V3 W1 START: Fable-5 preflight deltas (10-point), WAL-shaped writer probes, adoption-scan baseline, D-PERT-1#630
Conversation
Records the merged V3 substrate entry-point arc: .claude/v3/ tree, mailbox-kanban doctrine locks, ractor helper-scope ruling, and the mailbox-as-owner compile attestation. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
WAL-shaped W1 writer (M24: board = write-ahead log), M7 re-scope ruling recommendation, W6a baseline inversion, W3 oracle ratchet metric, W2 probe-first reorder, ractor batching by API shape, D-PERT-1 + M21 pull-forwards into W1, gate-run rule, supervisor stays thin. EPIPHANIES E-V3-PREFLIGHT-1. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
…ht deltas applied) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
|
Warning Review limit reached
Next review available in: 43 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (5)
📝 WalkthroughWalkthroughThis PR bundles V3 planning/board documentation updates, a new classid adoption-scan counting module in lance-graph-contract, a stubbed BatchWriter skeleton with probe tests in lance-graph-planner, and a rename of ResonanceDto to PerturbationDto (with deprecated alias) across thinking-engine and its callers. ChangesV3 Planning Documentation
Classid Adoption Scan
Batch Writer Skeleton
ResonanceDto → PerturbationDto Rename
Estimated code review effort: 3 (Moderate) | ~25 minutes Sequence Diagram(s)sequenceDiagram
participant Caller
participant BatchWriter
participant Board as "Board (moves)"
participant AckSet
Caller->>BatchWriter: cast(on_behalf, moves, payload)
BatchWriter->>Board: record intent moves (ahead of ack)
BatchWriter-->>Caller: CastId
Caller->>BatchWriter: intent_moves(cast_id)
BatchWriter->>Board: read recorded moves
Board-->>Caller: moves (crash-replay visible)
Caller->>BatchWriter: ack(cast_id)
BatchWriter->>AckSet: mark cast acked
Caller->>BatchWriter: unacked()
BatchWriter-->>Caller: remaining un-acked casts
Poem
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
…ph handler (M25) Operator-directed rs-graph-llm + rig evaluation folded into W2e/W3b/W3c; Session transitions persist as KanbanMoves through the W1b writer, so replay = rebuild from the board (unifies with M24 board-as-WAL). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 185425a601
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| 2. **M7 ruling recommendation:** re-scope `SoaEnvelope` as the spec/descriptor | ||
| certification surface (`verify_layout` + field-isolation matrix are the | ||
| value; trait polymorphism has zero production impls). Doc-line ruling, | ||
| unblocks W1 without refactor. |
There was a problem hiding this comment.
This re-scopes SoaEnvelope on the premise that trait polymorphism has zero production impls, but the contract already has a non-test implementation at crates/lance-graph-contract/src/canonical_node.rs:1275 for NodeRowPacket<'a>, which exposes NodeRow slices to Lance through this trait. W1 implementers following this addendum could incorrectly treat the trait as descriptor-only and skip preserving/testing the live NodeRowPacket envelope path and its owner/byte-layout behavior.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Confirmed and corrected in e2ee6fbd — the premise was wrong. impl<'a> SoaEnvelope for NodeRowPacket<'a> (canonical_node.rs:1275) is a live production impl: the Lance-facing zero-copy envelope exposing &[NodeRow] as LE bytes (NODE_ROW_COLUMNS / NODE_ROW_STRIDE / as_le_bytes over repr(C, align(64)) with const size asserts).
Revised M7 ruling (plan Correction section + ENTROPY M7 row + E-V3-PREFLIGHT-1-CORRECTION): the two surfaces are complementary, not duplicates — SoaEnvelope is the storage-boundary surface (certification + the canonical Lance byte path, NodeRowPacket as its live impl); MailboxSoaView/Owner is the runtime read/mutate surface. W1 implementers MUST route storage bytes through the NodeRowPacket envelope path and preserve/test its owner/byte-layout behavior — the trait is explicitly NOT descriptor-only. Also folded the process lesson: negative-existence claims in rulings now carry the same exhaustive-grep declaration required of worker reports (the original claim was inherited from a fleet inventory row and repeated unverified).
Generated by Claude Code
…robes WAL-shaped BatchWriter<P> skeleton (M24: cast = intent record, ack = confirmation, unacked = replay surface, resolve_owner = W1c delegation cache) with todo! bodies. Three #[ignore]d probes fail as designed (ahead-update ordering, kill-after-cast replay, delegation miss-then-hit) — the W1b implementation commit un-ignores them. Uses the SHIPPED contract types (KanbanMove kanban.rs:151, KanbanColumn, MailboxId collapse_gate.rs:121) — no mint needed; the brief's conditional mint did not fire. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
…urbationDto (M2) The mechanical Morton-tile inverse-pyramid field (Psi) renames to PerturbationDto with a deprecated ResonanceDto alias at the definition site; the perspectival (Piaget Three-Mountains) ResonanceDto in awareness_dto.rs KEEPS its name, untouched. 8 files; cascade key-tier vocabulary untouched. thinking-engine 362 tests green; shader-driver (with-engine) 100 green; zero in-crate deprecated-alias usage remains. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
…ormant; 3 deltas KanbanMove already unifies mailbox/witness/libet/exec at the contract level; supervisor + core + symbiont WIRED; planner crate is the gap (zero SoA-view refs, zero classid awareness, GATE-1 false friend). W2b gains real-owner integration probe; M15 upgraded to blocking before any planner->kanban emission. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
… count_adoption Mirrors classid_canon_compat's decision procedure exactly (zero bit math): CanonHigh vs the three documented legacy shapes (zero-prefix, pre-flip V3-marker-high, render-prefix-high) per routing.md §5. Ambiguous variant reserved (#[non_exhaustive]) for the documented 0x1000-domain collision bound; never constructed today. Tests via sanctioned composers only (F12). 10/10 new tests; full contract suite 771 green. Baseline finding: no real classid-keyed Lance corpora in this container (toy fixture only) — t0 baseline runs where data lives once the CLI wrapper lands. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
… M25 design finalized ~0.4-0.5 us/step measured; KanbanSessionStorage = snapshot upsert + move cast via W1b writer; the append-only move log already exists (graph-flow-kanban KanbanPlanEnvelope). rig = oracle-frequency only. AdaWorldAPI/burn 403 build wall recorded. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
W1 fleet complete — all four workers landed; verdicts and numbersFleet landings (each centrally verified, gates in AGENT_LOG):
Planner-SoA reality audit ( graph-flow speed (W2e input, release, steady-state batch 4096): ~408–471 ns/task-step (ContinueAndExecute) vs ~512–538 ns stepwise — the ~80–100 ns delta is the per-step SessionStorage round-trip. Two-speed ruling confirmed with data: graph-flow = the replayable orchestration layer; sub-µs hot dispatch stays with M25 KanbanSessionStorage design finalized: rig: Task-wrapping production-proven ( Ops finding: workspace-root cargo in rs-graph-llm/rig 403s on the Next: W1b implementation (turns the 3 probes green, un-ignores them) → M15 rename → W2b real-owner integration probe. 🤖 Generated with Claude Code Generated by Claude Code |
…elope impl (codex #630 P2) canonical_node.rs:1275 ships the Lance-facing zero-copy LE envelope over &[NodeRow]. Revised ruling: SoaEnvelope (storage boundary) and MailboxSoaView/Owner (runtime view) are complementary surfaces; W1 routes storage bytes through the NodeRowPacket path and preserves/tests its byte-layout behavior. Negative-existence claims in rulings now require the exhaustive-grep declaration (guardrails rule 10, applied to the orchestrator). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
…n, mutual masking via phase machine The cast carries a descriptor (mailbox, dirty row-range, cycle), never bytes; the sink reads deltas through NodeRowPacket::as_le_bytes at flush; rows in sink phase are mutation-frozen until ack while thinking proceeds elsewhere — write masks thinking and vice versa. M24 gate extended with descriptor-not-bytes + mutation-freeze tests. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
…L entries (Addendum-7) Corrects Addendum-6's mutation-freeze over-design (it contradicted 'updates reprioritize, never gate'). The thinker reports and moves on; >=3 stacked writes per row are stacked WAL entries with independent acks; the sink coalesces naturally by reading the live store at flush. Probe 4 (probe_stacked_casts_never_refused) pins it — 4 ignored probes now define W1b green. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
… WAL ruling (Addendum-8) Stacked-casts-never-refused is provable (Strict readers cannot observe past their V_ref horizon); M24/M25/time-travel collapse into ONE mechanism (QueryReference::at + deinterlace); W1b ack carries the assigned LanceVersion as the WAL<->temporal join; the ractor actor's payload is its V_ref reading-horizon. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (2)
crates/lance-graph-planner/src/batch_writer.rs (1)
79-114: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winNo fallible surface for
ack/resolve_owneron malformed input.
ack(&mut self, _cast: CastId)and the rest of the write-path API return noResult, so once implemented there's no defined way to signal an unknown/already-ackedCastId(double-ack) or a resolver failure. Since the API surface is still open at the stub stage, it may be worth deciding now whether these should returnResult<_, SomeSnafuError>per the repo's error-handling convention, rather than silently no-op'ing or panicking once real bodies land.As per coding guidelines, crates/**/*.rs should "reuse
snafuerror patterns" for error handling.🤖 Prompt for 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. In `@crates/lance-graph-planner/src/batch_writer.rs` around lines 79 - 114, The write-path API in BatchWriter currently has no way to report malformed input or resolver failures, so decide the error contract before implementing the stubs. Update the `ack`, `unacked`, `intent_moves`, and `resolve_owner` methods in `batch_writer.rs` to use the crate’s `snafu`-based error pattern where appropriate, especially for unknown or double-acked `CastId` and any resolver failure, rather than silently no-op’ing or panicking. Keep the signatures and internal handling consistent with the existing `BatchWriter`/`resolve_owner` design so callers can distinguish success from failure.Source: Coding guidelines
crates/lance-graph-planner/tests/w1_probes.rs (1)
1-105: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueIntegration-style probes vs.
#[cfg(test)]guideline.These probes live in
tests/w1_probes.rsas a separate integration test crate rather than a#[cfg(test)]module alongsidebatch_writer.rs. This is a deliberate probe-first design choice (the doc comments inbatch_writer.rsreference this file directly), and testing the exportedBatchWriterAPI as a black box has some merit, but it runs counter to the stated repo convention.As per coding guidelines, crates/**/*.rs should "Add Rust unit tests alongside implementations via
#[cfg(test)]modules; prefer focused scenarios over broad integration tests."🤖 Prompt for 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. In `@crates/lance-graph-planner/tests/w1_probes.rs` around lines 1 - 105, The probe tests in w1_probes.rs are written as a separate integration test crate, which conflicts with the repo’s `#[cfg(test)]` unit-test convention. Move these scenarios into a `#[cfg(test)]` module alongside `BatchWriter` in the implementation area, keeping the same focused checks for `cast`, `ack`, `unacked`, `intent_moves`, and `resolve_owner`. Preserve the black-box assertions against the public `BatchWriter` API, but relocate them so they run as unit tests rather than under tests/w1_probes.rs.Source: Coding guidelines
🤖 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 `@crates/lance-graph-planner/src/batch_writer.rs`:
- Around line 35-67: The BatchWriter fields are only initialized in
BatchWriter::new and not yet read because the methods are still todo!(), so
rustc/clippy will report dead_code on next_id, board, acked, delegation_cache,
and pending_payloads. Add a temporary dead_code suppression on the BatchWriter
struct (or otherwise gate it for the W1b work) and leave a short note pointing
to the pending implementation in BatchWriter so the lint doesn’t block CI until
the fields are actually used.
---
Nitpick comments:
In `@crates/lance-graph-planner/src/batch_writer.rs`:
- Around line 79-114: The write-path API in BatchWriter currently has no way to
report malformed input or resolver failures, so decide the error contract before
implementing the stubs. Update the `ack`, `unacked`, `intent_moves`, and
`resolve_owner` methods in `batch_writer.rs` to use the crate’s `snafu`-based
error pattern where appropriate, especially for unknown or double-acked `CastId`
and any resolver failure, rather than silently no-op’ing or panicking. Keep the
signatures and internal handling consistent with the existing
`BatchWriter`/`resolve_owner` design so callers can distinguish success from
failure.
In `@crates/lance-graph-planner/tests/w1_probes.rs`:
- Around line 1-105: The probe tests in w1_probes.rs are written as a separate
integration test crate, which conflicts with the repo’s `#[cfg(test)]` unit-test
convention. Move these scenarios into a `#[cfg(test)]` module alongside
`BatchWriter` in the implementation area, keeping the same focused checks for
`cast`, `ack`, `unacked`, `intent_moves`, and `resolve_owner`. Preserve the
black-box assertions against the public `BatchWriter` API, but relocate them so
they run as unit tests rather than under tests/w1_probes.rs.
🪄 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: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 7cd0214e-4c17-4ed7-a16b-b49bc70b0b90
⛔ Files ignored due to path filters (1)
crates/thinking-engine/Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (19)
.claude/board/EPIPHANIES.md.claude/board/LATEST_STATE.md.claude/board/PR_ARC_INVENTORY.md.claude/board/STATUS_BOARD.md.claude/v3/ENTROPY-MILESTONES.md.claude/v3/INTEGRATION-PLAN.mdcrates/cognitive-shader-driver/src/engine_bridge.rscrates/lance-graph-contract/src/classid_scan.rscrates/lance-graph-contract/src/lib.rscrates/lance-graph-planner/src/batch_writer.rscrates/lance-graph-planner/src/lib.rscrates/lance-graph-planner/tests/w1_probes.rscrates/thinking-engine/src/bf16_engine.rscrates/thinking-engine/src/composite_engine.rscrates/thinking-engine/src/dto.rscrates/thinking-engine/src/dual_engine.rscrates/thinking-engine/src/engine.rscrates/thinking-engine/src/f32_engine.rscrates/thinking-engine/src/signed_engine.rs
| pub struct BatchWriter<P> { | ||
| /// Monotonic id generator for the next cast. | ||
| next_id: u64, | ||
| /// Board: intent moves recorded per cast, keyed by `CastId`, alongside the | ||
| /// mailbox the cast was recorded on behalf of. Visible between `cast()` | ||
| /// and `ack()` (and beyond, for crash-replay via `unacked()`). | ||
| board: HashMap<CastId, (MailboxId, Vec<KanbanMove>)>, | ||
| /// Casts that have been confirmed (acked). A cast present in `board` but | ||
| /// absent from `acked` is the crash-replay surface (`unacked()`). | ||
| acked: std::collections::HashSet<CastId>, | ||
| /// W1c delegation cache: `on_behalf` mailbox -> resolved owner mailbox. | ||
| delegation_cache: HashMap<MailboxId, MailboxId>, | ||
| /// Payloads recorded per cast (payload-generic; the writer never inspects `P`). | ||
| pending_payloads: Vec<(CastId, P)>, | ||
| } | ||
|
|
||
| impl<P> Default for BatchWriter<P> { | ||
| fn default() -> Self { | ||
| Self::new() | ||
| } | ||
| } | ||
|
|
||
| impl<P> BatchWriter<P> { | ||
| /// Construct an empty batch writer. | ||
| pub fn new() -> Self { | ||
| Self { | ||
| next_id: 0, | ||
| board: HashMap::new(), | ||
| acked: std::collections::HashSet::new(), | ||
| delegation_cache: HashMap::new(), | ||
| pending_payloads: Vec::new(), | ||
| } | ||
| } |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
All struct fields are currently unread → expect dead_code warnings.
Every method body is todo!(), so next_id, board, acked, delegation_cache, and pending_payloads are written in new() but never read anywhere in the crate yet. rustc's dead_code lint fires on private fields that are never read (independent of the containing struct being pub), so cargo clippy --all-targets --all-features will likely flag all five fields until the method bodies are filled in. If CI treats warnings as errors this could block the pipeline before the W1b implementation lands.
Consider a temporary #[allow(dead_code)] with a comment pointing to the W1b tracking item, or confirm CI doesn't deny warnings on this crate during the probe-first phase.
🔧 Suggested temporary suppression
+#[allow(dead_code, reason = "fields wired up in the W1b implementation commit; see tests/w1_probes.rs")]
pub struct BatchWriter<P> {As per coding guidelines, "Run cargo clippy --all-targets --all-features to catch lint regressions in Rust code."
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| pub struct BatchWriter<P> { | |
| /// Monotonic id generator for the next cast. | |
| next_id: u64, | |
| /// Board: intent moves recorded per cast, keyed by `CastId`, alongside the | |
| /// mailbox the cast was recorded on behalf of. Visible between `cast()` | |
| /// and `ack()` (and beyond, for crash-replay via `unacked()`). | |
| board: HashMap<CastId, (MailboxId, Vec<KanbanMove>)>, | |
| /// Casts that have been confirmed (acked). A cast present in `board` but | |
| /// absent from `acked` is the crash-replay surface (`unacked()`). | |
| acked: std::collections::HashSet<CastId>, | |
| /// W1c delegation cache: `on_behalf` mailbox -> resolved owner mailbox. | |
| delegation_cache: HashMap<MailboxId, MailboxId>, | |
| /// Payloads recorded per cast (payload-generic; the writer never inspects `P`). | |
| pending_payloads: Vec<(CastId, P)>, | |
| } | |
| impl<P> Default for BatchWriter<P> { | |
| fn default() -> Self { | |
| Self::new() | |
| } | |
| } | |
| impl<P> BatchWriter<P> { | |
| /// Construct an empty batch writer. | |
| pub fn new() -> Self { | |
| Self { | |
| next_id: 0, | |
| board: HashMap::new(), | |
| acked: std::collections::HashSet::new(), | |
| delegation_cache: HashMap::new(), | |
| pending_payloads: Vec::new(), | |
| } | |
| } | |
| #[allow(dead_code, reason = "fields wired up in the W1b implementation commit; see tests/w1_probes.rs")] | |
| pub struct BatchWriter<P> { | |
| /// Monotonic id generator for the next cast. | |
| next_id: u64, | |
| /// Board: intent moves recorded per cast, keyed by `CastId`, alongside the | |
| /// mailbox the cast was recorded on behalf of. Visible between `cast()` | |
| /// and `ack()` (and beyond, for crash-replay via `unacked()`). | |
| board: HashMap<CastId, (MailboxId, Vec<KanbanMove>)>, | |
| /// Casts that have been confirmed (acked). A cast present in `board` but | |
| /// absent from `acked` is the crash-replay surface (`unacked()`). | |
| acked: std::collections::HashSet<CastId>, | |
| /// W1c delegation cache: `on_behalf` mailbox -> resolved owner mailbox. | |
| delegation_cache: HashMap<MailboxId, MailboxId>, | |
| /// Payloads recorded per cast (payload-generic; the writer never inspects `P`). | |
| pending_payloads: Vec<(CastId, P)>, | |
| } |
🤖 Prompt for 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.
In `@crates/lance-graph-planner/src/batch_writer.rs` around lines 35 - 67, The
BatchWriter fields are only initialized in BatchWriter::new and not yet read
because the methods are still todo!(), so rustc/clippy will report dead_code on
next_id, board, acked, delegation_cache, and pending_payloads. Add a temporary
dead_code suppression on the BatchWriter struct (or otherwise gate it for the
W1b work) and leave a short note pointing to the pending implementation in
BatchWriter so the lint doesn’t block CI until the fields are actually used.
Source: Coding guidelines
…nstruct it in the dispatch test) + rustfmt sweep The C6 fixture's TemplateGet variant was matched but never constructed, tripping dead_code under -D warnings in CI. Exercise it through the same generic consumer instead of allow()ing it. rustfmt also reflows classid_scan.rs (fmt-dirty since #630) and the new modules. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
2 participants
What
Starts the V3 W1 wave (the keystone) on top of merged #629, carrying the operator-authorized Fable-5 preflight pass so every other session converges on the same deltas.
Docs / governance (landed)
.claude/v3/INTEGRATION-PLAN.mdAddendum 2026-07-02,E-V3-PREFLIGHT-1): ten deltas, headline ones:SoaEnvelopere-scoped as the spec/certification surface (zero production impls;verify_layout+ field-isolation matrix are the value).Vec<KanbanMove>per message — helper-scope by construction), D-PERT-1 + M21 pull-forwards, gate-run rule, supervisor stays thin.Code (landing on this branch as the fleet completes)
batch_writerskeleton + 3 ignored failing probes (ahead-ordering, kill-after-cast replay, delegation miss).ClassidForm,count_adoption) — all three legacy shapes via the compat-reader procedure, no bit math.ResonanceDto→PerturbationDto(+ deprecated alias); perspectivalawareness_dto.rsuntouched.Why
The preflight collapses were all reorders/merges of machinery the plan already carried (the V3-shape test) — landing them before W1 code means every later wave consumes them instead of rediscovering them.
Verification
cargo test -p lance-graph-supervisor --features supervisor— 22 green (ractor ownership attestation, AdaWorldAPI fork).cargo test+ central/v3-auditgate-run before each push (results in AGENT_LOG).🤖 Generated with Claude Code
https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
Generated by Claude Code
Summary by CodeRabbit
New Features
Bug Fixes
Tests