Conversation
Without provider quotes, compute complete workload costs for each physically compiled candidate from workload demand, applicable ERP measurements and analytical resource estimates, including native operator CPU, heap workspace and maintenance programs. Report each candidate's resource breakdown in cost_comparison; complete provider quotes remain an optional override. Reject candidates that fail ERP accuracy admission before pricing, and accept a bounded classic-HLL confidence contract in scoped evidence for backend-local Regular HLL readouts. Split from #761 so #728's structural checks precede costing. The tree equals the pre-split #728 head. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This was referenced Sep 29, 2026
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Compute ERP and analytical workload costs
Split from #761 so that #728's structural checks come before any costing. This PR contains only the ERP/analytical pricing layer; native execution, SDS binding and scoped accuracy evidence are in #761.
Stack: main → #763 → #765 → #761 → #728 → this PR → #780 → #742 → #775. Diagnostics #756 branch from this PR.
Before this PR
Deployment required complete provider quotes (
workload_cost_evidence); without them, the snapshot compiler refused to deploy. ERP measurements informed only logical selection, and HLL readouts could never be certified.After this PR
Without provider quotes, Backend computes complete workload costs for each physically compiled candidate from workload demand, applicable ERP measurements and analytical resource estimates, including native operator CPU, heap workspace and maintenance programs.
cost_comparisonreports each candidate's resource breakdown (automatic_cost) and model version. Complete provider quotes remain an optional override.Accuracy admission stays separate from pricing: a candidate whose ERP accuracy admission fails is rejected before pricing, and missing or stale proof cannot be replaced by a favorable cost. Scoped evidence may carry a bounded classic-HLL confidence contract, which certifies backend-local Regular HLL readouts only.
Changes
workload_cost/automatic.rs: automatic workload resource model;select_candidatesprices without quotes.erp.rs,post_asap/cost_model.rs: ERP resource precedence over analytical estimates; HLL confidence model.compiler.rs: HLL contract validation and backend-local binding; ERP resource costs survive confidence targets.Validation
native_snapshot_topk,native_rate_topk); real-process bounded HLL execution.6042bb0a), so this PR's diff is exactly the ERP layer removed from feat: install, execute and recover native candidates over dataset-bound SDS #761. The full suite was not re-run on this tree during the split.The final tree equals the pre-split #728 head, so #780 and later PRs are unchanged.
🤖 Generated with Claude Code