Repository navigation
Conversation
…g legacy option behavior (#144)
…00 paging regression
…ndidate per dataset
…ting false or zero
…tic zero or false
…t synthetic values
…rdinal and valid false/zero
… gate without reducing safety
…utstanding window while retaining serial per-RCB reads
…chment worker count
…e on discovery fast path
…ad of publishing partial enrichment as complete
Consumer integration checkpointExact engine HEAD |
… dropping baseline Merge source HEADs: #189 24aae46, #186 d095b3a, #187 4ed76cc, #182 ded5036. Eleven path updates: ten literal source/test file blobs verified unchanged against individual merge-bases, plus deliberate CI workflow integration instead of #182's historical limited-branch trigger. Synthetic-merge checkout and all stacked branch PR triggers. This is a draft integration candidate, NOT field-qualified, NOT ARSAS repin or release. No main modification.
|
Repository consolidation audit, 2026-10-22: exact source HEAD Closing this original PR as CONSOLIDATED / SUPERSEDED BY #190, WITHOUT merging it directly to |
1 participant
Field evidence
SIEBCU: RCB=66, 6281 MMS variables, 191 directory requests, negotiatedCalling=5, directory=15590ms, types=2837ms, projection=2327ms, total=21397ms. Exact physical repeat needed; no speedup claimed.
Change
Previously report inventory enrichment was one RCB at a time. A finite read-only worker pool now fills at most the negotiated MMS outstanding-calling window, bounded by <=8 and existing candidate count, retaining the same 64-candidate limit. Every RCB runs base structured Read and missing exact fields sequentially, no new names or writes; KPI recorder records each request; deterministic signature sorts request identities.
Risk and release controls
Engine source from pinned 9c7cb40... PR #184. Draft trial; ARSAS consumer lock remains unchanged until CI and physical checks. Ensure exact Request counts/identity, no duplicates/sequence loss, accepted data model/RCB parity and low latency on SIEBCU before public release.