Repository navigation
Conversation
Contributor
Author
Independent serving split source review — 2026-10-21Verdict: SOURCE APPROVE for the frozen file hashes below, based on 437ee5f. No concrete source blocker found. This is not compilation, merge, deployment, performance or learning proof. CI remains required at the eventual exact head. Read AGENTS.md plan and repository README/CLAUDE guidance. This is phase 1 / No monolith: serving ownership becomes reviewable in smaller units while persistent residents retain the same lifecycle protection. Reconcile ownership seamReviewed reconcile_to_plan -> ReconcileLaunch -> launch_reconciled_plan and ServingOperation acquire/Drop. Acquire is still before reading the planned/intent snapshots. Early returns in the decision half drop the same operation; no new await or asynchronous handoff exists at the extracted synchronous call. The owned struct immediately destructures into locals, including the operation, and the same async move task captures it with an explicit local binding before any await. Abort-before-first-poll therefore still drops the captured operation; normal return/cancellation drops the local after inner lifecycle work. No resource guard is left in the caller across the extracted call: the watch and pending-model mutex guards end in existing blocks; occupancy is value data. Changed drop ordering of moved snapshots/strings does not release engine authority: they are not lifecycle guards. No cloned operation or early explicit release was introduced. Equivalence and contractsInspected and reran the author's read-only audit script (its prefix defines parser/read helpers only). It reports 288 original / 289 resulting function bodies; exactly three changed original bodies: reconcile, teacher fixture selector, production_source. Recomposing reconcile decision body and launch tail equals the original under its documented normalization. Selector differs only by the inserted teacher module. This script is supporting source evidence, not a Rust parser/typechecker; whitespace normalization alone cannot prove arbitrary string equivalence. Manually inspected the substantive seams and public facade reexports. Original public and crate-visible lifecycle/intent/demand/cache/catalog interfaces retain parent paths; sibling helpers are pub(super), and ServingDaemonModule remains at its original path. The source-contract fixture now includes the facade and all extracted production files, excludes test files, and preserves its anti-self-matching intent. Unchanged academy_batch/footprint remain outside its historical source scope. The public tests::exercise_cancelled_exclusive_teacher entry remains intact. Teacher --exact selector matches the new test module. Audit reproduced: all 3,640 original full comment lines retained; 94 test annotations retained; 81 diagnostic calls retained with explicit original serving target on moved calls. The shared probe macro target arm preserves INFO and probe_class and leaves the prior arm intact, matching the previously reviewed llama split arm. git diff --check passes (only Git's CRLF normalization warning). No local compiler or runtime operation was run, and candidate files were not edited. Reuse assessment: the change extracts existing responsibilities behind the existing module facade and one owned launch input; it introduces no second serving authority or copied policy algorithm. The existing diagnostic macro is extended once for target preservation. CI must cover full core tests, Windows test compilation, warning/hygiene gates and execution of the teacher subprocess fixture. Exact reviewed bytes (SHA256)
|
joelteply
force-pushed
the
codex/serving-daemon-by-owner
branch
from
October 21, 2026 08:58
3c47d2b to
039df9a
Compare
git grep -n -E AdmissionHold|GateRefusal|PromptCacheGrant -- core apps found only serving_daemon definitions and implementation references, no external named caller. Retain original pub(crate) root definitions for the pure-split visibility contract; children consume those types directly. Removes redundant reexports and artificial super-qualified signature uses; preserves type comments and behavior.
Contributor
Author
|
Independent source review APPROVE at 3676ae9: AdmissionHold, GateRefusal and PromptCacheGrant retain their original facade paths, fields and ownership semantics. Descendants access the parent definitions through existing super imports; AdmissionHold._operation is private as originally. GateRefusal inherent methods remain in their implementation owner. Independent comment multiset comparison across the three changed files found zero removed or added comment lines. No compilation performed by reviewer; current-head CI is still required. This supersedes the artificial facade-use warning correction. |
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
… measurement (#5229-#5233, #5240, #5244) (#5245) * code/file_engine.rs split by owner: 3,935 lines -> 144 + seven files, each under 1,000 Split by the names already in the file: suggest (nearest paths), ops (read, write, edit, delete, diff, undo, history), introspect (path resolution, exists, list_dir, glob), edit_checks (parse checks, notices, repair hints), apply (applying and reversing an edit), tests/{mod,filesystem}. Pure moves: children `use super::*`, moved private items become pub(super), the parent globs them back so every path holds; super::workspace_events becomes crate::code::workspace_events (one level deeper); gutter_line_number keeps its pub(crate) path through a re-export. file_engine tests: 58 before, 58 after, all pass. Lines past 1000: 155,781 -> 152,846. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 18a6f6d) * Split cognition/resource_admission.rs by owner (metrics, turns, lanes, gate, tests) Pure move: no behaviour change. Every file is at most 1,000 lines; public paths unchanged via glob re-exports. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 251e832) * Split working_memory.rs by owner (pure move) working_memory.rs 4138 -> 618 lines (types, snapshot, Drop; re-exports WorkingMemoryFaculty). Moved verbatim into working_memory/: ledger.rs 451 (construction, action statements, fingerprints, receipts) continuity.rs 429 (pins, notices, facts, snapshot/restore, claim-gone) perception.rs 490 (readers, dispatch, collapse, settlement) faculty.rs 213 (WorkingMemoryFaculty + render_trail) tests/{mod,actions,custody,faculty,open_views,pins,provenance,rebuilt_marker,trail}.rs (all <=423 lines; the two test mods become one tests tree) Widened to pub(super): push_receipt_head, collapse_results_over_cap, publish_live_handles. Test files carry #![cfg(test)] so source-hygiene keeps counting them as test code. Tests: working_memory filter 49 passed before and after. Lib warnings 122 -> 122, lib-test warnings 74 -> 74. source-hygiene baselines lowered: lines-over-1000 155781 -> 152643, excess test mods 19 -> 18. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 088a602) * Split openai_adapter.rs by owner (near pure move) openai_adapter.rs 3730 -> 267 lines (config/types; re-exports extract_reasoning). Moved into ai/openai_adapter/: runtime.rs 423 (construction, endpoints, runtime models, slot pool) lora.rs 198, registry.rs 139, reasoning.rs 107, embedding.rs 72 provider.rs 482 (AIProviderAdapter impl) stream.rs 789 (generate_stream_checked body) tests/{mod,lane_transport,request_shaping,lora_page_in, prefill_progress_is_liveness,unguaranteed_model_refusal_...}.rs (all <=593 lines) Not a pure move in two places, both forced by the 1000-line cap: - The AIProviderAdapter impl was a single 1250-line trait impl (one 790-line method), which cannot span files. generate_stream_checked's body moved verbatim to the inherent `stream_checked` in stream.rs; the trait method is now a one-line forward. - A doc link to crate::cognition::embedding in provider.rs became plain code formatting so the file-per-upward-import layer ratchet did not rise (cognition was named by two files after the split). Widened to pub(super): endpoints, endpoints_for_model, targets_local_serving_lane, refresh_runtime_models, resolve_dmr_model_name, runtime_models_contain, ensure_slot_pool, lora_scale_vector, match_lora_index, map_finish_reason, and the embedding helpers. Tests: openai_adapter filter 41 passed before and after. Lib warnings 122 -> 122, lib-test 74 -> 74. source-hygiene lines-over-1000 baseline 152643 -> 149913. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 463d122) * Split dream_consolidation.rs by owner (pure move + 9 justifications) dream_consolidation.rs 2817 -> 200 lines (module header, BrainRegion impl, engram builders, PersonaReflectionSource impl for the registry; re-exports distiller::* and region::*). Moved into cognition/dream_consolidation/: distiller.rs 484 (DistilledFact, Lens, SemanticDistiller) region.rs 358 (reflection source, region state, boundaries, entry points) consolidate.rs 516 (consolidate_with, review-only, received passes) passes.rs 475 (dream/review/supersession passes) tests/mod.rs 307 (incl. pub(crate) helpers training_producer imports), tests/region.rs 505 (pub(crate) mod region, path kept) Widened to pub(super): region consts/fields/ReviewBoundaries/DreamPermit/ CONSOLIDATE_EXECUTOR/consolidate_cooldown_elapsed, consolidate_with, the pass fns, observations_block, union_recall_keys. The original file's first code-level #[cfg(test)] (a field gate at line 694) hid ~1,300 lines of production from the unwrap ratchet; split files expose 9 unjustified unwraps. Each now carries a same-line JUSTIFIED comment (comments only) so the baseline does not rise. Tests: dream_consolidation filter 18 passed before and after. Lib warnings 122 -> 122, lib-test 74 -> 74. source-hygiene baselines lowered: lines-over-1000 149913 -> 148096, unwired 60 -> 59. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit afa6700) * Move the faculty tests out of llm_deliberation_faculty.rs into tests/ by theme (pure move) Each nested test module becomes its own file; the loose prompt-shaping and verdict tests group by theme as on the owner split. include_str! paths gain two ../ for the added directory depth. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 28e811b) * Move the tool-surface selection out of llm_deliberation_faculty.rs (pure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 3b34e76) * Move the request builder out of llm_deliberation_faculty.rs (pure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit cfcf416) * Move verdict production out of llm_deliberation_faculty.rs (pure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit e3bda14) * Move the prompt owner (types, view, messages, fit, system, segments) out of llm_deliberation_faculty.rs (pure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 5e3eced) * Move the reply claim and the hands surface out of llm_deliberation_faculty.rs (pure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 9817768) * Fix a super:: path the move rewrote one level short (pure move) The depth rewrite skipped 'in super::' meant for pub(in super::..) but also matched 'for fact in super::perception_facts'. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 24e88f6) * Widen the canary-only fit items the sibling modules use (visibility only) ShedPreview, ContextComposition's framing fields and the GroundingPiece notice helpers sit in prompt/ now; private -> pub(in super::super) like their siblings. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit c776905) * act_observe: split mod.rs by responsibility (no behaviour change) mod.rs was a 49-line coordinator plus a 2,460-line inline test module. The test module moves to act_observe/tests/ (declared `#[cfg(test)] mod tests;`): - tests/mod.rs shared fixtures: RecordingExecutor, FailingExecutor, ScriptedExecutor, ActThenSpeak, AlwaysAct, CountingSpeaker, body/body_with_wm/admission/acts_of helpers (410 lines) - tests/chain.rs causal ActChain linking scenarios (125) - tests/apply.rs apply_act scoping, folded/running alias, no-hands/failure (440) - tests/drive.rs drive_to_settle / settle_step scenarios + nudge rearm (534) - tests/credit.rs directed credit and live room input scenarios (509) - tests/investigation.rs multi-act investigation, re-awaken, promise/echo gates (507) Test bodies are moved verbatim (dedented); two relative paths gain one `super::` (settle::NARRATION_BUDGET, perception::batch_evidence). Uncompiled locally. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 946a8c0) * act_observe tests: mark moved test files #![cfg(test)] so ratchets read them as test code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 13a29ca) * act_observe tests: restore per-file imports (drive.rs lost them) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 13e77d7) * refactor(serving): split llama server by lifecycle responsibility (cherry picked from commit fc90c20) * fix serving split imports and ratchet reduced file excess (cherry picked from commit 08608e7) * Ratchet measured llama split file excess (cherry picked from commit e887947) * refactor(serving): split daemon responsibilities preserving lifecycle ownership (cherry picked from commit 26ff921) * fix(serving): preserve trait and shared test helper visibility (cherry picked from commit 039df9a) * fix(serving): use retained facade types in owner signatures (cherry picked from commit 597b0f1) * refactor(serving): retain contract types in their original facade git grep -n -E AdmissionHold|GateRefusal|PromptCacheGrant -- core apps found only serving_daemon definitions and implementation references, no external named caller. Retain original pub(crate) root definitions for the pure-split visibility contract; children consume those types directly. Removes redundant reexports and artificial super-qualified signature uses; preserves type comments and behavior. (cherry picked from commit 3676ae9) * source-hygiene: baselines measured once on the merged split train The only Rust file past 1,000 lines is the vendored quantized_llama.rs. Every changed count names its cause beside it; cargo test -p source-hygiene passes 43/43 on a fresh build of this tree. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
Author
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…iblings (#5246) Card d4909786. the_lived_copy_is_shed_by_the_substrate_not_by_her_identity was the only data-writing test in training_producer without HomeGuard. Her store resolves through the process home (`@persona:kimi` -> $HOME/.continuum/ personas/kimi/data/longterm.db). A parallel test that pinned and restored HOME could move it between this test's ensure-schema and its batch, so the two calls hit different databases: "Delete failed: no such table: staged_credit". Seen on #5239 at 815bdba and on #5244 at 3676ae9, while canary passed. It also wrote into the real ~/.continuum/personas/kimi on every machine that ran it. The fix: a tempdir home under HomeGuard::set, held for the whole test. This is the same pattern its siblings in part_2..part_6 already use. A scan of every tokio::test in training_producer/tests that touches data found no other test without the guard. Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
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
1 participant
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.
Split serving_daemon by responsibility so the serving lifecycle for persistent residents has smaller reviewable owners (architecture phase 1). The facade is 742 lines; extracted files are at most 798. Public paths, 94 tests, 3,640 original full comment lines and 81 diagnostic targets are retained.
Reconcile decisions pass owned inputs to a synchronous launch method; the same ServingOperation remains owned through the existing asynchronous task. Test source contracts include every extracted production owner and exclude tests. The teacher subprocess selector follows its moved test.
Independent source review APPROVE covers the frozen source hashes and operation/guard lifetime, public visibility, comments and test contracts. Source audit and diff checks pass. Rebased onto canary 8e26442 after #5239; direct source census measures 34,293 excess lines across 10 files. CI is required; no local compilation, deployment, latency or learning claim. Reconcile the shared probe macro and remeasure baseline if #5233 lands first.
CI at the prior head caught test helper/trait visibility errors; corrected and independently reviewed. Current head 3676ae9 also restores AdmissionHold, GateRefusal and PromptCacheGrant definitions to their original facade owner, preserving paths and comments without artificial import usage or warning suppressions. Independent source review approves this delta; exact-head core/Windows checks remain pending. Proposed integration is through Fable's reviewed union batch after green; original PR stays open until landed.