Repository navigation
Split merge-train: the last seven oversized-file splits, one baseline measurement (#5229-#5233, #5240, #5244) - #5245
Merged
Merged
Conversation
… 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)
…, 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)
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)
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)
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)
… 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)
…ure move) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 3b34e76)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit cfcf416)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit e3bda14)
…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)
…culty.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)
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)
…nly) 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)
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)
…ad 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)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo (cherry picked from commit 13e77d7)
(cherry picked from commit fc90c20)
(cherry picked from commit 08608e7)
(cherry picked from commit e887947)
… ownership (cherry picked from commit 26ff921)
(cherry picked from commit 039df9a)
(cherry picked from commit 597b0f1)
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)
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
Contributor
Author
|
Union/source equivalence APPROVE for b7ddb23, conditional on exact-head CI and existing constituent reviews. Compared every changed continuum-core source path from each of #5229, #5230, #5231, #5232, #5233, #5240 and #5244 with its current exact PR head: zero final source differences for all seven (8/6/35/30/35/7/33 paths respectively). Per-PR source range-diffs match; differences are cherry-pick provenance metadata and the shared probe target arm already supplied by #5233, coalesced once for #5244. Direct Git-blob census of the union finds only vendored quantized_llama.rs above 1,000 lines (1,364), so file-size excess 364 is confirmed. Other hygiene counts still rely on the required batch CI. This is an integration equivalence review, not a replacement for constituent independent reviews or tests. #5244 CI has a training_producer staged_credit-table failure tracked separately by Fable as d4909786; no green or merge claim until gates pass. |
This was referenced Oct 21, 2026
This was referenced Oct 21, 2026
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…; a dense model's GGUF is read once (card 60f74ac5, f3ba3b56 item 19) Re-applied onto canary 445ae4b after #5245 split serving_daemon.rs (the builder and its readers now live in serving_daemon/placement.rs and moe_cache.rs) and #5249 moved the probe macro to continuum_base. moe_serving_context returned None both for a dense model and for a GGUF it could not read (open or header error, no architecture, a layout error other than NotMoe, no router top-k, no block_count). A MoE with unreadable geometry was served dense, with no expert placement and no word. It now returns Result<Option<ctx>, why>: Ok(None) is dense (NotMoe), and an expert count with no stacked tensors or a malformed layout is an error, not dense. The placement cache held only a MoE context, so for a dense model every call re-opened and parsed the GGUF header. The guard now caches the verdict per model (Option<ctx>, dense included) and probes serving.moe.geometry_unreadable once when it is built. A GGUF that does not resolve yet is not cached, so it is read once it lands. The two readers match Some((id, Some(ctx))). source-hygiene 43/43 on IntelMac. Core compile is CI's. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FRQWzgSfo79JtnwZywHKrE
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…s are re-asked (card 6080a28f, f3ba3b56 items 11 and 13) Re-applied onto canary 445ae4b after #5245 split llama_server.rs (prepare.rs, kv_support.rs, genome.rs) and #5249 moved the probe macro to continuum_base. Item 13: server_bin() falls through to a bare 'llama-server' from PATH when no override, slot or legacy install resolves. A stock build there runs without the fork's routes (slot save-size, /train, MTP). prepare_local_launch now calls engine_slots::note_unowned_launch, which probes serving.engine.unslotted when the engine is a bare PATH name no operator override chose. Resolved::Path's doc says what actually happens. Item 11: ensure_engine_kv_support_recorded latched once per process before probing, so a failed or timed-out --help probe left /train off and KV to the backend table for the core's life, and an engine promote kept the old binary's answers. The latch is now the binary that answered; a failure is asked again at the next launch, a promoted binary afresh, and a failure resets ENGINE_TRAIN_DIR and ENGINE_QUANTIZED_KV to unknown. source-hygiene 43/43 on IntelMac. Core compile is CI's. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FRQWzgSfo79JtnwZywHKrE
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…ned over a resident (card 1ca5abe1, f3ba3b56 item 4) Re-applied onto canary 445ae4b after #5245 split slots.rs and openai_adapter.rs (the request placement now lives in openai_adapter/stream.rs) and #5249 moved the probe macro to continuum_base. The slots test move this PR used to carry is already on canary. A Turn with no (persona, room) key (a nil-room eval on her live lane, or a caller that dropped its room) went out UNPINNED with cache_prompt on whenever the pool had a scratch slot, so the server placed it by similarity and could overwrite a resident's KV. The one placement rule is now slots::placement (slots/placement.rs): only a keyed Turn owns a slot. Every other request, a keyless Turn included, takes the slot its admission borrowed or the scratch slot, and without a scratch slot never persists a cache. A keyless Turn probes inference.turn.keyless with its purpose, and only a keyed Turn registers an in-flight slot guard. Test: slots::tests::only_a_keyed_turn_owns_a_slot pins the four cases. source-hygiene 43/43 on IntelMac. Core compile is CI's. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FRQWzgSfo79JtnwZywHKrE
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…s, not a silent re-prefill (card f3ba3b56, items 1, 3 and 6) Re-applied onto canary 445ae4b after #5245 split openai_adapter.rs (the completion probe now lives in openai_adapter/stream.rs) and #5249 moved the probe macro to continuum_base. The turn_admission test move this PR used to carry is already on canary. - turn_admission: TurnAdmission records how its KV reached the slot, as PageIn {None, Resident, Restored, RestoreRejected}, instead of a restored bool. A restore the engine refuses is probed at the decision (inference.restored_turn outcome=restore_rejected), so it counts even for a turn that never completes. - stream.rs: at completion, Restored is judged warm/cold as before. A Resident slot that reused nothing is its own class (inference.resident_prefix_lost), kept out of the restore verdict. - track-canary.sh: the deploy self-check counts restore_rejected against the economy, reports lost resident prefixes on their own line (a failure when they outnumber warm restores), and reports refused KV page ledger adoptions with their reasons. source-hygiene 43/43 on IntelMac. Core compile is CI's. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FRQWzgSfo79JtnwZywHKrE
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…state sized it (card 37eb1925) Re-applied onto canary 445ae4b after #5245 split serving_daemon.rs: the derivation, with_measured_state and the decision type all live in serving_daemon/prompt_cache.rs, which now owns sized_by and its test too. A sizing raised by a measured saved state reported the KV estimate's reason, so the receipt read the same before and after card 37eb1925's fix, and a fresh core's first sizing could not be told from the defect (Kimi's node, 8eaed76: one_conversation 4,816 MiB from a 5,050,734,260-byte page, under resident_demand_kv_estimate). derived_prompt_cache_mib keeps the estimate beside the raised demands, and sized_by relabels the decision resident_demand_measured_state only when a measured state raised a demand; a prior is never relabeled. The receipt's serialized_state_bytes reports the measured bytes instead of 'unobserved'. Test: a_measured_state_sizing_says_so_and_a_prior_is_never_relabeled. source-hygiene 43/43 on IntelMac. Core compile is CI's. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FRQWzgSfo79JtnwZywHKrE
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…ard cb3221a5) Re-ported onto canary 445ae4b after the #5245 file splits. Three flaky lib tests in a row turned canary red and refused every node's deploy on 2026-10-15 (#5066, #5067): tests that only READ the process home (continuum_home, HOME, home_dir) took no lock, so a guarded fixture could swap the home between two reads. Five readers take HomeGuard::preserve_blocking() (one async test takes preserve().await). The probe-env pair (tracing_init, probe_file_sink) both set/remove CONTINUUM_PROBE_DIR/CLASSES unlocked and restore each other's values mid-assertion; they take the same lock. test_env's doc says the ONE lock covers every test that mutates process env, so there is one env lock, not one per variable. The earlier forge test move is dropped: canary already split modules/forge. No production change. source-hygiene 43/43 on IntelMac, no baseline change. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01FRQWzgSfo79JtnwZywHKrE
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…split canary as one commit) Squash of the #5215 branch (9 commits plus its canary merge) re-applied onto current canary after #5245 split the files it edited and #5249 moved the probe macros: the same edits, landed in the split child modules. Ratchet baselines are canary's; CI measures the exact lowered counts. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo
joelteply
added a commit
that referenced
this pull request
Oct 21, 2026
…st-only items (rebased onto split canary as one commit) Squash of the #5195 branch (8 commits plus review fix) re-applied onto current canary: the same deletions landed in the split child modules from #5245, the probe path from #5249, and files #5200 already removed are skipped. Ratchet baselines are canary's; CI measures the exact lowered counts. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo
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.
Lands seven reviewed split PRs together. Each lowers the same lines-over baseline, so landing them one by one would mean a re-measure, a new head and about 45 minutes of Windows CI per PR.
Contents
Each PR's own commits, cherry-picked onto canary 8e26442 unchanged (
-xrecords the originals):code/file_engine.rscognition/resource_admission.rsworking_memory,openai_adapter,dream_consolidationllm_deliberation_faculty.rs(12,678 lines)act_observetestsinference/llama_server.rs(Codex)modules/serving_daemon.rs(Codex)routing/macros.rsarm (probe!(target: …), identical in serving: split llama-server lifecycle into bounded owner modules #5233 and refactor(serving): split daemon responsibilities and preserve lifecycle ownership #5244): it is applied once. Git merged the identical change cleanly.quantized_llama.rs(1,364 lines). The sum checks: 45,131 − (2,935 + 1,299 + 7,685 + 11,678 + 1,509 + 8,823 + 10,838) = 364.cargo test -p source-hygienepasses 43/43 on a fresh build.Verify
The union check is
git range-diffof each original PR range against its picks here. Codex has agreed to audit union and source equivalence. The originals stay open until this merges, then close as landed via this PR.🤖 Generated with Claude Code
https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo