Versioned facts + anti-entropy reconciliation (kills first-write-wins) - #323
Conversation
|
Warning Review limit reached
Next review available in: 3 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: defaults Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (7)
✨ Finishing Touches🧪 Generate unit tests (beta)
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 |
Content-addressed VersionedFact with verifyFact, reconcile() that kills first-write-wins (kept/adopted/healed/rejected_corrupt outcomes), and antiEntropySync() for pairwise neighbor repair. Pure TS, no NestJS deps, buildable to an importable CommonJS artifact for external harnesses. Tests cover the five Spec 16 A4 cases: heal early-hop corruption, corrupt never overwrites verified, anti-entropy repair+count, digest breaks on any single-token mutation, higher verified version supersedes. Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>
An empty node now adopts any arrival (verified or a provisional corrupt copy) so it becomes informed and stays healable, instead of refusing corrupt copies to null. rejected_corrupt is reserved for Rule 3 (holding a verified copy, refusing a corrupt one). This makes fidelity recover BY HEALING in the exp-08 retest rather than by damping spread, and keeps coverage/saturation intact. Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>
0a4910d to
3586f5d
Compare
✅ Recall Benchmark ResultsFull outputCommit: f507dd3 |
Summary
src/reconciliation/module:VersionedFact(provenance + version vector),reconcile(verified-copy wins, no first-write-wins), andanti-entropygossip repairExp-08 retest evidence (swarmlab, real package linked, 30 trials/cell, seed
rumor-mill-v1)Healing accounting: ~279
healed/ ~2768rejected_corruptper trial (0/0 in noiseless column) — fidelity recovers by verified copies overwriting corrupt ones, not by throttling spread.Honest caveat: anti-entropy over live edges is also a propagation channel — TTS dropped sharply (e.g. 17.27→1.10 rounds). No regression, but this is repair plus accelerated diffusion, not spread-neutral.
Run IDs: baseline
rm-baseline-mr7uziwl, engramrm-engram-mr7uzjds(both replay-verified).🤖 Generated with Claude Code