feat(contract): D-MBX-A6-P2 — Rubicon kanban lifecycle enforcement + ExecTarget strategy tag#439
Merged
Merged
Conversation
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughAdds Rubicon lifecycle enforcement: DAG-based Kanban transition helpers, an ExecTarget tag on KanbanMove, a RubiconTransitionError type, MailboxSoaOwner::try_advance_phase() that rejects illegal transitions without mutation, expanded crate re-exports, and related documentation updates. ChangesRubicon Lifecycle Enforcement
Sequence DiagramsequenceDiagram
participant Client
participant MailboxSoaOwner
participant KanbanColumn
Client->>MailboxSoaOwner: try_advance_phase(to)
MailboxSoaOwner->>KanbanColumn: can_transition_to(to)
alt legal
MailboxSoaOwner->>MailboxSoaOwner: advance_phase(to)
MailboxSoaOwner-->>Client: Ok(KanbanMove{ exec })
else illegal
MailboxSoaOwner-->>Client: Err(RubiconTransitionError{from,to})
Note over MailboxSoaOwner: no state mutation
end
Possibly related PRs
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
AdaWorldAPI
pushed a commit
that referenced
this pull request
May 30, 2026
…h, gating OQs Maps the whole unified-soa-convergence arc to completion: what's shipped (A1/A6-P1) vs in-PR (A6-P2 #439) vs queued; the dependency DAG; the critical path (A6-P3 -> A2 -> A3 -> A5 -> D-MBX-7 -> D-MBX-9 -> D-MBX-12); the gating open questions (OQ-11.6 surreal fork blocks the D-MBX-9 payoff); LE-1..4 + EW64 fold-in; recommended autonomous order (D-MBX-11 mechanical, then A6-P3 consumer wiring). https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.claude/board/D-MBX-COMPLETION-MAP.md:
- Line 15: The fenced code block starting at the shown fence in
D-MBX-COMPLETION-MAP.md is missing a language tag (triggering MD040); update the
opening fence to include a language identifier (for example change ``` to
```text) so the block is labeled (e.g., use "text" if it’s plain output),
leaving the block content and closing fence unchanged.
In `@crates/lance-graph-contract/src/kanban.rs`:
- Around line 148-168: Add an implementation of the Error trait for
RubiconTransitionError to match crate convention: add an empty impl
core::error::Error for RubiconTransitionError block (do not change the existing
Display impl), mirroring how CodecParamsError implements core::error::Error and
avoiding adding std::error::Error unless other types in this crate for the same
error category also implement std::error::Error (see
WidgetRenderError/CodecParamsError for reference).
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: dbda1b19-bf8f-4d37-8856-b3868fa97d14
📒 Files selected for processing (7)
.claude/board/D-MBX-COMPLETION-MAP.md.claude/board/EPIPHANIES.md.claude/board/LATEST_STATE.md.claude/board/STATUS_BOARD.mdcrates/lance-graph-contract/src/kanban.rscrates/lance-graph-contract/src/lib.rscrates/lance-graph-contract/src/soa_view.rs
AdaWorldAPI
pushed a commit
that referenced
this pull request
May 30, 2026
…D040 fence tag (CodeRabbit #439) - RubiconTransitionError now impls core::error::Error (empty block), matching the crate convention (cam::CodecParamsError) so it composes with ? / Box<dyn Error>. - D-MBX-COMPLETION-MAP.md: language-tag the DAG fence (markdownlint MD040). - Board: E-VERSION-ARC-IS-THE-KANBAN finding (the mailbox Lance version timeline IS the kanban arc for free; consume via surreal LIVE subscription like a GitHub CI/PR subscription; witness pointer = (mailbox_id, lance_version); D-MBX-9 collapses to subscription-wiring, substrate-free). 489 lib tests green; kanban rustfmt-clean. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.claude/board/EPIPHANIES.md:
- Around line 7-15: The markdown table block currently isn't surrounded by blank
lines which triggers MD058; edit the EPIPHANIES.md content so there is an empty
line immediately before the line starting with "**The consumption pattern..."
and an empty line immediately after the closing table markup (the line with the
last table row or the paragraph that follows the table) so the table is
separated by blank lines and satisfies markdownlint MD058.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: f6f236ba-fb1c-4637-8996-2c6b8839b6e6
📒 Files selected for processing (3)
.claude/board/D-MBX-COMPLETION-MAP.md.claude/board/EPIPHANIES.mdcrates/lance-graph-contract/src/kanban.rs
✅ Files skipped from review due to trivial changes (1)
- .claude/board/D-MBX-COMPLETION-MAP.md
🚧 Files skipped from review as they are similar to previous changes (1)
- crates/lance-graph-contract/src/kanban.rs
AdaWorldAPI
pushed a commit
that referenced
this pull request
May 30, 2026
…8 table blank-lines (CodeRabbit #439) - E-SUBSTRATE-IS-THE-SCHEDULER: surreal time-series/LIVE over the version arc is a cheap planner->execution scheduler firing back INTO the mailbox. Completes the substrate<->mailbox loop (outbound = E-VERSION-ARC free kanban; inbound = surreal schedules next advance_phase). Two-clock decoupling for free; ExecTarget::SurrealQl made literal (scheduled query = trigger + backend). - D-MBX-COMPLETION-MAP: D-MBX-9 reframed as a bidirectional subscription (not a build). - MD058: blank lines around tables in both new EPIPHANIES entries (swept both, not one-at-a-time, to stop the markdownlint round-trip). https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
AdaWorldAPI
pushed a commit
that referenced
this pull request
May 30, 2026
…i4-32D style -> tau -> JITson/Cranelift template -> KernelHandle); recipes are what styles select Corrects the recipe-centric framing. Grounded: contract::thinking = 36 styles/6 clusters, each mapped to a tau macro address for JIT; contract::jit = JitTemplate -> JitCompiler (ndarray Cranelift) -> KernelHandle -> StyleRegistry cache. Full substrate pipeline (all shipped, unwired into planner): i4-32D style -> tau -> JITSON template -> Cranelift -> native KernelHandle; style also selects the recipe/Tactic. Grounds #439's ExecTarget::Jit (= jit.rs path) and ::Elixir (= recipe_kernels interpreted layer). Minimal slice should wire STYLE SELECTION first, then recipe/kernel dispatch. Re-centering the in-flight Opus map. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…W64 coref pointer, cold SPO+AriGraph unify, Rubicon commit, BindSpace consumers) E-DUPLICATION-IS-INTRINSIC-VS-TEMPORAL: the fixed-SoA 'duplicate' = intrinsic awareness (CE64/gestalt) vs temporal belief-state arc (EW64/witness) = AriGraph Es/Ee = SoA1:SoA2; TD-RESONANCEDTO-DUP-1 is name-it-not-collapse-it. LE-1 EW64 as relativ-pronoun witness pointer (DeepNSM>Markov>grammar, no bundling); LE-2 unify cold SPO + cold AriGraph (EW64 = cheap witness pointer); LE-3 mailbox-cycle-end Rubicon commits SPO-W to cold + SLA/goalstate; LE-4 Odoo/OWL business substrate = other session. + BindSpace consumer map. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…get strategy tag
Schedule-layer spine for the ractor SoA owner + planner candidate-gen:
- KanbanColumn::{next_phases, can_transition_to} — the Rubicon lifecycle DAG
(Planning->CognitiveWork->Evaluation->{Commit|Plan|Prune}; Planning->Prune Libet
veto; Plan->Planning re-deliberate; Commit/Prune absorbing).
- KanbanColumn::is_absorbing — cycle-END (Commit/Prune, tombstone) vs Plan
(terminal decision that re-deliberates); ractor driver tombstones iff absorbing
(LE-3 cycle-end commit/SLA hooks here).
- MailboxSoaOwner::try_advance_phase — checked default; legal edge -> KanbanMove,
illegal -> RubiconTransitionError (no mutation). Lifecycle enforcement is now a
contract-level testable invariant the ractor driver uses.
- ExecTarget{Native,Jit,SurrealQl,Elixir} on KanbanMove — planner JIT-adjacent
execution-target tag (resolves the #437 deferred NOTE; KanbanMove still <=16 B).
Zero-dep preserved; 489 contract lib tests (+4); planner/shader-driver/supervisor
cargo-check clean. Board: STATUS_BOARD D-MBX-A6-P2 + EPIPHANIES (same commit).
https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…h, gating OQs Maps the whole unified-soa-convergence arc to completion: what's shipped (A1/A6-P1) vs in-PR (A6-P2 #439) vs queued; the dependency DAG; the critical path (A6-P3 -> A2 -> A3 -> A5 -> D-MBX-7 -> D-MBX-9 -> D-MBX-12); the gating open questions (OQ-11.6 surreal fork blocks the D-MBX-9 payoff); LE-1..4 + EW64 fold-in; recommended autonomous order (D-MBX-11 mechanical, then A6-P3 consumer wiring). https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…D040 fence tag (CodeRabbit #439) - RubiconTransitionError now impls core::error::Error (empty block), matching the crate convention (cam::CodecParamsError) so it composes with ? / Box<dyn Error>. - D-MBX-COMPLETION-MAP.md: language-tag the DAG fence (markdownlint MD040). - Board: E-VERSION-ARC-IS-THE-KANBAN finding (the mailbox Lance version timeline IS the kanban arc for free; consume via surreal LIVE subscription like a GitHub CI/PR subscription; witness pointer = (mailbox_id, lance_version); D-MBX-9 collapses to subscription-wiring, substrate-free). 489 lib tests green; kanban rustfmt-clean. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…8 table blank-lines (CodeRabbit #439) - E-SUBSTRATE-IS-THE-SCHEDULER: surreal time-series/LIVE over the version arc is a cheap planner->execution scheduler firing back INTO the mailbox. Completes the substrate<->mailbox loop (outbound = E-VERSION-ARC free kanban; inbound = surreal schedules next advance_phase). Two-clock decoupling for free; ExecTarget::SurrealQl made literal (scheduled query = trigger + backend). - D-MBX-COMPLETION-MAP: D-MBX-9 reframed as a bidirectional subscription (not a build). - MD058: blank lines around tables in both new EPIPHANIES entries (swept both, not one-at-a-time, to stop the markdownlint round-trip). https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…lot/4096 surface Grounded read (file:line agent map): polyglot frontends (Cypher/GQL/Gremlin/SPARQL) ARE wired to one LogicalOperator IR, but (1) '4096 surface' is 4 distinct uses of the magic number with NO canonical type, and (2) frontend->4096->SoA does NOT exist (IR is string-keyed end-to-end; DeepNSM's 4096 vocab runs parallel, unintegrated). SQL=backend only, NARS=truth-only (no parser), GEL=absent. Real shippable slice = the polyglot IR-conformance loop (grounds the untested hub via the existing conformance harness), then sequenced P-B unify-4096 / P-C NARS-frontend / P-D frontend->4096->SoA. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…raph, BEAM-analogous), NOT a query dialect Corrects my 'GEL absent' misread (I pattern-matched to EdgeDB's rebrand; wrong). Per user: GEL is their coinage = a graph substrate representing ANY language in graph form, analogous to OTP/BEAM. Repositions GEL as the IR/substrate layer (what frontends lower INTO), not a frontend. The BEAM analogy is already partly built: ractor = BEAM actor model in Rust; ExecTarget::Elixir = the BEAM-analogous exec path; version arc = the GEL execution trace; SoA = GEL state. GEL is the NAME for the D-MBX convergence, not a new component. Notes holograph 'GEL=global execution layer' as adjacent prior-art to reconcile. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…user pointer to planner path) Verified arena.push counts: Gremlin(37)+SPARQL(31) transpile to LogicalOp IR in-strategy; Cypher(0)+GQL(0) feature-detect then build via shared ArenaIR (#2). Two IR-construction routes converging on one LogicalOp arena that NOTHING asserts agree — exactly the gap the polyglot conformance loop (P-A) closes (GQL only flips feature bits today). 4096-in-planner = AutocompleteCache 64x64 heads (NOT deepnsm COCA-4096); parser->4096 join = cache/convergence.rs (the same half-built p64-drift terminus). Revises P-A to assert both dialect- AND route-equivalence. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…ned for efficiency, not 4 magic numbers Corrects my E-POLYGLOT-4096-IS-CONJECTURAL 'four unrelated 4096s' claim. Per user: the surfaces (COCA vocab, 64x64 attention heads, BindSpace 16x256, distance matrix) were deliberately aligned to the 0xFFF/12-bit address width so a parsed symbol / vocab rank / head / SoA slot co-index WITHOUT translation (the efficiency reason). CAM addressing invariant: content -> fixed-width address -> every layer indexes by it. Still-true: end-to-end frontend->0xFFF->SoA resolution is not yet called (string-keyed IR; convergence.rs partial) — but alignment makes wiring it cheap. Reframes P-B from 'merge 4 things' to 'declare the 0xFFF width as one canonical 12-bit type'. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…process layer Answers the transport question from ractor source: local actor msg = Box<dyn Any> pointer-move over Tokio mpsc + downcast (message.rs:62,102), ZERO serialize; gRPC (protobuf+HTTP/2+socket) is strictly slower and is LAB-ONLY (lab-vs-canonical:52-54, feature-gated, tonic optional). ractor's own distributed path = raw TCP serialize (ractor_cluster), not gRPC. The gRPC/Ballista/Flight + CAM/0xFFF occurrences are the cross-PROCESS movement layer where CAM = the compact 12-bit/6-byte wire form. Maps to inside(Tokio-Baton)/outside(cluster-TCP or Flight-CAM); gRPC=outside+lab, never inside. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…ept-for-concept to the D-MBX arc Cloned adaworldapi/firefly read-only (public, outside authorized scope). It's a runnable GEL substrate (any-lang->1.25KB Hamming node->executor; RUBBERDUCK/PYTHONIC/JAVELIN/RUSTLER compilers) + the OUTSIDE-transport prototype. CORRECTION: transport is Redis-streams mRNA, NOT gRPC (the gRPC framing is the analog). FireflyPacket (80B LE header + 1250B 10K-Hamming 4-zone + ctx) = the serialized cross-process Baton; addresses route / payload rides = '0xFFF as transport'. Independent convergence on hot-parallel/commit-single (RISC inv 4), backpressure (inv 8), 4-signature decomposition (= SoA column families), witness-as-trace (ttl/hop/trace=R4). Open: 256-bit SHA addr vs 12-bit 0xFFF; Redis-mRNA vs cluster-TCP vs Flight-CAM; trinity vs unify-on-Lance. Kept read-only ref; not in push scope. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…x planning substrate; distribution = one gRPC packet (BEAM location transparency) The planning-substrate answer for D-MBX-A6-P3: plan = a packet hopping a node-graph (firefly model), each hop = phase commit = version, trace = witness arc, ExecTarget = how each node runs. Punchline: because the substrate IS packet-based, the in-mailbox plan state is already a complete serializable execution context, so cross-server distribution = serialize once + gRPC hop to another instance (location transparency, the BEAM/ractor model). Reconciles with E-RACTOR-WANTS-TOKIO-NOT-GRPC: inside hop = move struct zero-serialize; outside hop = serialize ONCE at the boundary. Records the 5 design forks (DTO shape, 0xFFF-vs-SHA addr width, by-ref-not-copy SoA per R1, wire format, routing table) + the minimal PlanPacket first slice. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…ser: 'just brainstorming') Right-size the status: it's adjacent inspiration to remember, not a ratified build plan. The 5 forks + PlanPacket slice are sketch options, not a commitment. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…DTOs exist, wiring missing) Grounded: contract::recipes (34 Recipe catalogue) + recipe_kernels (34 Tactic kernels, 'Elixir-like layer') EXIST; planner references NEITHER (verified gap). Consumer DTOs ready: callcenter membrane, ontology OntologyRegistry (OGIT classes on SoA), rbac Policy/evaluate. Chain to wire: recipe -> candidate -> rbac gate -> ontology resolve -> membrane commit. This IS the Elixir-like template layer made real, part of D-MBX-A6-P3. Opus design map in flight for dep directions + minimal first slice + rbac-gate placement. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…i4-32D style -> tau -> JITson/Cranelift template -> KernelHandle); recipes are what styles select Corrects the recipe-centric framing. Grounded: contract::thinking = 36 styles/6 clusters, each mapped to a tau macro address for JIT; contract::jit = JitTemplate -> JitCompiler (ndarray Cranelift) -> KernelHandle -> StyleRegistry cache. Full substrate pipeline (all shipped, unwired into planner): i4-32D style -> tau -> JITSON template -> Cranelift -> native KernelHandle; style also selects the recipe/Tactic. Grounds #439's ExecTarget::Jit (= jit.rs path) and ::Elixir (= recipe_kernels interpreted layer). Minimal slice should wire STYLE SELECTION first, then recipe/kernel dispatch. Re-centering the in-flight Opus map. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…the planning substrate The thinking-style planning substrate: a PlanStrategy (#18 in default_strategies) that resolves the active ThinkingStyle -> cluster() -> mechanism -> selects which of the 34 recipe_kernels Tactics fire over a ThoughtCtx built from PlanContext markers. The style SELECTS the recipe (by cluster->mechanism), not a hardcoded id list; it also carries the tau() JIT address (grounds ExecTarget::Jit). Mirrors the mul::escalation precedent (thin planner module over zero-dep contract substrate). Planner already deps contract; no new edge; contract stays zero-dep. First slice: runs style-selected recipes over ThoughtCtx (the substrate the planner did not consume before); LogicalPlan passes through unchanged. Deferred: i4-32D style-vector argmax decode, Outcome->Candidate/KanbanMove, tau->JIT compile, membrane commit. 3 new tests + 192 planner lib green; rustfmt-clean. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
AdaWorldAPI
force-pushed
the
claude/sleepy-cori-aRK2x
branch
from
46d8716 to
fe1d6fb
Compare
…ical i4-32D style home but stale+orphaned; #439 StyleStrategy correctly on live recipes/recipe_kernels (adjacent, not wrong) Three recipe modules disambiguated: recipe.rs (i4-32D StyleRecipe->PersonaRecipe->JIT, the canonical style substrate) vs recipes.rs (34 tactics) vs recipe_kernels.rs (34 Tactic impls) + ontology OdooStyleRecipe (#433 renamed to avoid collision). The rot: recipe.rs's D-ATOM-1 blocker LANDED (contract::atoms real: I4x32/Atom/CANONICAL_ATOMS[33]) but recipe.rs never migrated off I4x32Stub/AtomStub AND isn't exported in lib.rs -> dead code. #439 wired the LIVE recipes/recipe_kernels (correct shippable choice); canonical-home migration = scoped follow-up D-MBX-A6-P3b (council-gated per #433). #439 stands as-is. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…or insight, not a guardian of who may think Per user: the council multiplies thinking (spawn savant lenses to GENERATE reframings), it does not police recording. LAND/REVISE/REJECT = synthesis output, not permission. Fixes the two places I framed it as a gate (P3b 'gate it through' + the 'council-pending' process note). Findings are deposited priors; the council is an amplifier to invoke, not a checkpoint to clear. P3b can proceed without it. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…M-SYN) — ratification = determinism firewall, council = catalyst Two of my earlier session entries reused the gate misframing. D-ARM-SYN ratification is the determinism firewall (nondeterministic-stays-upstream), not the brainstorm-council. Remaining 'council-gated' strings are pre-session or the E-SOA-VIEW status line; left intact (append-only) and superseded by the top-of-file correction. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…ssthrough theater; map the real style->PlannerDTO->truth-gated-Rubicon wiring Catalyst council surfaced: (1) honest correction — StyleStrategy::plan() discards its Outcome/ThoughtCtx (dead-store no-op), resolve_style ignores thinking_style (constant DEFAULT), try_advance_phase is dead (no real MailboxSoaOwner impl), exec/Libet test-only; my commit overstated a passthrough. (2) don't revive recipe.rs (4 savants: Jirak-uncited thresholds, 32v33, fifth-column drift, cross-crate cascade). (3) keystone = I4x32->argmax-> ThinkingStyle->cluster->tau->KernelHandle (the deferred decode IS the spine; no StyleRecipe). (4) PlannerDTO is drift; canonical home = D-MBX-A6 KanbanMove output. (5) Rubicon gates DAG-only, truth (f,c) latent-unwired (TruthGate/counterfactual_ready zero callers). Next build: argmax decode + plan() emits KanbanMove + truth-gate the transition; PROBE R-GATE first. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…lity not validity; 'truth' is a wisdom marker; validity is conferred externally at/after Commit The substrate's (f,c) is reliability machinery (confidence = accumulated-evidence coefficient, Cronbach/ICC/Jirak family); it never measures validity (ground-truth correspondence) — NARS is non-axiomatic, Stockfish is the external validity oracle. Reliability necessary-but-not-sufficient for validity; substrate can be confidently wrong. 'Truth' = the wisdom/settledness axis (phi^-1 ceiling = permanent humility) mislabeled. Splits the Rubicon gate: RELIABILITY gate hot (Evaluation->Commit on f/c+SD, crystallize), VALIDITY gate cold (post-commit, Stockfish/GoBD/reciprocal/FailureTicket). Corrects R-GATE to probe reliability-thresholding, validity stays the external post-commit gate. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…ility_of R-GATE measurable, honest passthrough Fixes the no-op the council's brutally-honest-tester caught in D-MBX-A6-P3a: - resolve_style now DECODES the 23D style vector (dominant axis -> ThinkingStyle), killing the constant-DEFAULT_STYLE bug (selection was identical for every query). - reliability_of(style,ctx) -> f32: the R-GATE measurable (style-selected recipes -> accumulated confidence). Reliability/settledness, NOT validity (external, post-commit; per E-RELIABILITY-NOT-VALIDITY). - plan() honestly labeled pure-passthrough: computes reliability, emits nothing (no faked KanbanMove pre-A6-overhaul). Dead-store theater removed. Probe-first (reviewers' rule): r_gate_reliability_varies_by_style asserts Analytical vs Creative select distinct mechanisms (non-cosmetic) BEFORE any gate field; +decode test +honest-passthrough test. 5 style_strategy + full planner lib green; fmt clean pre-commit. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…tuned Rubicon (f,c)/SD thresholds with measured Cronbach/ICC/Spearman/Pearson Follows E-RELIABILITY-NOT-VALIDITY: (f,c) is a reliability coefficient, so calibrate it with real reliability stats instead of hand-tuning. The crates exist: thinking-engine/ cronbach.rs (cronbach_alpha + bands, doc says it already 'replaces the BF16 heuristic with empirical test'), jc spearman_rho, codebook_pearson, calibrate_lenses (ICC), bgz-tensor cam_pq_calibrate/quality. Resolves the iron-rule-savant VIOLATES-I-NOISE-FLOOR-JIRAK (uncited 0.2/0.8/0.15/0.35). Float-boundary preserved: calibrate OFFLINE-float in thinking-engine -> emit frozen bands -> contract const (cited) -> hot Rubicon reads integer. Zero-dep boundary respected (contract carries the calibrated const, not the calc). Makes R-GATE rigorous: measured-band-changes-outcome, not 'styles differ'. Direction, not yet built; sequenced after R-GATE live-trace probe. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…tive: reliability = required-rung/checklist coverage over a knowns/unknowns SoA bitmask (AND-test, one cycle), no float/corpus User's cheap alternative to psychometric calibration: reliability becomes structural+prior = coverage of a normalized eval set. 10-rung ladder (rung:u8 ALREADY on ThoughtCtx/SPO/ CausalEdge64, E-LADDER doctrine) = depth axis; per-domain class_id-keyed checklist (HHTL- inherited) = eval axis; knowns/unknowns = presence bitmask (cognitive-risc-classes N3). Commit=required&known==required (SIMD AND+popcount, one cycle); Plan=named known-unknowns remain; Prune=unsatisfiable. DISSOLVES the 0.2/0.8/0.15/0.35 threshold problem (no calibrate, no Jirak) -> retires the iron-rule VIOLATES. Unknown-unknowns stay the cold Stockfish validity gate (checklist-completeness audit). Reuses rung+bitmask+class_id(#439); no float, no thinking-engine dep -> lighter than the psychometric slice. Complementary: cheap hot gate + psychometric offline audit. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
… = per-domain checklist = Odoo D-Atoms (one object); #433 built half of it User unification: the checklist IS the reasoning template; in Odoo = tax/billable/account fields. Grounded in #433 odoo_blueprint/style_recipe.rs: DAtom catalogue (FiscalCtx/Money/ ApplyRate/Tax/Law/...) = the checklist items as real variants; OdooStyleRecipe.atoms = which items a template REQUIRES; the 5-lit/6-dark split = the knowns/unknowns coverage bitmask already present. Reliability = required atoms LIT (required & known == required). Elixir open/closed = add-field(data) vs add-template(structure). CORRECTS my earlier 'OdooStyleRecipe = unrelated codegen' filing — it's the domain-instance face of the template-as-checklist; other domains write their own D-Atom catalogue. Cheap gate reads #433's atoms wholesale, no new checklist type. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…SIBILITIES (P1-P11) for council+brutal recalibration Possibility menu (not a committed sequence) for the reliability/checklist/template arc: P1 cheap coverage gate, P2 R-GATE probe, P3 StyleStrategy emit, P4 MailboxSoaOwner impl, P5 argmax-decode keystone, P6 delete recipe.rs, P7 reconcile 4-way recipe surface, P8 psychometric offline audit, P9 lance bump, P10 polyglot conformance, P11 class_id->checklist. Each cites its session finding + size/deps/risk. Council + brutal reviewers recalibrate next. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…rc — keystone is M1 Tactic::requires()->AtomMask (extraction not construction) Panel (cascade-impact + brutally-honest-tester + creative-explorer) converges: the keystone is the UNLISTED M1 = Tactic::requires()->AtomMask (the 34 kernels each already declare a latent checklist via which ThoughtCtx fields apply() reads; reify it -> P1/P7/P11 derived). Corrections: P2's witness corpus doesn't exist (build it first); P5 P0-blocked (I4x32 pack/unpack todo!() + 32-vs-33 fork); P3/P4 AP6 theater (zero prod callers); P1 dups recipes::Coverage; P6 cosmetic (recipe.rs never compiles in); P9 still blocked; P10 off-arc. Added M2 (completeness/unknown-unknown auditor) + M3 (version-arc scheduler). Recalibrated order: M1 -> P2+corpus -> M2. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
… checklist made data; reliability=coverage) The panel-recalibrated keystone of reliability-checklist-arc-v1. Adds ThoughtField (8-field enum) + ThoughtMask(u8, zero-dep: of/has/len/covered_by) + NON-defaulted Tactic::requires(); all 34 tactics declare which ThoughtCtx fields their apply() reads (audited from bodies; 4 algebraic constant-only tactics legitimately empty). covered_by (required&known==required) = the reliability-coverage gate. Reliability is a DECLARED ACCESSOR not a constructed gate (extraction not construction) -> makes P1/P7/P11 derived. Non-defaulted + teeth-test (requires_masks_are_varied_not_a_constant_stub: exactly-4-empty, >=8 distinct) = no silent theater. 9 recipe_kernels + full contract lib green; fmt clean pre-commit. Board: STATUS_BOARD D-MBX-A6-P3-M1 + EPIPHANIES shipped entry. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
…it assertion (CI #439) clippy::erasing_op on recipe_kernels.rs:1074: `m.0 & !0xFF` where m.0:u8 is always 0 (a u8 can't hold bits above 0xFF). The 'no stray high bits' guard was meaningless on a u8; the real invariant (len() <= 8) is already asserted. Removed the tautological line. https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
AdaWorldAPI
pushed a commit
that referenced
this pull request
May 30, 2026
… + auto-resolve
User instruction: "create the integration plan as a list of possibilities,
then use the council and brutally honest review to recalibrate, then continue
autoattended autonomous decision making and auto resolve."
Composed `post-438-integration-options-v1.md` with 8 options + 6 combinations.
Spawned 4 council reviewers in parallel (Opus, single main-thread turn) with
identical brief but diverse lenses:
R1 (architectural-fit) — B+C conditional on N1 class_id paired pass + u16
R2 (prior-art / drift) — B+H; REJECTS C as re-litigating user-owned forks;
flags AriGraph hot↔cold bridge as the missing
integration target the options doc never named
R3 (integration-coordination) — B+D; REJECTS C — in-flight collision with #439
on lance-graph-contract (31 commits, unstable,
KanbanMove const _ ≤16B size assertion)
R4 (brutal-critic / bias-hunter)— B+G; REJECTS C as ego-shipping; tier-set spec
disagreement = SPEC FREEZE, not Claude-session;
cites session pattern of 3 prior hallucinations
Auto-resolve per options-doc §5:
- B unanimous (4/4) → EXECUTED (2 stale-citation fixes this commit)
- C: 3/4 reject → NOT EXECUTED
- OD-1/2/3 + canonical-spec-self-disagreement → escalated to ISSUES.md as
explicit SPEC-OWNER decisions (R2+R4 unanimous framing)
- R2's AriGraph-bridge finding → captured as EPIPHANY (new, missed by all
prior session work): the discovery_origin byte's natural home is a column
on the AriGraph hot↔cold bridge, not the mailbox-SoA byte
- G (chess bring-up): 2/4 endorse → queued to STATUS_BOARD as D-CHESS-BRINGUP-1
(NOT executed on this branch per R1 "needs its own branch + freeze-decision
authority"); unblocked by #436's Rust Aerial+ (user-flagged 2026-05-30)
- D (help #439): 1/4 → below threshold, not executed
- H (cargo clean, 3.3G): flagged for user, not auto-executed (workflow not
architecture)
Fixed stale citations in own docs:
- discovery-origin-provenance-reconciliation §8: specs are on main now
(via main commits d1635db, 93ac046, a16d0f4, 45276eb), not on a
separate cognitive-risc-core branch as v1 of the doc claimed pre-rebase
- odoo-blueprint-survival-dossier §7 "Concrete next moves": Wave 1 went
D-ARM-13/14 via #436/#438, not D-ARM-1/2 as originally planned
Documentation only. No code or in-flight contract modified. Contract carrier
work (D-ARM-1/D-ARM-2/D-MBX-A6-P3) explicitly held until the spec owner
reconciles the canonical tier-set disagreement and the user picks
proposer-id width (OD-1).
https://claude.ai/code/session_017gZ6sPRXYPj5n7uJ7NBtRv
AdaWorldAPI
pushed a commit
that referenced
this pull request
May 30, 2026
… + auto-resolve
User instruction: "create the integration plan as a list of possibilities,
then use the council and brutally honest review to recalibrate, then continue
autoattended autonomous decision making and auto resolve."
Composed `post-438-integration-options-v1.md` with 8 options + 6 combinations.
Spawned 4 council reviewers in parallel (Opus, single main-thread turn) with
identical brief but diverse lenses:
R1 (architectural-fit) — B+C conditional on N1 class_id paired pass + u16
R2 (prior-art / drift) — B+H; REJECTS C as re-litigating user-owned forks;
flags AriGraph hot↔cold bridge as the missing
integration target the options doc never named
R3 (integration-coordination) — B+D; REJECTS C — in-flight collision with #439
on lance-graph-contract (31 commits, unstable,
KanbanMove const _ ≤16B size assertion)
R4 (brutal-critic / bias-hunter)— B+G; REJECTS C as ego-shipping; tier-set spec
disagreement = SPEC FREEZE, not Claude-session;
cites session pattern of 3 prior hallucinations
Auto-resolve per options-doc §5:
- B unanimous (4/4) → EXECUTED (2 stale-citation fixes this commit)
- C: 3/4 reject → NOT EXECUTED
- OD-1/2/3 + canonical-spec-self-disagreement → escalated to ISSUES.md as
explicit SPEC-OWNER decisions (R2+R4 unanimous framing)
- R2's AriGraph-bridge finding → captured as EPIPHANY (new, missed by all
prior session work): the discovery_origin byte's natural home is a column
on the AriGraph hot↔cold bridge, not the mailbox-SoA byte
- G (chess bring-up): 2/4 endorse → queued to STATUS_BOARD as D-CHESS-BRINGUP-1
(NOT executed on this branch per R1 "needs its own branch + freeze-decision
authority"); unblocked by #436's Rust Aerial+ (user-flagged 2026-05-30)
- D (help #439): 1/4 → below threshold, not executed
- H (cargo clean, 3.3G): flagged for user, not auto-executed (workflow not
architecture)
Fixed stale citations in own docs:
- discovery-origin-provenance-reconciliation §8: specs are on main now
(via main commits d1635db, 93ac046, a16d0f4, 45276eb), not on a
separate cognitive-risc-core branch as v1 of the doc claimed pre-rebase
- odoo-blueprint-survival-dossier §7 "Concrete next moves": Wave 1 went
D-ARM-13/14 via #436/#438, not D-ARM-1/2 as originally planned
Documentation only. No code or in-flight contract modified. Contract carrier
work (D-ARM-1/D-ARM-2/D-MBX-A6-P3) explicitly held until the spec owner
reconciles the canonical tier-set disagreement and the user picks
proposer-id width (OD-1).
https://claude.ai/code/session_017gZ6sPRXYPj5n7uJ7NBtRv
AdaWorldAPI
pushed a commit
that referenced
this pull request
May 31, 2026
Full-breadth integration spec wiring D-MBX kanban contract through witness commit (D-ATOM-5), surreal LIVE -> Rubicon kanban flip, ExecTarget backends, head2head two-view superposition in the shader driver, EW64-Markov Hebbian prefetch, language->SPO landing (D-LWS), and BindSpace decommission. Grounded against current main (#437/#439/ #444/#445) + two recon passes; flags the two hard blockers (lance-7 witness API, surreal fork dep / OQ-11.6) and the stale-branch caveat. https://claude.ai/code/session_01PLf95mURCY96TvKBFvSWEQ
AdaWorldAPI
pushed a commit
that referenced
this pull request
Jun 16, 2026
…oaOwner cherry-pick + LanceVersionScheduler + SurrealMailboxView (D-PG-6) Lands four tasks from the shortest-unblocking-path list surfaced after PR #497-#501 + the AdaWorldAPI/surrealdb fork bump (lance/lance-index =7.0.0, lancedb =0.30.0, ndarray exact-rev). All four meet at the single contract trait `MailboxSoaView`, closing the cascade in one commit (E-UNBLOCK-CASCADE-1). ## Task 3 — `NiblePath::{from_guid_prefix, prefix}` (zero-dep, foundational) Ontology-side keystone follow-up of PR #498's `classid → ReadMode` LE contract. The 20-nibble `classid(8) | HEEL(4) | HIP(4) | TWIG(4)` prefix overflows the 16-nibble MAX_DEPTH: the deterministic fold drops the canon-reserved high u16 of classid (root-first pack: `classid_lo(4) | HEEL(4) | HIP(4) | TWIG(4)`), returning None when the fold would be lossy. `prefix(d)` is the O(1) ancestor view; `prefix(d).is_ancestor_of(self)` holds for every d ≤ self.depth (the routing-cache view of a deeper class path). +7 tests in `hhtl::tests`; contract lib 619 → 632 green. ## Task 2 — `impl MailboxSoaView + MailboxSoaOwner for MailboxSoA<N>` Cherry-pick of jolly-cori-clnf9 commit 463d71b (integrated-cognitive-planner-v1 §2 Seam #3, +149 LOC). Adds `pub phase: KanbanColumn` field + zero-copy repr(transparent) slice impls (edges_raw, meta_raw) + the in-RAM Rubicon driving-loop test (`test_in_ram_driving_loop_walks_rubicon_to_commit`). The contract spine (#437/#439) now drives an actual loop end-to-end — no surreal, no ractor bus needed for the in-process case. +1 driving-loop test; cognitive-shader-driver lib 85 → 86 green. ## Task 1 — `LanceVersionScheduler` over `VersionedGraph::versions()` D-MBX-9-IN core impl (the CI-gated twin of the contract slice shipped 2026-05-31). Lives in `crates/lance-graph/src/graph/scheduler.rs`. Wraps a `VersionedGraph` + inner `VersionScheduler<S = NextPhaseScheduler>` and exposes: - `drive_once(view, exec)` — read current Lance version, lower to a move - `drive_at_latest(view, exec)` — fold `versions().last()` into a move - `current_dataset_version()` — typed `DatasetVersion` over nodes head Closes `E-SUBSTRATE-IS-THE-SCHEDULER`'s OUT-direction end-to-end. The OUT direction stays propose-not-dispose (R1): returned `KanbanMove` is for the caller's `MailboxSoaOwner::try_advance_phase` to apply. +5 tests with real on-disk tempdir Lance (no mocks). ## Task 4 — `SurrealMailboxView<'a>` (D-PG-6 contract slice) Read-only `MailboxSoaView` adapter the SurrealQL projection populates via `from_columns(...)` — pure zero-copy borrow over the kv-lance scan's byte buffers. Imports `MailboxSoaView` but NOT `MailboxSoaOwner` (compile-time enforcement of `kanban.rs:1-21` "surreal=project-read-only, callcenter=commit"). `read_via_kv_lance()` returns the new typed `SurrealContainerError::BlockedColdBuild` until the surrealdb fork dep in `Cargo.toml` is uncommented — kept off by default to avoid the ~10 min cold surrealdb build for contributors who don't need it. The contract surface is available today; the integration is one Cargo.toml edit + a SurrealQL projection body in `view.rs`. +4 tests; new `lance-graph-contract` dep in surreal_container/Cargo.toml; BLOCKED(C) marker flipped to RESOLVED. ## What this unblocks - **D-MBX-9-IN-impl** — SHIPPED (the contract trait now has a Lance-backed implementor). - **D-MBX-A6-P3** — still queued, BUT Seam #3 (the in-RAM loop) is now in-tree; a downstream session can wire the emit-side without depending on the unmerged jolly branch. - **D-PG-6 (Rubicon kanban VIEW)** — contract slice SHIPPED; impl-side gated on `BlockedColdBuild` flip-on (one Cargo.toml uncomment + projection body). - **Identity-architecture v1 §3 P-SCOPE-CLASSIFY** — solved (the bijection-width fix is deterministic + ancestor-preserving + falsifiable by tests). ## Tests + clippy - lance-graph-contract: **632** (+7 hhtl) - cognitive-shader-driver: **86** (+1 driving-loop) - lance-graph::scheduler: **5** (new module, real Lance tempdir) - surreal_container::view: **4** (new module) All clippy `-D warnings` clean on the new files. Pre-existing lints in lance-graph-ontology / lance-graph-planner / ndarray_bridge.rs are out of session scope. ## Board hygiene (mandatory rule) - LATEST_STATE.md — Contract Inventory PREPEND for the new types. - EPIPHANIES.md — E-UNBLOCK-CASCADE-1: three independent landings converge on one trait surface, closing four queued deliverables in one commit. - AGENT_LOG.md — task-by-task summary with test counts. https://claude.ai/code/session_01Xzyc27Nx3f8WC5KzwfWfjx
AdaWorldAPI
pushed a commit
that referenced
this pull request
Jun 16, 2026
…oaOwner cherry-pick + LanceVersionScheduler + SurrealMailboxView (D-PG-6) Lands four tasks from the shortest-unblocking-path list surfaced after PR #497-#501 + the AdaWorldAPI/surrealdb fork bump (lance/lance-index =7.0.0, lancedb =0.30.0, ndarray exact-rev). All four meet at the single contract trait `MailboxSoaView`, closing the cascade in one commit (E-UNBLOCK-CASCADE-1). ## Task 3 — `NiblePath::{from_guid_prefix, prefix}` (zero-dep, foundational) Ontology-side keystone follow-up of PR #498's `classid → ReadMode` LE contract. The 20-nibble `classid(8) | HEEL(4) | HIP(4) | TWIG(4)` prefix overflows the 16-nibble MAX_DEPTH: the deterministic fold drops the canon-reserved high u16 of classid (root-first pack: `classid_lo(4) | HEEL(4) | HIP(4) | TWIG(4)`), returning None when the fold would be lossy. `prefix(d)` is the O(1) ancestor view; `prefix(d).is_ancestor_of(self)` holds for every d ≤ self.depth (the routing-cache view of a deeper class path). +7 tests in `hhtl::tests`; contract lib 619 → 632 green. ## Task 2 — `impl MailboxSoaView + MailboxSoaOwner for MailboxSoA<N>` Cherry-pick of jolly-cori-clnf9 commit 463d71b (integrated-cognitive-planner-v1 §2 Seam #3, +149 LOC). Adds `pub phase: KanbanColumn` field + zero-copy repr(transparent) slice impls (edges_raw, meta_raw) + the in-RAM Rubicon driving-loop test (`test_in_ram_driving_loop_walks_rubicon_to_commit`). The contract spine (#437/#439) now drives an actual loop end-to-end — no surreal, no ractor bus needed for the in-process case. +1 driving-loop test; cognitive-shader-driver lib 85 → 86 green. ## Task 1 — `LanceVersionScheduler` over `VersionedGraph::versions()` D-MBX-9-IN core impl (the CI-gated twin of the contract slice shipped 2026-05-31). Lives in `crates/lance-graph/src/graph/scheduler.rs`. Wraps a `VersionedGraph` + inner `VersionScheduler<S = NextPhaseScheduler>` and exposes: - `drive_once(view, exec)` — read current Lance version, lower to a move - `drive_at_latest(view, exec)` — fold `versions().last()` into a move - `current_dataset_version()` — typed `DatasetVersion` over nodes head Closes `E-SUBSTRATE-IS-THE-SCHEDULER`'s OUT-direction end-to-end. The OUT direction stays propose-not-dispose (R1): returned `KanbanMove` is for the caller's `MailboxSoaOwner::try_advance_phase` to apply. +5 tests with real on-disk tempdir Lance (no mocks). ## Task 4 — `SurrealMailboxView<'a>` (D-PG-6 contract slice) Read-only `MailboxSoaView` adapter the SurrealQL projection populates via `from_columns(...)` — pure zero-copy borrow over the kv-lance scan's byte buffers. Imports `MailboxSoaView` but NOT `MailboxSoaOwner` (compile-time enforcement of `kanban.rs:1-21` "surreal=project-read-only, callcenter=commit"). `read_via_kv_lance()` returns the new typed `SurrealContainerError::BlockedColdBuild` until the surrealdb fork dep in `Cargo.toml` is uncommented — kept off by default to avoid the ~10 min cold surrealdb build for contributors who don't need it. The contract surface is available today; the integration is one Cargo.toml edit + a SurrealQL projection body in `view.rs`. +4 tests; new `lance-graph-contract` dep in surreal_container/Cargo.toml; BLOCKED(C) marker flipped to RESOLVED. ## What this unblocks - **D-MBX-9-IN-impl** — SHIPPED (the contract trait now has a Lance-backed implementor). - **D-MBX-A6-P3** — still queued, BUT Seam #3 (the in-RAM loop) is now in-tree; a downstream session can wire the emit-side without depending on the unmerged jolly branch. - **D-PG-6 (Rubicon kanban VIEW)** — contract slice SHIPPED; impl-side gated on `BlockedColdBuild` flip-on (one Cargo.toml uncomment + projection body). - **Identity-architecture v1 §3 P-SCOPE-CLASSIFY** — solved (the bijection-width fix is deterministic + ancestor-preserving + falsifiable by tests). ## Tests + clippy - lance-graph-contract: **632** (+7 hhtl) - cognitive-shader-driver: **86** (+1 driving-loop) - lance-graph::scheduler: **5** (new module, real Lance tempdir) - surreal_container::view: **4** (new module) All clippy `-D warnings` clean on the new files. Pre-existing lints in lance-graph-ontology / lance-graph-planner / ndarray_bridge.rs are out of session scope. ## Board hygiene (mandatory rule) - LATEST_STATE.md — Contract Inventory PREPEND for the new types. - EPIPHANIES.md — E-UNBLOCK-CASCADE-1: three independent landings converge on one trait surface, closing four queued deliverables in one commit. - AGENT_LOG.md — task-by-task summary with test counts. https://claude.ai/code/session_01Xzyc27Nx3f8WC5KzwfWfjx
AdaWorldAPI
pushed a commit
that referenced
this pull request
Jun 22, 2026
Full-breadth integration spec wiring D-MBX kanban contract through witness commit (D-ATOM-5), surreal LIVE -> Rubicon kanban flip, ExecTarget backends, head2head two-view superposition in the shader driver, EW64-Markov Hebbian prefetch, language->SPO landing (D-LWS), and BindSpace decommission. Grounded against current main (#437/#439/ #444/#445) + two recon passes; flags the two hard blockers (lance-7 witness API, surreal fork dep / OQ-11.6) and the stale-branch caveat. https://claude.ai/code/session_01PLf95mURCY96TvKBFvSWEQ
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
2 participants
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.
D-MBX-A6 Phase 2 — Rubicon kanban lifecycle enforcement + planner exec-target tag
Builds on #437 (D-MBX-A6-P1, merged). The schedule-layer spine that the ractor SoA owner drives and the planner candidate-generation emits into — all in the zero-dep
lance-graph-contract.Added
KanbanColumn::next_phases()+can_transition_to()— the Rubicon lifecycle DAG as a contract-level, testable invariant:KanbanColumn::is_absorbing()— disambiguates cycle-end columns (Commitcalcify /Prunedrop, tombstone now) fromPlan(terminal decision that re-deliberates back toPlanning). The ractor driver tombstones iff absorbing; the cycle-end commit/SLA decision (board LE-3) hooks here. (Fixes the latentis_terminal-vs-has-successor ambiguity a reviewer would flag.)MailboxSoaOwner::try_advance_phase()— checked default method: validates the edge against the DAG, returnsKanbanMoveon a legal transition orRubiconTransitionErroron an illegal one (no mutation). Lifecycle enforcement the ractor lifecycle driver uses — illegal transition is a typed error, not silent corruption.ExecTarget { Native, Jit, SurrealQl, Elixir }— the planner JIT-adjacent execution-target (strategy) tag, now a field onKanbanMove. Resolves the deferred NOTE from feat(contract): D-MBX-A6 Phase 1 — planner⟷ractor⟷surreal meta-DTO (kanban + soa_view + StepDomain::Kanban) #437. Distinct from the planner's 16 composable planning strategies — this names where the precipitated plan runs (native engine / Cranelift-JITson kernel / lowered SurrealQL / Elixir-like template).Invariants
[dependencies]stays empty (zero-dep).KanbanMoveis stillCopyand ≤ 16 B (const _size assertion holds with the newExecTargetfield).Verification
cargo test -p lance-graph-contract→ 489 lib tests (+4 new: lifecycle DAG, absorbing-vs-terminal,try_advance_phaseenforcement,ExecTargetdefault) + all integration suites green.cargo checkclean onlance-graph-planner,cognitive-shader-driver,lance-graph-supervisor(theKanbanMovefield addition is handled in thesoa_viewfake-impl; no exhaustive-match breakage).cargo macheteclean; touched files rustfmt-clean (pinned 1.95.0).Board hygiene (same PR)
STATUS_BOARD.mdD-MBX-A6-P2 row +EPIPHANIES.mdentry.Next (deferred, P3)
The consumer side:
impl MailboxSoaOwner for MailboxSoA<N>(ractor-owned, incognitive-shader-driver) wiringtry_advance_phaseinto the actual lifecycle, and planner candidate-gen emittingKanbanMoves with the selectedExecTarget.https://claude.ai/code/session_01R9AWgFa65uPnLyS2my2d2R
Generated by Claude Code
Summary by CodeRabbit
New Features
Documentation