Code e39267343612ed8b594d15db3e161465c25b8c2f pins that source. Code CI 38681259931 passes 405 checks with 4 explicit skips; all 114 source blobs match the merge checkout. The extra skip is the synthetic namespace test on a runner lacking bwrap, not actual sandbox acceptance. Original log ZIP SHA-256: 0b9d42a5f0526ada26cf56166b2698d48b83510a3d6ffb5d5950a113d3b9e93c.

Exactly one actual native follow-up 38681471469 is now running against that Code/core pair with the unchanged explicit Qwen4B profile, native backend, prompt/output/resource/time limits, task and acceptance rules. It is not a success claim or a presumed fix for the six execution failures. Core ledger-only commit 11bffbb6 records this evidence; the tested execution pin remains 5607f117.

[erikleblansch]

Follow-up candidate 3a5c1f5 changes only the two native core-pin lines to core3327115e07d2c4fa976b2d0c625729a68df0aaf6. The original failed actual38726624377 and its closed audit above remain unchanged.

Core332 adds a native-only verified-source descriptor lifecycle before GGUF preparation; all complete source/GGUF verification, memory floors, execution bounds, model/runtime pins and tool/workspace permissions remain intact. It is a candidate for the observed preparation-time pressure, not proof that specific pages were reclaimed or that coding now completes.

Core targeted checks:191 inert casesPASS, independent reviewPASS. Code targeted native suite:37PASS,2explicit retained-source skips; the exact new-core import case was then separately run andPASS (the retained patched-loader case was not rerun). No native library/model was loaded locally. Automatic source-check run38733022967 is terminalSUCCESS; its original source-bound audit follows. CoreQuality38733078778 and CodeQL38732695543 are the only new pending core source gates. No new actual has been dispatched and no budget was raised.

[erikleblansch]

All exact-source prerequisites for Code3a5c1f59b282952e378e9525f2cf47da43ee19b2/core3327115e07d2c4fa976b2d0c625729a68df0aaf6 have now passed and their original logs/checkout bindings were audited: Code Source38733022967 (410PASS/4SKIP), core Quality38733078778 and four-language CodeQL38732695543. Core Quality original ZIP SHA-256 dbe843ee87550cf3b38fed5b839e27ef395f684b358b1f0ec625818015b684ac. Exactly one actual read/edit/test follow-up is now running: https://github.com/VOLPAROSSA/volparossa-code/actions/runs/38737284583 (job116429700011, created2026-10-15T12:46:59Z). Same Qwen3-4B-Instruct-2507 profile, llama_cpp_bf16_v1, 180-minute job/115-minute actual timeout, task and resource/admission limits; no fallback or weaker guard. Both candidates remain frozen. Previous38726624377 remains a failed two-admission/one-read trial with no edit/test. This source-cache lifecycle candidate and its new pending trial do not establish confidential peer execution or general coding quality.

[erikleblansch]

Original hosted trial 38737284583 is terminal FAILURE on Code 3a5c1f5 and core 3327115e07d2c4fa976b2d0c625729a68df0aaf6.

The pinned source/build/preflight stages passed. During guest-inference the owned QEMU unit reported oom-kill / signal 9; SSH returned 255 and no guest-result.json was recovered. This does not establish which memory boundary caused the kill, which guest phase had been reached, or whether the new source-cache hint ran. It is not the same observed guest memory-guard failure as the preceding trial.

QEMU join, removal of the private VM scratch/overlay, unchanged observed host routes/DNS, runner restoration and build cleanup are confirmed. Guest-internal worker join/private cleanup are not independently evidenced without the missing guest result. No successful edit/test/model-execution or confidential-execution claim.

Original artifact 11974713652 contains 10 closed JSON records, SHA256 031678c85a91bc8210c5dff121b8a1037cdb31e9519bc5a19cfdd36ce39d0cf2 (API digest matches); log ZIP SHA256 358264ace79ac02eee6d355a8d095a0245b20d5ed14623a5bad14af48c480d39. Both ZIP CRC checks pass. Exact checkouts, 33 Code source blobs and the core Git archive were independently audited.

No duplicate actual is running. Next scoped work is bounded host-cgroup diagnostics in a separate Code candidate, without changing model, task, resource limits, cache settings or privacy.

[erikleblansch]

Original actual run 38742699743 is FAILURE, not complete coding-agent acceptance. Exact Code 8b21775 / core 3327115e; source/archive/runtime joins and original artifact/log ZIPs were independently audited.

This time provisioning and the BF16 native build completed, the VM remained active, and the real model produced two function calls observed as completed native read actions. The third request reached compute_deadline: the retained state has 4,987 prompt tokens, first forward beginning at 143,352 ms and finishing at 585,635 ms, then 26 generated tokens at 595,782 ms. No edit or test completed; no response was injected and no private peer execution is claimed.

The new source-bound QEMU observer joined and retained 2,345 samples. Its own 13 GiB cgroup reached the limit and recorded max events, but oom/oom_kill remained zero. These sequential observations do not prove the cause of all latency. Guest worker/runtime cleanup, private fixture removal, outer QEMU join and scratch removal are confirmed. Semantic host routes/DNS are unchanged; raw route bytes are not claimed unchanged.

Artifact ZIP SHA-256: 721d57778d0dd5f2bc259921f9cee24caaa0d2a98fc094d35c21f4a3cb1ed4da. Logs ZIP SHA-256: 5b03fddfa33dcf2d4514598297cdcdf06b5e9d66ec024ff5c06123fb3dc2ee98.

Next: diagnose preparation and prefill costs before changing the implementation. No repeated actual, larger budget, weaker model or relaxed privacy/cleanup gate has been selected.

https://github.com/VOLPAROSSA/volparossa-code/actions/runs/38742699743

[erikleblansch]

New ordinary-runtime candidate: native model connections are scoped to one OpenCode task, keep the full ordinary tool/context contract, distinguish per-request file removal from joined session cleanup, and never replay an admitted failed turn. The final task frame waits for provider/session close. Build4/backend3 and all seven native wrapper sources are bound by the closed provenance checker. Targeted evidence: 67 Node checks, 4 packer checks; native-provenance suite 42 passes / 2 explicit retained-source skips. These are source/Unix/HTTP contracts, not real model reuse or acceleration. The current actual core pin is deliberately unchanged until the paired core candidate passes hosted Rust/native checks; no actual trial was dispatched.

[erikleblansch]

Original source check 38759140548 passed for 97dfa4a: 455 passing checks and 4 explicit skips. Its PR merge checkout c71716b has the exact same complete tree as the candidate (8d83e3be36e289c4610504fd8206458e187da782). Original logs ZIP was preserved and CRC-checked, SHA256 1af925d30b96513c947aac5a99d63125bc2d9345b001f071dd6ab0b10445ff4f. This verifies source contracts, packaging and adapter/lifecycle behavior only, not real inference, acceleration or a finished OpenCode task. The native core candidate still needs its Rust quality run; the actual workflow retains the old core pin until the compatible source has been verified. No model run was retried.

[erikleblansch]

Source acceptance and one pinned actual trial:

• Core 0de65f423a6d96fa565facf6f283efae70adf2bd passes original-audited Quality 38766972432 and CodeQL 38766966694. The previously recorded upstream C++ sanitizer failure remains separate and unresolved.
• Code 266d1a03f0b327cc7221610372a8c626fe4b50d9 changes only the native core pin and its exact test expectation. Source run 38770956502 passes 455 checks with four explicit skips; merge checkout tree matches the candidate. Original log ZIP SHA-256: 6e355b256a5a84c4dde01ffe58bf283aadcd1dac0f061eef3b544c90408ebea3.
• The local inert native contract checks pass 43 cases, including exact retained-core builder import; one patched-loader-source fixture is explicitly skipped. No local model or native build ran.
• Exactly one actual was dispatched: https://github.com/VOLPAROSSA/volparossa-code/actions/runs/38771413590 with expected Code SHA above, qwen3-4b-instruct-2507-v1, and llama_cpp_bf16_v1. Model, upstream source, ordinary toolset and resource/time limits are unchanged.

This is a pending hosted disposable-VM experiment, not a successful edit/test, measured prefix acceleration, protected peer execution or completed coding agent.

[erikleblansch]

The single paired actual trial 38771413590 completed with FAILURE on exact Code 266d1a0 (tree 647f9b708849178719eab22a97f8db70fb30e40f) and core 0de65f423a6d96fa565facf6f283efae70adf2bd. No retry has been dispatched.

Original receipts show successful pinned model provisioning and native build-v4/backend-v3 construction. The first task request ends as execution_failed at about 521 seconds; no native tool call, edit or test completes, and the fixture is unchanged. The retained worker observation reaches generation of one token after a 4,572-token prompt, but this is not a successful model turn, prefix-reuse proof or completed coding task. The closed task diagnostic is unclassified and cache_hint reports supervisor_error; neither identifies a root cause. Independent source/receipt review is ongoing before proposing a fix.

The final receipts confirm task runtime cleanup, removal of the project and private guest data, empty owned units, QEMU join, VM scratch removal, and restoration of CI changes. Observed host routes and DNS match both semantically and byte-for-byte. No confidential or private-peer execution is claimed. The upstream C++ sanitizer issue remains open independently.

Original artifact ZIP SHA-256: 4dcedc22d678a0516b89d6ffd1d213ead188ab5e58827d94396a47857d19a0fa (matches GitHub artifact digest); log ZIP SHA-256: c3825b4f3de24599ad20dbf14499dc556dbefb5aa43173af6c4e3c2b262457cc. Both archives pass CRC checks. This does not supersede the earlier failed trial or convert passing source checks into actual acceptance.

https://github.com/VOLPAROSSA/volparossa-code/actions/runs/38771413590

[erikleblansch]

Current integration checkpoint — 15 October

Child PR16 and PR17 are merged into this candidate branch at cce5a21 (tree 085884d0e0c8e790484a288792e381b1aad7a495). They bind the core CPU-alignment provenance and normalize an exactly bound single-edit target before the unchanged owner approval policy. Hosted source run 38805641941 passes 504 unique tests with five explicit retained-source/bubblewrap skips; all 43 new permission cases pass. Root and independent reviews are complete.

One actual trial is running: https://github.com/VOLPAROSSA/volparossa-code/actions/runs/38806073909, pinned to this exact Code revision and core 6d65bb12ed7b8720d43868c9306313133a5b96ae. Model, full toolset, resource limits and execution permissions are unchanged. The core normal and complete ASan/UBSan source suites now pass, including the prior upstream alignment regression.

The preceding actual 38797816570 completed three model calls and two reads but not an edit/test; its missing private permission payload does not establish the exact retrospective rejection cause. The new trial remains unaccepted until original task and cleanup evidence is audited.

This PR stays draft and stacked on PR10. A read-only merge preview against current main 88c4dd7 is conflict-free, but is not an integration or runtime result. Preserve the source-bound trial and integrate the stack only after the required actual and final-source gates.