V3 W1b LIVE: WAL batch writer implemented (4 probes green), M15 rename, temporal deinterlace synthesis, live oracle measurements#631
Conversation
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
…green; M15 MulGateDecision rename
W1b: BTreeMap board (monotonic CastId = cast order), ack(cast,
LanceVersion) with acked_version getter (the WAL<->temporal join,
Addendum-8), delegation cache, never-refuses stacking. All 4 probes
un-ignored and passing.
M15 (blocking-before-W2): planner-local GateDecision{Proceed,Sandbox,
Compass} -> MulGateDecision + deprecated alias; contract kanban gate
{Flow,Hold,Block} keeps the name; zero internal alias usage. Kills the
false friend before any planner->kanban emission.
Planner lib 204 green; clippy clean on all six touched files
(pre-existing nars_engine.rs:492 deprecated-CausalEdge64 debt noted,
untouched).
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
…-8.7 s LLM round trip (Addendum-9) Live rig(xai)->FlowRunner oracle-node run; self-hosted JITSON serve.rs test-bench delta recorded; rs-graph-llm design-intent + GraphRAG-rs references folded. 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: 11 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 (12)
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 |
![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: 563f560a08
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".
| let cast = CastId(self.next_id); | ||
| self.next_id += 1; | ||
| self.board.insert(cast, (on_behalf, moves)); | ||
| self.pending_payloads.push((cast, payload)); |
There was a problem hiding this comment.
For any real W1b sink or crash-replay path, this stores the descriptor P in a private vector but no method ever exposes or drains pending_payloads; unacked() returns only CastId and intent_moves() only returns kanban moves. A writer can therefore record an unacked cast that replay code cannot retrieve the SoA descriptor for, while acknowledged casts also retain their payload forever. Please add a pending/drain accessor or include the payload in the replay surface.
Useful? React with 👍 / 👎.
| pub use compass::{CompassDecision, CompassResult}; | ||
| pub use dk::{DkDetector, DkPosition}; | ||
| pub use gate::GateDecision; | ||
| pub use gate::MulGateDecision; |
There was a problem hiding this comment.
This rename adds a deprecated gate::GateDecision alias, but the public mul prelude no longer re-exports it. Any downstream code using the previous path lance_graph_planner::mul::GateDecision now gets an unresolved import even though the alias exists one module lower; if the M15 rename is meant to be a deprecation instead of a hard break, re-export both GateDecision and MulGateDecision here.
Useful? React with 👍 / 👎.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
2 participants
What
The W1 keystone goes live, plus the operator rulings that shaped it and the measurements that close the two-speed architecture's empirical loop.
Code
batch_writer.rs):BTreeMapWAL board (monotonicCastId= cast order for free),ack(cast, LanceVersion)+acked_version()— the CastId↔LanceVersion join wiring the WAL into the temporal classifier — delegation cache, never-refuses stacking. All 4 probes un-ignored and passing; planner lib 204 green.GateDecision{Proceed,Sandbox,Compass}→MulGateDecision+ deprecated alias. The contract kanban gate{Flow,Hold,Block}keeps the name; the false friend that would have routed the wrong gate intoadvance_on_gateis dead. Confirmed:collapse_gate::GateDecisionis a third, distinct type (engine_bridge consumer) — untouched.probe_stacked_casts_never_refused— 3 stacked casts, distinct WAL entries, out-of-order acks.Operator rulings pinned (plan Addenda 6–9)
NodeRowPacket::as_le_bytes; eager ASAP drain; "the write masks the thinking and vice versa."E-V3-TEMPORAL-DEINTERLACE-1): replay =QueryReference::at(v, rung)+deinterlace— M24 crash-replay, M25 session-replay, and time-travel are ONE mechanism; no-refusal is provable (Strict readers can't observe past theirV_ref).Measurements (live, evidence in EPIPHANIES)
serve.rs(JITSON) as the local zero-cost CI oracle.automataIA/graphrag-rs) as RAG prior art; API-design inventory in flight.Verification
cargo test -p lance-graph-planner --test w1_probes— 4/4 passing (0 ignored)cargo test -p lance-graph-planner --lib— 204 greennars_engine.rs:492deprecated-CausalEdge64 debt noted, untouched)🤖 Generated with Claude Code
https://claude.ai/code/session_01MLBnPuScZy6w9di2QEjsXM
Generated by Claude Code