@tangle-network/agent-eval 0.208.2 → 0.210.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +26 -0
- package/README.md +1 -1
- package/dist/adapters/http.d.ts +2 -2
- package/dist/{agent-profile-CEhxR_I3.d.ts → agent-profile-BGyt2z8U.d.ts} +3 -3
- package/dist/{agent-profile-CEhxR_I3.d.ts.map → agent-profile-BGyt2z8U.d.ts.map} +1 -1
- package/dist/{agent-profile-cell-BHcB-mZX.d.ts → agent-profile-cell-BNp-Js9F.d.ts} +2 -3
- package/dist/agent-profile-cell-BNp-Js9F.d.ts.map +1 -0
- package/dist/{agent-profile-cell-DcFW4vOc.js → agent-profile-cell-BYc0bxsL.js} +2 -2
- package/dist/agent-profile-cell-BYc0bxsL.js.map +1 -0
- package/dist/analyst/index.d.ts +12 -12
- package/dist/analyst/index.js +5 -5
- package/dist/{analyze-runs-GZ0VSwui.js → analyze-runs-3xpby3p2.js} +6 -6
- package/dist/analyze-runs-3xpby3p2.js.map +1 -0
- package/dist/{analyze-runs-CTJyUplR.d.ts → analyze-runs-BlslTkOJ.d.ts} +4 -4
- package/dist/{analyze-runs-CTJyUplR.d.ts.map → analyze-runs-BlslTkOJ.d.ts.map} +1 -1
- package/dist/{benchmark-command-BC265T1r.js → benchmark-command-S8tCtkxh.js} +6 -6
- package/dist/{benchmark-command-BC265T1r.js.map → benchmark-command-S8tCtkxh.js.map} +1 -1
- package/dist/benchmarks/index.d.ts +5 -7
- package/dist/benchmarks/index.d.ts.map +1 -1
- package/dist/benchmarks/index.js +3 -3
- package/dist/builder-eval/index.js +1 -1
- package/dist/campaign/index.d.ts +6 -6
- package/dist/campaign/index.js +10 -10
- package/dist/{campaign-evidence-DW3Krxov.js → campaign-evidence-OcUciD-N.js} +4 -4
- package/dist/campaign-evidence-OcUciD-N.js.map +1 -0
- package/dist/{campaign-C5pjR-Qn.js → campaign-wvvJA1ps.js} +14 -14
- package/dist/{campaign-C5pjR-Qn.js.map → campaign-wvvJA1ps.js.map} +1 -1
- package/dist/claude-jsonl-dZir12m6.js +86 -0
- package/dist/claude-jsonl-dZir12m6.js.map +1 -0
- package/dist/cli.js +7 -7
- package/dist/contract/index.d.ts +10 -10
- package/dist/contract/index.js +92 -16
- package/dist/contract/index.js.map +1 -1
- package/dist/counterfactual-DtMWvn7D.js.map +1 -1
- package/dist/{default-registry-DZFWY2_E.d.ts → default-registry-Crk3QEy1.d.ts} +6 -6
- package/dist/{default-registry-DZFWY2_E.d.ts.map → default-registry-Crk3QEy1.d.ts.map} +1 -1
- package/dist/{default-registry-DX5R7_n9.js → default-registry-DF4Yp7HQ.js} +2 -2
- package/dist/{default-registry-DX5R7_n9.js.map → default-registry-DF4Yp7HQ.js.map} +1 -1
- package/dist/{define-agent-eval-C8zBSbsc.d.ts → define-agent-eval-B_zZY2AZ.d.ts} +7 -7
- package/dist/{define-agent-eval-C8zBSbsc.d.ts.map → define-agent-eval-B_zZY2AZ.d.ts.map} +1 -1
- package/dist/{define-agent-eval-BjKXcDQp.js → define-agent-eval-CoXXDCMu.js} +7 -7
- package/dist/{define-agent-eval-BjKXcDQp.js.map → define-agent-eval-CoXXDCMu.js.map} +1 -1
- package/dist/diagnosis/index.d.ts +2 -5
- package/dist/diagnosis/index.d.ts.map +1 -1
- package/dist/diagnosis/index.js +2 -26
- package/dist/diagnosis/index.js.map +1 -1
- package/dist/{dspy-rlm-engine-sHwdSxK3.js → dspy-rlm-engine-BpaKVKJA.js} +2 -2
- package/dist/{dspy-rlm-engine-sHwdSxK3.js.map → dspy-rlm-engine-BpaKVKJA.js.map} +1 -1
- package/dist/{engine-DProXJ7A.d.ts → engine-D-YZ32ZJ.d.ts} +5 -12
- package/dist/engine-D-YZ32ZJ.d.ts.map +1 -0
- package/dist/{eval-campaign-DU_p8qVt.js → eval-campaign-4byTtqrf.js} +4 -4
- package/dist/{eval-campaign-DU_p8qVt.js.map → eval-campaign-4byTtqrf.js.map} +1 -1
- package/dist/evaluation.d.ts +2 -2
- package/dist/{exact-types-Cq6xXNVR.d.ts → exact-types-DjhfZglz.d.ts} +2 -2
- package/dist/{exact-types-Cq6xXNVR.d.ts.map → exact-types-DjhfZglz.d.ts.map} +1 -1
- package/dist/experiment/index.d.ts +2 -2
- package/dist/experiment/index.js +8 -8
- package/dist/{feedback-trajectory-C6Z8PRoF.d.ts → feedback-trajectory-CX01700m.d.ts} +3 -3
- package/dist/{feedback-trajectory-C6Z8PRoF.d.ts.map → feedback-trajectory-CX01700m.d.ts.map} +1 -1
- package/dist/hosted/index.js +1 -1
- package/dist/{index-CbBfavbT.d.ts → index-D0_l25We.d.ts} +21 -42
- package/dist/{index-CbBfavbT.d.ts.map → index-D0_l25We.d.ts.map} +1 -1
- package/dist/{index-nwNoS582.d.ts → index-DnZdBxCL.d.ts} +14 -11
- package/dist/{index-nwNoS582.d.ts.map → index-DnZdBxCL.d.ts.map} +1 -1
- package/dist/{index-DkucR0KJ.d.ts → index-UxNTvn-w.d.ts} +6 -6
- package/dist/{index-DkucR0KJ.d.ts.map → index-UxNTvn-w.d.ts.map} +1 -1
- package/dist/{index-Bzto04CY.d.ts → index-xnhRdqik.d.ts} +9 -20
- package/dist/{index-Bzto04CY.d.ts.map → index-xnhRdqik.d.ts.map} +1 -1
- package/dist/index.d.ts +28 -37
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +28 -90
- package/dist/index.js.map +1 -1
- package/dist/judge-gate-C59znGmR.js +638 -0
- package/dist/judge-gate-C59znGmR.js.map +1 -0
- package/dist/judge-gate-D2hCppem.d.ts +439 -0
- package/dist/judge-gate-D2hCppem.d.ts.map +1 -0
- package/dist/{kind-factory-Cn1lbsAd.js → kind-factory-ivTx6koH.js} +2 -11
- package/dist/{kind-factory-Cn1lbsAd.js.map → kind-factory-ivTx6koH.js.map} +1 -1
- package/dist/{llm-judge-HLWASwUJ.js → llm-judge-Tr7MBUnf.js} +9 -9
- package/dist/llm-judge-Tr7MBUnf.js.map +1 -0
- package/dist/{matrix-Cqkk7eYI.d.ts → matrix-jHTsHEkR.d.ts} +2 -2
- package/dist/{matrix-Cqkk7eYI.d.ts.map → matrix-jHTsHEkR.d.ts.map} +1 -1
- package/dist/meta-eval/index.d.ts +3 -133
- package/dist/meta-eval/index.d.ts.map +1 -1
- package/dist/meta-eval/index.js +3 -276
- package/dist/meta-eval/index.js.map +1 -1
- package/dist/{mint-C1W8w5wo.d.ts → mint-CDy-brUR.d.ts} +2 -2
- package/dist/{mint-C1W8w5wo.d.ts.map → mint-CDy-brUR.d.ts.map} +1 -1
- package/dist/{mint-CWG-s2mI.js → mint-Csor1BLJ.js} +2 -2
- package/dist/{mint-CWG-s2mI.js.map → mint-Csor1BLJ.js.map} +1 -1
- package/dist/multishot/golden/index.d.ts +1 -1
- package/dist/multishot/index.d.ts +2 -2
- package/dist/openapi.json +1 -1
- package/dist/{paired-arms-CgWT7sHy.js → paired-arms-DzKn0TfY.js} +3 -5
- package/dist/{paired-arms-CgWT7sHy.js.map → paired-arms-DzKn0TfY.js.map} +1 -1
- package/dist/{paired-binary-b2L3WKqr.js → paired-binary-Q36bhn2R.js} +13 -9
- package/dist/paired-binary-Q36bhn2R.js.map +1 -0
- package/dist/{paired-promotion-decision-2deJ7sDX.js → paired-promotion-decision-BN6IQTiu.js} +2 -2
- package/dist/{paired-promotion-decision-2deJ7sDX.js.map → paired-promotion-decision-BN6IQTiu.js.map} +1 -1
- package/dist/pipelines/index.d.ts +1 -1
- package/dist/pipelines/index.js +2 -2
- package/dist/{prime-protocol-Csr9kcOM.d.ts → prime-protocol-B7et3tI7.d.ts} +2 -2
- package/dist/{prime-protocol-Csr9kcOM.d.ts.map → prime-protocol-B7et3tI7.d.ts.map} +1 -1
- package/dist/{produced-state-Bzlh09oj.js → produced-state-CKybuR9F.js} +3 -3
- package/dist/produced-state-CKybuR9F.js.map +1 -0
- package/dist/profile-cell.d.ts +2 -2
- package/dist/profile-cell.js +2 -2
- package/dist/{promotion-policy-CCmyEH_c.js → promotion-policy-CWW35UWq.js} +3 -3
- package/dist/{promotion-policy-CCmyEH_c.js.map → promotion-policy-CWW35UWq.js.map} +1 -1
- package/dist/{query-DlV0pdx4.js → query-B1D1qp2Z.js} +2 -2
- package/dist/{query-DlV0pdx4.js.map → query-B1D1qp2Z.js.map} +1 -1
- package/dist/{query-DBUQhehH.d.ts → query-DK1nWcmp.d.ts} +2 -3
- package/dist/{query-DBUQhehH.d.ts.map → query-DK1nWcmp.d.ts.map} +1 -1
- package/dist/{registry-DMkmhBQ5.d.ts → registry-BL7H33tj.d.ts} +3 -3
- package/dist/{registry-DMkmhBQ5.d.ts.map → registry-BL7H33tj.d.ts.map} +1 -1
- package/dist/{release-confidence-CkoXFx9I.d.ts → release-confidence-C4kI7lQi.d.ts} +4 -12
- package/dist/release-confidence-C4kI7lQi.d.ts.map +1 -0
- package/dist/{release-confidence-CiFg1EPq.js → release-confidence-Dbe9AS_X.js} +2 -2
- package/dist/{release-confidence-CiFg1EPq.js.map → release-confidence-Dbe9AS_X.js.map} +1 -1
- package/dist/{report-command-pXtxg5YK.js → report-command-CtIIsQKP.js} +3 -28
- package/dist/{report-command-pXtxg5YK.js.map → report-command-CtIIsQKP.js.map} +1 -1
- package/dist/reporting.d.ts +3 -3
- package/dist/reporting.js +3 -3
- package/dist/{researcher-CZFZjj-u.d.ts → researcher-DuIJ9IaA.d.ts} +4 -4
- package/dist/{researcher-CZFZjj-u.d.ts.map → researcher-DuIJ9IaA.d.ts.map} +1 -1
- package/dist/{reward-hacking-Cyk-QHk9.d.ts → reward-hacking-C6Kp54Av.d.ts} +2 -2
- package/dist/{reward-hacking-Cyk-QHk9.d.ts.map → reward-hacking-C6Kp54Av.d.ts.map} +1 -1
- package/dist/reward-nw2xZGZG.js.map +1 -1
- package/dist/rl.d.ts +10 -29
- package/dist/rl.d.ts.map +1 -1
- package/dist/rl.js +7 -7
- package/dist/rl.js.map +1 -1
- package/dist/rollout/index.d.ts +3 -3
- package/dist/rollout/index.js +4 -4
- package/dist/{rollout-l9w3Gdvh.js → rollout-D9bKKBnZ.js} +30 -165
- package/dist/rollout-D9bKKBnZ.js.map +1 -0
- package/dist/{rubric-predictive-validity-DIUwJJ9Y.d.ts → rubric-predictive-validity-BiNNEjpG.d.ts} +2 -2
- package/dist/{rubric-predictive-validity-DIUwJJ9Y.d.ts.map → rubric-predictive-validity-BiNNEjpG.d.ts.map} +1 -1
- package/dist/{run-record-BRlM99rX.js → run-record-DlTZ_lqe.js} +3 -13
- package/dist/run-record-DlTZ_lqe.js.map +1 -0
- package/dist/{run-record-BB290Y9A.js → run-record-ES07wfz9.js} +2 -2
- package/dist/{run-record-BB290Y9A.js.map → run-record-ES07wfz9.js.map} +1 -1
- package/dist/{run-record-DF-fuu31.d.ts → run-record-p2AYTrCn.d.ts} +3 -7
- package/dist/run-record-p2AYTrCn.d.ts.map +1 -0
- package/dist/search/index.d.ts +2 -24
- package/dist/search/index.d.ts.map +1 -1
- package/dist/search/index.js +3 -3
- package/dist/{search-kernel-CIie6xig.js → search-kernel-BIXWM7YV.js} +4 -4
- package/dist/search-kernel-BIXWM7YV.js.map +1 -0
- package/dist/{search-kernel-uKdPVRAe.d.ts → search-kernel-ah2GESe3.d.ts} +3 -3
- package/dist/{search-kernel-uKdPVRAe.d.ts.map → search-kernel-ah2GESe3.d.ts.map} +1 -1
- package/dist/{search-ledger-CXQPwdWz.js → search-ledger-aMTNQn_m.js} +2 -2
- package/dist/{search-ledger-CXQPwdWz.js.map → search-ledger-aMTNQn_m.js.map} +1 -1
- package/dist/{search-shipper-DhhofkDX.js → search-shipper-DKXWuLnp.js} +2 -2
- package/dist/{search-shipper-DhhofkDX.js.map → search-shipper-DKXWuLnp.js.map} +1 -1
- package/dist/{search-summary-BV5uehLd.js → search-summary-nwrtXxC_.js} +3 -3
- package/dist/{search-summary-BV5uehLd.js.map → search-summary-nwrtXxC_.js.map} +1 -1
- package/dist/{semantic-concept-judge-D1wH-YKc.js → semantic-concept-judge-BQxZyXdN.js} +3 -3
- package/dist/{semantic-concept-judge-D1wH-YKc.js.map → semantic-concept-judge-BQxZyXdN.js.map} +1 -1
- package/dist/{sequential-DMqr3sni.js → sequential-D-1PPCmM.js} +3 -3
- package/dist/{sequential-DMqr3sni.js.map → sequential-D-1PPCmM.js.map} +1 -1
- package/dist/{skillopt-optimization-method-Dcnh5Pvr.js → skillopt-optimization-method-DujPOiwD.js} +5 -5
- package/dist/{skillopt-optimization-method-Dcnh5Pvr.js.map → skillopt-optimization-method-DujPOiwD.js.map} +1 -1
- package/dist/{statistical-heldout-BP6iEJvR.js → statistical-heldout-1Gbmex7v.js} +2 -2
- package/dist/{statistical-heldout-BP6iEJvR.js.map → statistical-heldout-1Gbmex7v.js.map} +1 -1
- package/dist/{statistical-heldout-DFKtp-Qm.d.ts → statistical-heldout-8RGPIeyo.d.ts} +17 -14
- package/dist/statistical-heldout-8RGPIeyo.d.ts.map +1 -0
- package/dist/{store-otlp-BA7G1Du-.js → store-otlp-__ebQ6Hn.js} +2 -2
- package/dist/{store-otlp-BA7G1Du-.js.map → store-otlp-__ebQ6Hn.js.map} +1 -1
- package/dist/{store-tool-spans-DKoVbE-w.js → store-tool-spans-TbRBuIM7.js} +3 -3
- package/dist/{store-tool-spans-DKoVbE-w.js.map → store-tool-spans-TbRBuIM7.js.map} +1 -1
- package/dist/{store-tool-spans-9YJNMP-z.d.ts → store-tool-spans-wQ9idsnn.d.ts} +5 -5
- package/dist/{store-tool-spans-9YJNMP-z.d.ts.map → store-tool-spans-wQ9idsnn.d.ts.map} +1 -1
- package/dist/{summary-report-DvvesgR_.d.ts → summary-report-CK0ueoJd.d.ts} +2 -2
- package/dist/{summary-report-DvvesgR_.d.ts.map → summary-report-CK0ueoJd.d.ts.map} +1 -1
- package/dist/{summary-report-fXurE2TR.js → summary-report-DtdRucM0.js} +3 -3
- package/dist/{summary-report-fXurE2TR.js.map → summary-report-DtdRucM0.js.map} +1 -1
- package/dist/supervisor-run/index.d.ts +0 -8
- package/dist/supervisor-run/index.d.ts.map +1 -1
- package/dist/supervisor-run/index.js +36 -71
- package/dist/supervisor-run/index.js.map +1 -1
- package/dist/{systemone-review-Bt4JcxMc.d.ts → systemone-review-BReh3mvm.d.ts} +2 -2
- package/dist/{systemone-review-Bt4JcxMc.d.ts.map → systemone-review-BReh3mvm.d.ts.map} +1 -1
- package/dist/systemone.d.ts +3 -3
- package/dist/{tool-groups-j19pmOfP.d.ts → tool-groups-v28A-ZAF.d.ts} +3 -3
- package/dist/tool-groups-v28A-ZAF.d.ts.map +1 -0
- package/dist/{tool-waste-x_E8h9Ih.js → tool-waste-BYt5rdt8.js} +2 -2
- package/dist/{tool-waste-x_E8h9Ih.js.map → tool-waste-BYt5rdt8.js.map} +1 -1
- package/dist/{tool-waste-DxlebQkx.d.ts → tool-waste-Wi4g6Ekb.d.ts} +2 -2
- package/dist/{tool-waste-DxlebQkx.d.ts.map → tool-waste-Wi4g6Ekb.d.ts.map} +1 -1
- package/dist/{trace-questions-dRvmsbTk.d.ts → trace-questions-C26UvI3D.d.ts} +3 -3
- package/dist/{trace-questions-dRvmsbTk.d.ts.map → trace-questions-C26UvI3D.d.ts.map} +1 -1
- package/dist/trace-repair/index.d.ts +2 -9
- package/dist/trace-repair/index.d.ts.map +1 -1
- package/dist/trace-repair/index.js +1 -1
- package/dist/traces.d.ts +6 -8
- package/dist/traces.d.ts.map +1 -1
- package/dist/traces.js +6 -6
- package/dist/trajectory-replay/index.d.ts +0 -2
- package/dist/trajectory-replay/index.d.ts.map +1 -1
- package/dist/trajectory-replay/index.js +1 -1
- package/dist/{tree-CJOXbVcv.js → tree-kc1d80yF.js} +5 -25
- package/dist/tree-kc1d80yF.js.map +1 -0
- package/dist/{types-CFCUZk0B.d.ts → types-Dv9sYwBB.d.ts} +4 -4
- package/dist/{types-CFCUZk0B.d.ts.map → types-Dv9sYwBB.d.ts.map} +1 -1
- package/dist/{types-DAPpAkd4.d.ts → types-KEkKkETv.d.ts} +2 -2
- package/dist/{types-DAPpAkd4.d.ts.map → types-KEkKkETv.d.ts.map} +1 -1
- package/dist/{types-BlSBUNRd.d.ts → types-tWFY6bON.d.ts} +3 -3
- package/dist/{types-BlSBUNRd.d.ts.map → types-tWFY6bON.d.ts.map} +1 -1
- package/dist/wire/index.d.ts +1 -1
- package/docs/concepts.md +13 -0
- package/docs/experiment.md +10 -0
- package/docs/public-api.md +399 -424
- package/docs/trace-analysis.md +3 -0
- package/package.json +6 -5
- package/dist/agent-profile-cell-BHcB-mZX.d.ts.map +0 -1
- package/dist/agent-profile-cell-DcFW4vOc.js.map +0 -1
- package/dist/analyze-runs-GZ0VSwui.js.map +0 -1
- package/dist/campaign-evidence-DW3Krxov.js.map +0 -1
- package/dist/claude-jsonl-CG3JUbTJ.js +0 -211
- package/dist/claude-jsonl-CG3JUbTJ.js.map +0 -1
- package/dist/engine-DProXJ7A.d.ts.map +0 -1
- package/dist/llm-judge-HLWASwUJ.js.map +0 -1
- package/dist/paired-binary-b2L3WKqr.js.map +0 -1
- package/dist/produced-state-Bzlh09oj.js.map +0 -1
- package/dist/release-confidence-CkoXFx9I.d.ts.map +0 -1
- package/dist/rollout-l9w3Gdvh.js.map +0 -1
- package/dist/run-record-BRlM99rX.js.map +0 -1
- package/dist/run-record-DF-fuu31.d.ts.map +0 -1
- package/dist/search-kernel-CIie6xig.js.map +0 -1
- package/dist/series-convergence-CjO2QdRW.js +0 -43
- package/dist/series-convergence-CjO2QdRW.js.map +0 -1
- package/dist/series-convergence-hEfKdoi-.d.ts +0 -126
- package/dist/series-convergence-hEfKdoi-.d.ts.map +0 -1
- package/dist/statistical-heldout-DFKtp-Qm.d.ts.map +0 -1
- package/dist/tool-groups-j19pmOfP.d.ts.map +0 -1
- package/dist/tree-CJOXbVcv.js.map +0 -1
|
@@ -1,4 +1,4 @@
|
|
|
1
|
-
import {
|
|
1
|
+
import { c as validateRunRecord } from "./run-record-DlTZ_lqe.js";
|
|
2
2
|
//#region src/campaign/run-record.ts
|
|
3
3
|
/**
|
|
4
4
|
* A campaign cell carried a judge score without a `dimensions` record.
|
|
@@ -220,4 +220,4 @@ function mean(values) {
|
|
|
220
220
|
//#endregion
|
|
221
221
|
export { campaignCellToRunRecord as n, projectCampaignCellQuality as r, campaignCellCostProvenance as t };
|
|
222
222
|
|
|
223
|
-
//# sourceMappingURL=run-record-
|
|
223
|
+
//# sourceMappingURL=run-record-ES07wfz9.js.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"run-record-BB290Y9A.js","names":[],"sources":["../src/campaign/run-record.ts"],"sourcesContent":["import type { AgentProfileCell } from '../agent-profile-cell'\nimport type { CostProvenance } from '../cost-ledger'\nimport type {\n JudgeScoresRecord,\n RunCostProvenance,\n RunOutcome,\n RunRecord,\n RunSearchCoordinates,\n RunSplitTag,\n RunTerminalOutcome,\n RunTraceRef,\n} from '../run-record'\nimport { validateRunRecord } from '../run-record'\nimport type { CampaignCellResult, JudgeScore } from './types'\n\nexport interface CampaignCellRunRecordOptions {\n /** With `search`, this must be `searchCellRunId(search)`. */\n runId: string\n experimentId: string\n candidateId: string\n model: string\n promptHash: string\n configHash: string\n commitSha: string\n splitTag: RunSplitTag\n seed?: number\n scenarioId?: string\n /** Estimated total for a cell that proved no spend at all. A cell that\n * proved part of its spend keeps that floor instead. */\n defaultCostUsd?: number\n agentProfile?: AgentProfileCell\n raw?: Record<string, number>\n search?: RunSearchCoordinates\n traceRef?: RunTraceRef\n}\n\nexport interface CampaignCellQualityProjection {\n score?: number\n judgeScores?: JudgeScoresRecord\n successfulJudgeScores: Record<string, JudgeScore>\n failedJudges: string[]\n raw: Record<string, number>\n}\n\n/**\n * A campaign cell carried a judge score without a `dimensions` record.\n *\n * Two exported types share the name `JudgeScore`: the campaign verdict\n * (`{ dimensions, composite, notes }` from `@tangle-network/agent-eval/campaign`)\n * and the root export's flat per-dimension row (`{ judgeName, dimension, score }`\n * from `@tangle-network/agent-eval`). Campaign aggregation accepts only the\n * campaign shape; the flat shape previously crashed here with an opaque\n * TypeError deep inside aggregation.\n */\nexport class CampaignJudgeScoreShapeError extends TypeError {\n constructor(judgeName: string) {\n super(\n `campaign cell judge '${judgeName}' carries a score without a 'dimensions' record. ` +\n `Use the campaign JudgeScore ({ dimensions, composite, notes }) from ` +\n `'@tangle-network/agent-eval/campaign'; the root export's JudgeScore ` +\n `({ judgeName, dimension, score }) is a different type with the same name.`,\n )\n this.name = 'CampaignJudgeScoreShapeError'\n }\n}\n\nexport interface CampaignCellExecutionEvidence {\n terminalOutcome: RunTerminalOutcome\n executionErrorCount?: number\n judgeErrorCount?: number\n unclassifiedErrorCount?: number\n terminalFailureReason?: string\n}\n\n/**\n * Project one campaign cell into the canonical run format.\n *\n * A dispatch error establishes terminal execution failure. A judge error only\n * establishes that quality measurement failed after dispatch completed.\n * Failures without a stage remain unknown. No failure becomes a zero-quality\n * label.\n */\nexport function campaignCellToRunRecord<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n options: CampaignCellRunRecordOptions,\n): RunRecord {\n const quality = projectCampaignCellQuality(cell)\n const execution = campaignCellExecutionEvidence(cell)\n const judgeErrorCount = Math.max(\n quality.raw.judge_error_count ?? 0,\n execution.judgeErrorCount ?? 0,\n )\n const costProvenance = campaignCellRunCostProvenance(cell, options.defaultCostUsd)\n const costUsd = costProvenance.usd\n const raw: Record<string, number> = {\n ...finiteMetrics(options.raw),\n ...quality.raw,\n rep: cell.rep,\n duration_ms: cell.durationMs,\n ...(costUsd === null ? {} : { cost_usd: costUsd }),\n tokens_input: cell.tokenUsage.input,\n tokens_output: cell.tokenUsage.output,\n tokens_known: cell.tokenUsage.tokensKnown === false ? 0 : 1,\n latency_ms: cell.durationMs,\n ...(execution.executionErrorCount === undefined\n ? {}\n : { execution_error_count: execution.executionErrorCount }),\n ...(judgeErrorCount > 0 ? { judge_error_count: judgeErrorCount } : {}),\n ...(execution.unclassifiedErrorCount === undefined\n ? {}\n : { unclassified_error_count: execution.unclassifiedErrorCount }),\n }\n if (typeof cell.generation === 'number') raw.generation = cell.generation\n if (cell.tokenUsage.reasoning !== undefined) {\n raw.tokens_reasoning = cell.tokenUsage.reasoning\n }\n if (cell.tokenUsage.cached !== undefined) raw.tokens_cached = cell.tokenUsage.cached\n if (cell.tokenUsage.cacheWrite !== undefined) {\n raw.tokens_cache_write = cell.tokenUsage.cacheWrite\n }\n if (cell.tokenUsage.tokensKnown !== false && costUsd !== null && costUsd > 0) {\n raw.tokens_per_dollar = (cell.tokenUsage.input + cell.tokenUsage.output) / costUsd\n }\n if (costUsd !== null && quality.score !== undefined && quality.score > 0.01) {\n raw.cost_per_quality = costUsd / quality.score\n }\n\n const outcome: RunOutcome = {\n raw,\n ...(quality.judgeScores ? { judgeScores: quality.judgeScores } : {}),\n }\n if (quality.score !== undefined) {\n if (options.splitTag === 'holdout') outcome.holdoutScore = quality.score\n else outcome.searchScore = quality.score\n }\n\n return validateRunRecord({\n runId: options.runId,\n experimentId: options.experimentId,\n candidateId: options.candidateId,\n seed: options.seed ?? cell.seed,\n model: options.model,\n promptHash: options.promptHash,\n configHash: options.configHash,\n commitSha: options.commitSha,\n wallMs: cell.durationMs,\n costUsd,\n costProvenance,\n tokenUsage: { ...cell.tokenUsage },\n terminalOutcome: execution.terminalOutcome,\n ...(execution.terminalFailureReason\n ? { terminalFailureReason: execution.terminalFailureReason }\n : {}),\n outcome,\n splitTag: options.splitTag,\n scenarioId: options.scenarioId ?? cell.scenarioId,\n ...(options.agentProfile ? { agentProfile: options.agentProfile } : {}),\n ...(options.search ? { search: options.search } : {}),\n ...(options.traceRef ? { traceRef: options.traceRef } : {}),\n })\n}\n\n/**\n * A cell's cost as a RunRecord states it. An uncaptured cell's `costUsd` is\n * the spend its receipts proved, so a positive one is the run's floor. The\n * caller's `defaultCostUsd` stands in only when nothing was proved: a proven\n * floor is never replaced by a guess.\n */\nfunction campaignCellRunCostProvenance<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n defaultCostUsd: number | undefined,\n): RunCostProvenance {\n const provenance = campaignCellCostProvenance(cell)\n if (provenance.kind !== 'uncaptured') return provenance\n if (cell.costUsd > 0) return { kind: 'lower-bound', usd: null, knownLowerBoundUsd: cell.costUsd }\n return defaultCostUsd === undefined ? provenance : { kind: 'estimated', usd: defaultCostUsd }\n}\n\n/**\n * Validate the cost fields that cross campaign cache and RunRecord boundaries.\n * `costUsd` is a known subtotal for uncaptured cells, but it must equal the\n * authoritative total whenever that total is observed or estimated.\n */\nexport function campaignCellCostProvenance<TArtifact>(\n cell: Pick<CampaignCellResult<TArtifact>, 'cellId' | 'costUsd' | 'costProvenance'>,\n): CostProvenance {\n if (!Number.isFinite(cell.costUsd) || cell.costUsd < 0) {\n throw new Error(`campaign cell '${cell.cellId}' has invalid costUsd`)\n }\n const provenance = cell.costProvenance\n if (!provenance || typeof provenance !== 'object') {\n throw new Error(`campaign cell '${cell.cellId}' has no costProvenance`)\n }\n if (provenance.kind === 'uncaptured') {\n if (provenance.usd !== null) {\n throw new Error(`campaign cell '${cell.cellId}' has invalid uncaptured costProvenance`)\n }\n return { kind: 'uncaptured', usd: null }\n }\n if (\n (provenance.kind !== 'observed' && provenance.kind !== 'estimated') ||\n !Number.isFinite(provenance.usd) ||\n provenance.usd < 0\n ) {\n throw new Error(`campaign cell '${cell.cellId}' has invalid costProvenance`)\n }\n if (provenance.usd !== cell.costUsd) {\n throw new Error(`campaign cell '${cell.cellId}' has costUsd inconsistent with costProvenance`)\n }\n return { kind: provenance.kind, usd: provenance.usd }\n}\n\nexport function campaignCellExecutionEvidence<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n): CampaignCellExecutionEvidence {\n if (cell.errorStage === 'dispatch') {\n return {\n terminalOutcome: 'failed',\n executionErrorCount: 1,\n ...(cell.error ? { terminalFailureReason: cell.error } : {}),\n }\n }\n if (cell.errorStage === 'judge') {\n return {\n terminalOutcome: 'succeeded',\n executionErrorCount: 0,\n judgeErrorCount: 1,\n }\n }\n if (!cell.error) {\n return { terminalOutcome: 'succeeded', executionErrorCount: 0 }\n }\n return {\n terminalOutcome: 'unknown',\n unclassifiedErrorCount: 1,\n }\n}\n\n/**\n * Produce the only task-quality view used by campaign aggregates and exports.\n *\n * Successful judge results remain available for diagnosis after another judge\n * fails, but a task score exists only for an error-free cell whose reported\n * judge values are all finite.\n */\nexport function projectCampaignCellQuality<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n): CampaignCellQualityProjection {\n if (cell.errorStage === 'dispatch') {\n return { successfulJudgeScores: {}, failedJudges: [], raw: {} }\n }\n\n const perJudge: Record<string, Record<string, number>> = {}\n const successfulJudgeScores: Record<string, JudgeScore> = {}\n const dimensionValues = new Map<string, number[]>()\n const composites: number[] = []\n const notes: string[] = []\n const failedJudges = new Set<string>(\n cell.errorStage === 'judge' ? [cell.errorJudge ?? 'unknown-judge'] : [],\n )\n const raw: Record<string, number> = {}\n\n for (const [judgeName, score] of Object.entries(cell.judgeScores)) {\n const dimensionsShape = (score as { dimensions?: unknown }).dimensions\n if (\n typeof dimensionsShape !== 'object' ||\n dimensionsShape === null ||\n Array.isArray(dimensionsShape)\n ) {\n throw new CampaignJudgeScoreShapeError(judgeName)\n }\n const finiteDimensions = Object.values(score.dimensions).every(Number.isFinite)\n if (score.failed || !Number.isFinite(score.composite) || !finiteDimensions) {\n failedJudges.add(judgeName)\n continue\n }\n\n composites.push(score.composite)\n successfulJudgeScores[judgeName] = score\n const dimensions = { ...score.dimensions }\n perJudge[judgeName] = dimensions\n for (const [dimension, value] of Object.entries(dimensions)) {\n raw[`${judgeName}.${dimension}`] = value\n const values = dimensionValues.get(dimension) ?? []\n values.push(value)\n dimensionValues.set(dimension, values)\n }\n if (score.notes) notes.push(`${judgeName}: ${score.notes}`)\n for (const failedJudge of score.failedJudges ?? []) {\n failedJudges.add(`${judgeName}/${failedJudge}`)\n }\n }\n\n if (failedJudges.size > 0) raw.judge_error_count = failedJudges.size\n const sortedFailedJudges = [...failedJudges].sort()\n if (composites.length === 0) {\n return {\n successfulJudgeScores,\n failedJudges: sortedFailedJudges,\n raw,\n }\n }\n\n const composite = mean(composites)\n const perDimMean = Object.fromEntries(\n [...dimensionValues.entries()].map(([dimension, values]) => [dimension, mean(values)]),\n )\n const complete =\n cell.error === undefined && cell.errorStage === undefined && failedJudges.size === 0\n if (complete) raw.composite = composite\n\n return {\n ...(complete ? { score: composite } : {}),\n raw,\n successfulJudgeScores,\n failedJudges: sortedFailedJudges,\n judgeScores: {\n perJudge,\n perDimMean,\n composite,\n ...(sortedFailedJudges.length > 0 ? { failedJudges: sortedFailedJudges } : {}),\n ...(notes.length > 0 ? { notes: notes.join(' | ') } : {}),\n },\n }\n}\n\n/** Read the canonical task score without recomputing cell quality. */\nexport function campaignCellTaskScore<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n): number | undefined {\n return projectCampaignCellQuality(cell).score\n}\n\n/** Read canonical successful judge dimensions without recomputing cell quality. */\nexport function campaignCellJudgeDimensions<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n): Record<string, Record<string, number>> {\n return projectCampaignCellQuality(cell).judgeScores?.perJudge ?? {}\n}\n\nfunction finiteMetrics(metrics: Record<string, number> | undefined): Record<string, number> {\n const finite: Record<string, number> = {}\n for (const [key, value] of Object.entries(metrics ?? {})) {\n if (Number.isFinite(value)) finite[key] = value\n }\n return finite\n}\n\nfunction mean(values: number[]): number {\n return values.reduce((sum, value) => sum + value, 0) / values.length\n}\n"],"mappings":";;;;;;;;;;;;AAsDA,IAAa,+BAAb,cAAkD,UAAU;CAC1D,YAAY,WAAmB;EAC7B,MACE,wBAAwB,UAAU,mQAIpC;EACA,KAAK,OAAO;CACd;AACF;;;;;;;;;AAkBA,SAAgB,wBACd,MACA,SACW;CACX,MAAM,UAAU,2BAA2B,IAAI;CAC/C,MAAM,YAAY,8BAA8B,IAAI;CACpD,MAAM,kBAAkB,KAAK,IAC3B,QAAQ,IAAI,qBAAqB,GACjC,UAAU,mBAAmB,CAC/B;CACA,MAAM,iBAAiB,8BAA8B,MAAM,QAAQ,cAAc;CACjF,MAAM,UAAU,eAAe;CAC/B,MAAM,MAA8B;EAClC,GAAG,cAAc,QAAQ,GAAG;EAC5B,GAAG,QAAQ;EACX,KAAK,KAAK;EACV,aAAa,KAAK;EAClB,GAAI,YAAY,OAAO,CAAC,IAAI,EAAE,UAAU,QAAQ;EAChD,cAAc,KAAK,WAAW;EAC9B,eAAe,KAAK,WAAW;EAC/B,cAAc,KAAK,WAAW,gBAAgB,QAAQ,IAAI;EAC1D,YAAY,KAAK;EACjB,GAAI,UAAU,wBAAwB,KAAA,IAClC,CAAC,IACD,EAAE,uBAAuB,UAAU,oBAAoB;EAC3D,GAAI,kBAAkB,IAAI,EAAE,mBAAmB,gBAAgB,IAAI,CAAC;EACpE,GAAI,UAAU,2BAA2B,KAAA,IACrC,CAAC,IACD,EAAE,0BAA0B,UAAU,uBAAuB;CACnE;CACA,IAAI,OAAO,KAAK,eAAe,UAAU,IAAI,aAAa,KAAK;CAC/D,IAAI,KAAK,WAAW,cAAc,KAAA,GAChC,IAAI,mBAAmB,KAAK,WAAW;CAEzC,IAAI,KAAK,WAAW,WAAW,KAAA,GAAW,IAAI,gBAAgB,KAAK,WAAW;CAC9E,IAAI,KAAK,WAAW,eAAe,KAAA,GACjC,IAAI,qBAAqB,KAAK,WAAW;CAE3C,IAAI,KAAK,WAAW,gBAAgB,SAAS,YAAY,QAAQ,UAAU,GACzE,IAAI,qBAAqB,KAAK,WAAW,QAAQ,KAAK,WAAW,UAAU;CAE7E,IAAI,YAAY,QAAQ,QAAQ,UAAU,KAAA,KAAa,QAAQ,QAAQ,KACrE,IAAI,mBAAmB,UAAU,QAAQ;CAG3C,MAAM,UAAsB;EAC1B;EACA,GAAI,QAAQ,cAAc,EAAE,aAAa,QAAQ,YAAY,IAAI,CAAC;CACpE;CACA,IAAI,QAAQ,UAAU,KAAA,GAAW;EAC/B,IAAI,QAAQ,aAAa,WAAW,QAAQ,eAAe,QAAQ;OAC9D,QAAQ,cAAc,QAAQ;CACrC;CAEA,OAAO,kBAAkB;EACvB,OAAO,QAAQ;EACf,cAAc,QAAQ;EACtB,aAAa,QAAQ;EACrB,MAAM,QAAQ,QAAQ,KAAK;EAC3B,OAAO,QAAQ;EACf,YAAY,QAAQ;EACpB,YAAY,QAAQ;EACpB,WAAW,QAAQ;EACnB,QAAQ,KAAK;EACb;EACA;EACA,YAAY,EAAE,GAAG,KAAK,WAAW;EACjC,iBAAiB,UAAU;EAC3B,GAAI,UAAU,wBACV,EAAE,uBAAuB,UAAU,sBAAsB,IACzD,CAAC;EACL;EACA,UAAU,QAAQ;EAClB,YAAY,QAAQ,cAAc,KAAK;EACvC,GAAI,QAAQ,eAAe,EAAE,cAAc,QAAQ,aAAa,IAAI,CAAC;EACrE,GAAI,QAAQ,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,CAAC;EACnD,GAAI,QAAQ,WAAW,EAAE,UAAU,QAAQ,SAAS,IAAI,CAAC;CAC3D,CAAC;AACH;;;;;;;AAQA,SAAS,8BACP,MACA,gBACmB;CACnB,MAAM,aAAa,2BAA2B,IAAI;CAClD,IAAI,WAAW,SAAS,cAAc,OAAO;CAC7C,IAAI,KAAK,UAAU,GAAG,OAAO;EAAE,MAAM;EAAe,KAAK;EAAM,oBAAoB,KAAK;CAAQ;CAChG,OAAO,mBAAmB,KAAA,IAAY,aAAa;EAAE,MAAM;EAAa,KAAK;CAAe;AAC9F;;;;;;AAOA,SAAgB,2BACd,MACgB;CAChB,IAAI,CAAC,OAAO,SAAS,KAAK,OAAO,KAAK,KAAK,UAAU,GACnD,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,sBAAsB;CAEtE,MAAM,aAAa,KAAK;CACxB,IAAI,CAAC,cAAc,OAAO,eAAe,UACvC,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,wBAAwB;CAExE,IAAI,WAAW,SAAS,cAAc;EACpC,IAAI,WAAW,QAAQ,MACrB,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,wCAAwC;EAExF,OAAO;GAAE,MAAM;GAAc,KAAK;EAAK;CACzC;CACA,IACG,WAAW,SAAS,cAAc,WAAW,SAAS,eACvD,CAAC,OAAO,SAAS,WAAW,GAAG,KAC/B,WAAW,MAAM,GAEjB,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,6BAA6B;CAE7E,IAAI,WAAW,QAAQ,KAAK,SAC1B,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,+CAA+C;CAE/F,OAAO;EAAE,MAAM,WAAW;EAAM,KAAK,WAAW;CAAI;AACtD;AAEA,SAAgB,8BACd,MAC+B;CAC/B,IAAI,KAAK,eAAe,YACtB,OAAO;EACL,iBAAiB;EACjB,qBAAqB;EACrB,GAAI,KAAK,QAAQ,EAAE,uBAAuB,KAAK,MAAM,IAAI,CAAC;CAC5D;CAEF,IAAI,KAAK,eAAe,SACtB,OAAO;EACL,iBAAiB;EACjB,qBAAqB;EACrB,iBAAiB;CACnB;CAEF,IAAI,CAAC,KAAK,OACR,OAAO;EAAE,iBAAiB;EAAa,qBAAqB;CAAE;CAEhE,OAAO;EACL,iBAAiB;EACjB,wBAAwB;CAC1B;AACF;;;;;;;;AASA,SAAgB,2BACd,MAC+B;CAC/B,IAAI,KAAK,eAAe,YACtB,OAAO;EAAE,uBAAuB,CAAC;EAAG,cAAc,CAAC;EAAG,KAAK,CAAC;CAAE;CAGhE,MAAM,WAAmD,CAAC;CAC1D,MAAM,wBAAoD,CAAC;CAC3D,MAAM,kCAAkB,IAAI,IAAsB;CAClD,MAAM,aAAuB,CAAC;CAC9B,MAAM,QAAkB,CAAC;CACzB,MAAM,eAAe,IAAI,IACvB,KAAK,eAAe,UAAU,CAAC,KAAK,cAAc,eAAe,IAAI,CAAC,CACxE;CACA,MAAM,MAA8B,CAAC;CAErC,KAAK,MAAM,CAAC,WAAW,UAAU,OAAO,QAAQ,KAAK,WAAW,GAAG;EACjE,MAAM,kBAAmB,MAAmC;EAC5D,IACE,OAAO,oBAAoB,YAC3B,oBAAoB,QACpB,MAAM,QAAQ,eAAe,GAE7B,MAAM,IAAI,6BAA6B,SAAS;EAElD,MAAM,mBAAmB,OAAO,OAAO,MAAM,UAAU,CAAC,CAAC,MAAM,OAAO,QAAQ;EAC9E,IAAI,MAAM,UAAU,CAAC,OAAO,SAAS,MAAM,SAAS,KAAK,CAAC,kBAAkB;GAC1E,aAAa,IAAI,SAAS;GAC1B;EACF;EAEA,WAAW,KAAK,MAAM,SAAS;EAC/B,sBAAsB,aAAa;EACnC,MAAM,aAAa,EAAE,GAAG,MAAM,WAAW;EACzC,SAAS,aAAa;EACtB,KAAK,MAAM,CAAC,WAAW,UAAU,OAAO,QAAQ,UAAU,GAAG;GAC3D,IAAI,GAAG,UAAU,GAAG,eAAe;GACnC,MAAM,SAAS,gBAAgB,IAAI,SAAS,KAAK,CAAC;GAClD,OAAO,KAAK,KAAK;GACjB,gBAAgB,IAAI,WAAW,MAAM;EACvC;EACA,IAAI,MAAM,OAAO,MAAM,KAAK,GAAG,UAAU,IAAI,MAAM,OAAO;EAC1D,KAAK,MAAM,eAAe,MAAM,gBAAgB,CAAC,GAC/C,aAAa,IAAI,GAAG,UAAU,GAAG,aAAa;CAElD;CAEA,IAAI,aAAa,OAAO,GAAG,IAAI,oBAAoB,aAAa;CAChE,MAAM,qBAAqB,CAAC,GAAG,YAAY,CAAC,CAAC,KAAK;CAClD,IAAI,WAAW,WAAW,GACxB,OAAO;EACL;EACA,cAAc;EACd;CACF;CAGF,MAAM,YAAY,KAAK,UAAU;CACjC,MAAM,aAAa,OAAO,YACxB,CAAC,GAAG,gBAAgB,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,WAAW,YAAY,CAAC,WAAW,KAAK,MAAM,CAAC,CAAC,CACvF;CACA,MAAM,WACJ,KAAK,UAAU,KAAA,KAAa,KAAK,eAAe,KAAA,KAAa,aAAa,SAAS;CACrF,IAAI,UAAU,IAAI,YAAY;CAE9B,OAAO;EACL,GAAI,WAAW,EAAE,OAAO,UAAU,IAAI,CAAC;EACvC;EACA;EACA,cAAc;EACd,aAAa;GACX;GACA;GACA;GACA,GAAI,mBAAmB,SAAS,IAAI,EAAE,cAAc,mBAAmB,IAAI,CAAC;GAC5E,GAAI,MAAM,SAAS,IAAI,EAAE,OAAO,MAAM,KAAK,KAAK,EAAE,IAAI,CAAC;EACzD;CACF;AACF;AAgBA,SAAS,cAAc,SAAqE;CAC1F,MAAM,SAAiC,CAAC;CACxC,KAAK,MAAM,CAAC,KAAK,UAAU,OAAO,QAAQ,WAAW,CAAC,CAAC,GACrD,IAAI,OAAO,SAAS,KAAK,GAAG,OAAO,OAAO;CAE5C,OAAO;AACT;AAEA,SAAS,KAAK,QAA0B;CACtC,OAAO,OAAO,QAAQ,KAAK,UAAU,MAAM,OAAO,CAAC,IAAI,OAAO;AAChE"}
|
|
1
|
+
{"version":3,"file":"run-record-ES07wfz9.js","names":[],"sources":["../src/campaign/run-record.ts"],"sourcesContent":["import type { AgentProfileCell } from '../agent-profile-cell'\nimport type { CostProvenance } from '../cost-ledger'\nimport type {\n JudgeScoresRecord,\n RunCostProvenance,\n RunOutcome,\n RunRecord,\n RunSearchCoordinates,\n RunSplitTag,\n RunTerminalOutcome,\n RunTraceRef,\n} from '../run-record'\nimport { validateRunRecord } from '../run-record'\nimport type { CampaignCellResult, JudgeScore } from './types'\n\nexport interface CampaignCellRunRecordOptions {\n /** With `search`, this must be `searchCellRunId(search)`. */\n runId: string\n experimentId: string\n candidateId: string\n model: string\n promptHash: string\n configHash: string\n commitSha: string\n splitTag: RunSplitTag\n seed?: number\n scenarioId?: string\n /** Estimated total for a cell that proved no spend at all. A cell that\n * proved part of its spend keeps that floor instead. */\n defaultCostUsd?: number\n agentProfile?: AgentProfileCell\n raw?: Record<string, number>\n search?: RunSearchCoordinates\n traceRef?: RunTraceRef\n}\n\nexport interface CampaignCellQualityProjection {\n score?: number\n judgeScores?: JudgeScoresRecord\n successfulJudgeScores: Record<string, JudgeScore>\n failedJudges: string[]\n raw: Record<string, number>\n}\n\n/**\n * A campaign cell carried a judge score without a `dimensions` record.\n *\n * Two exported types share the name `JudgeScore`: the campaign verdict\n * (`{ dimensions, composite, notes }` from `@tangle-network/agent-eval/campaign`)\n * and the root export's flat per-dimension row (`{ judgeName, dimension, score }`\n * from `@tangle-network/agent-eval`). Campaign aggregation accepts only the\n * campaign shape; the flat shape previously crashed here with an opaque\n * TypeError deep inside aggregation.\n */\nexport class CampaignJudgeScoreShapeError extends TypeError {\n constructor(judgeName: string) {\n super(\n `campaign cell judge '${judgeName}' carries a score without a 'dimensions' record. ` +\n `Use the campaign JudgeScore ({ dimensions, composite, notes }) from ` +\n `'@tangle-network/agent-eval/campaign'; the root export's JudgeScore ` +\n `({ judgeName, dimension, score }) is a different type with the same name.`,\n )\n this.name = 'CampaignJudgeScoreShapeError'\n }\n}\n\nexport interface CampaignCellExecutionEvidence {\n terminalOutcome: RunTerminalOutcome\n executionErrorCount?: number\n judgeErrorCount?: number\n unclassifiedErrorCount?: number\n terminalFailureReason?: string\n}\n\n/**\n * Project one campaign cell into the canonical run format.\n *\n * A dispatch error establishes terminal execution failure. A judge error only\n * establishes that quality measurement failed after dispatch completed.\n * Failures without a stage remain unknown. No failure becomes a zero-quality\n * label.\n */\nexport function campaignCellToRunRecord<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n options: CampaignCellRunRecordOptions,\n): RunRecord {\n const quality = projectCampaignCellQuality(cell)\n const execution = campaignCellExecutionEvidence(cell)\n const judgeErrorCount = Math.max(\n quality.raw.judge_error_count ?? 0,\n execution.judgeErrorCount ?? 0,\n )\n const costProvenance = campaignCellRunCostProvenance(cell, options.defaultCostUsd)\n const costUsd = costProvenance.usd\n const raw: Record<string, number> = {\n ...finiteMetrics(options.raw),\n ...quality.raw,\n rep: cell.rep,\n duration_ms: cell.durationMs,\n ...(costUsd === null ? {} : { cost_usd: costUsd }),\n tokens_input: cell.tokenUsage.input,\n tokens_output: cell.tokenUsage.output,\n tokens_known: cell.tokenUsage.tokensKnown === false ? 0 : 1,\n latency_ms: cell.durationMs,\n ...(execution.executionErrorCount === undefined\n ? {}\n : { execution_error_count: execution.executionErrorCount }),\n ...(judgeErrorCount > 0 ? { judge_error_count: judgeErrorCount } : {}),\n ...(execution.unclassifiedErrorCount === undefined\n ? {}\n : { unclassified_error_count: execution.unclassifiedErrorCount }),\n }\n if (typeof cell.generation === 'number') raw.generation = cell.generation\n if (cell.tokenUsage.reasoning !== undefined) {\n raw.tokens_reasoning = cell.tokenUsage.reasoning\n }\n if (cell.tokenUsage.cached !== undefined) raw.tokens_cached = cell.tokenUsage.cached\n if (cell.tokenUsage.cacheWrite !== undefined) {\n raw.tokens_cache_write = cell.tokenUsage.cacheWrite\n }\n if (cell.tokenUsage.tokensKnown !== false && costUsd !== null && costUsd > 0) {\n raw.tokens_per_dollar = (cell.tokenUsage.input + cell.tokenUsage.output) / costUsd\n }\n if (costUsd !== null && quality.score !== undefined && quality.score > 0.01) {\n raw.cost_per_quality = costUsd / quality.score\n }\n\n const outcome: RunOutcome = {\n raw,\n ...(quality.judgeScores ? { judgeScores: quality.judgeScores } : {}),\n }\n if (quality.score !== undefined) {\n if (options.splitTag === 'holdout') outcome.holdoutScore = quality.score\n else outcome.searchScore = quality.score\n }\n\n return validateRunRecord({\n runId: options.runId,\n experimentId: options.experimentId,\n candidateId: options.candidateId,\n seed: options.seed ?? cell.seed,\n model: options.model,\n promptHash: options.promptHash,\n configHash: options.configHash,\n commitSha: options.commitSha,\n wallMs: cell.durationMs,\n costUsd,\n costProvenance,\n tokenUsage: { ...cell.tokenUsage },\n terminalOutcome: execution.terminalOutcome,\n ...(execution.terminalFailureReason\n ? { terminalFailureReason: execution.terminalFailureReason }\n : {}),\n outcome,\n splitTag: options.splitTag,\n scenarioId: options.scenarioId ?? cell.scenarioId,\n ...(options.agentProfile ? { agentProfile: options.agentProfile } : {}),\n ...(options.search ? { search: options.search } : {}),\n ...(options.traceRef ? { traceRef: options.traceRef } : {}),\n })\n}\n\n/**\n * A cell's cost as a RunRecord states it. An uncaptured cell's `costUsd` is\n * the spend its receipts proved, so a positive one is the run's floor. The\n * caller's `defaultCostUsd` stands in only when nothing was proved: a proven\n * floor is never replaced by a guess.\n */\nfunction campaignCellRunCostProvenance<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n defaultCostUsd: number | undefined,\n): RunCostProvenance {\n const provenance = campaignCellCostProvenance(cell)\n if (provenance.kind !== 'uncaptured') return provenance\n if (cell.costUsd > 0) return { kind: 'lower-bound', usd: null, knownLowerBoundUsd: cell.costUsd }\n return defaultCostUsd === undefined ? provenance : { kind: 'estimated', usd: defaultCostUsd }\n}\n\n/**\n * Validate the cost fields that cross campaign cache and RunRecord boundaries.\n * `costUsd` is a known subtotal for uncaptured cells, but it must equal the\n * authoritative total whenever that total is observed or estimated.\n */\nexport function campaignCellCostProvenance<TArtifact>(\n cell: Pick<CampaignCellResult<TArtifact>, 'cellId' | 'costUsd' | 'costProvenance'>,\n): CostProvenance {\n if (!Number.isFinite(cell.costUsd) || cell.costUsd < 0) {\n throw new Error(`campaign cell '${cell.cellId}' has invalid costUsd`)\n }\n const provenance = cell.costProvenance\n if (!provenance || typeof provenance !== 'object') {\n throw new Error(`campaign cell '${cell.cellId}' has no costProvenance`)\n }\n if (provenance.kind === 'uncaptured') {\n if (provenance.usd !== null) {\n throw new Error(`campaign cell '${cell.cellId}' has invalid uncaptured costProvenance`)\n }\n return { kind: 'uncaptured', usd: null }\n }\n if (\n (provenance.kind !== 'observed' && provenance.kind !== 'estimated') ||\n !Number.isFinite(provenance.usd) ||\n provenance.usd < 0\n ) {\n throw new Error(`campaign cell '${cell.cellId}' has invalid costProvenance`)\n }\n if (provenance.usd !== cell.costUsd) {\n throw new Error(`campaign cell '${cell.cellId}' has costUsd inconsistent with costProvenance`)\n }\n return { kind: provenance.kind, usd: provenance.usd }\n}\n\nexport function campaignCellExecutionEvidence<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n): CampaignCellExecutionEvidence {\n if (cell.errorStage === 'dispatch') {\n return {\n terminalOutcome: 'failed',\n executionErrorCount: 1,\n ...(cell.error ? { terminalFailureReason: cell.error } : {}),\n }\n }\n if (cell.errorStage === 'judge') {\n return {\n terminalOutcome: 'succeeded',\n executionErrorCount: 0,\n judgeErrorCount: 1,\n }\n }\n if (!cell.error) {\n return { terminalOutcome: 'succeeded', executionErrorCount: 0 }\n }\n return {\n terminalOutcome: 'unknown',\n unclassifiedErrorCount: 1,\n }\n}\n\n/**\n * Produce the only task-quality view used by campaign aggregates and exports.\n *\n * Successful judge results remain available for diagnosis after another judge\n * fails, but a task score exists only for an error-free cell whose reported\n * judge values are all finite.\n */\nexport function projectCampaignCellQuality<TArtifact>(\n cell: CampaignCellResult<TArtifact>,\n): CampaignCellQualityProjection {\n if (cell.errorStage === 'dispatch') {\n return { successfulJudgeScores: {}, failedJudges: [], raw: {} }\n }\n\n const perJudge: Record<string, Record<string, number>> = {}\n const successfulJudgeScores: Record<string, JudgeScore> = {}\n const dimensionValues = new Map<string, number[]>()\n const composites: number[] = []\n const notes: string[] = []\n const failedJudges = new Set<string>(\n cell.errorStage === 'judge' ? [cell.errorJudge ?? 'unknown-judge'] : [],\n )\n const raw: Record<string, number> = {}\n\n for (const [judgeName, score] of Object.entries(cell.judgeScores)) {\n const dimensionsShape = (score as { dimensions?: unknown }).dimensions\n if (\n typeof dimensionsShape !== 'object' ||\n dimensionsShape === null ||\n Array.isArray(dimensionsShape)\n ) {\n throw new CampaignJudgeScoreShapeError(judgeName)\n }\n const finiteDimensions = Object.values(score.dimensions).every(Number.isFinite)\n if (score.failed || !Number.isFinite(score.composite) || !finiteDimensions) {\n failedJudges.add(judgeName)\n continue\n }\n\n composites.push(score.composite)\n successfulJudgeScores[judgeName] = score\n const dimensions = { ...score.dimensions }\n perJudge[judgeName] = dimensions\n for (const [dimension, value] of Object.entries(dimensions)) {\n raw[`${judgeName}.${dimension}`] = value\n const values = dimensionValues.get(dimension) ?? []\n values.push(value)\n dimensionValues.set(dimension, values)\n }\n if (score.notes) notes.push(`${judgeName}: ${score.notes}`)\n for (const failedJudge of score.failedJudges ?? []) {\n failedJudges.add(`${judgeName}/${failedJudge}`)\n }\n }\n\n if (failedJudges.size > 0) raw.judge_error_count = failedJudges.size\n const sortedFailedJudges = [...failedJudges].sort()\n if (composites.length === 0) {\n return {\n successfulJudgeScores,\n failedJudges: sortedFailedJudges,\n raw,\n }\n }\n\n const composite = mean(composites)\n const perDimMean = Object.fromEntries(\n [...dimensionValues.entries()].map(([dimension, values]) => [dimension, mean(values)]),\n )\n const complete =\n cell.error === undefined && cell.errorStage === undefined && failedJudges.size === 0\n if (complete) raw.composite = composite\n\n return {\n ...(complete ? { score: composite } : {}),\n raw,\n successfulJudgeScores,\n failedJudges: sortedFailedJudges,\n judgeScores: {\n perJudge,\n perDimMean,\n composite,\n ...(sortedFailedJudges.length > 0 ? { failedJudges: sortedFailedJudges } : {}),\n ...(notes.length > 0 ? { notes: notes.join(' | ') } : {}),\n },\n }\n}\n\nfunction finiteMetrics(metrics: Record<string, number> | undefined): Record<string, number> {\n const finite: Record<string, number> = {}\n for (const [key, value] of Object.entries(metrics ?? {})) {\n if (Number.isFinite(value)) finite[key] = value\n }\n return finite\n}\n\nfunction mean(values: number[]): number {\n return values.reduce((sum, value) => sum + value, 0) / values.length\n}\n"],"mappings":";;;;;;;;;;;;AAsDA,IAAa,+BAAb,cAAkD,UAAU;CAC1D,YAAY,WAAmB;EAC7B,MACE,wBAAwB,UAAU,mQAIpC;EACA,KAAK,OAAO;CACd;AACF;;;;;;;;;AAkBA,SAAgB,wBACd,MACA,SACW;CACX,MAAM,UAAU,2BAA2B,IAAI;CAC/C,MAAM,YAAY,8BAA8B,IAAI;CACpD,MAAM,kBAAkB,KAAK,IAC3B,QAAQ,IAAI,qBAAqB,GACjC,UAAU,mBAAmB,CAC/B;CACA,MAAM,iBAAiB,8BAA8B,MAAM,QAAQ,cAAc;CACjF,MAAM,UAAU,eAAe;CAC/B,MAAM,MAA8B;EAClC,GAAG,cAAc,QAAQ,GAAG;EAC5B,GAAG,QAAQ;EACX,KAAK,KAAK;EACV,aAAa,KAAK;EAClB,GAAI,YAAY,OAAO,CAAC,IAAI,EAAE,UAAU,QAAQ;EAChD,cAAc,KAAK,WAAW;EAC9B,eAAe,KAAK,WAAW;EAC/B,cAAc,KAAK,WAAW,gBAAgB,QAAQ,IAAI;EAC1D,YAAY,KAAK;EACjB,GAAI,UAAU,wBAAwB,KAAA,IAClC,CAAC,IACD,EAAE,uBAAuB,UAAU,oBAAoB;EAC3D,GAAI,kBAAkB,IAAI,EAAE,mBAAmB,gBAAgB,IAAI,CAAC;EACpE,GAAI,UAAU,2BAA2B,KAAA,IACrC,CAAC,IACD,EAAE,0BAA0B,UAAU,uBAAuB;CACnE;CACA,IAAI,OAAO,KAAK,eAAe,UAAU,IAAI,aAAa,KAAK;CAC/D,IAAI,KAAK,WAAW,cAAc,KAAA,GAChC,IAAI,mBAAmB,KAAK,WAAW;CAEzC,IAAI,KAAK,WAAW,WAAW,KAAA,GAAW,IAAI,gBAAgB,KAAK,WAAW;CAC9E,IAAI,KAAK,WAAW,eAAe,KAAA,GACjC,IAAI,qBAAqB,KAAK,WAAW;CAE3C,IAAI,KAAK,WAAW,gBAAgB,SAAS,YAAY,QAAQ,UAAU,GACzE,IAAI,qBAAqB,KAAK,WAAW,QAAQ,KAAK,WAAW,UAAU;CAE7E,IAAI,YAAY,QAAQ,QAAQ,UAAU,KAAA,KAAa,QAAQ,QAAQ,KACrE,IAAI,mBAAmB,UAAU,QAAQ;CAG3C,MAAM,UAAsB;EAC1B;EACA,GAAI,QAAQ,cAAc,EAAE,aAAa,QAAQ,YAAY,IAAI,CAAC;CACpE;CACA,IAAI,QAAQ,UAAU,KAAA,GAAW;EAC/B,IAAI,QAAQ,aAAa,WAAW,QAAQ,eAAe,QAAQ;OAC9D,QAAQ,cAAc,QAAQ;CACrC;CAEA,OAAO,kBAAkB;EACvB,OAAO,QAAQ;EACf,cAAc,QAAQ;EACtB,aAAa,QAAQ;EACrB,MAAM,QAAQ,QAAQ,KAAK;EAC3B,OAAO,QAAQ;EACf,YAAY,QAAQ;EACpB,YAAY,QAAQ;EACpB,WAAW,QAAQ;EACnB,QAAQ,KAAK;EACb;EACA;EACA,YAAY,EAAE,GAAG,KAAK,WAAW;EACjC,iBAAiB,UAAU;EAC3B,GAAI,UAAU,wBACV,EAAE,uBAAuB,UAAU,sBAAsB,IACzD,CAAC;EACL;EACA,UAAU,QAAQ;EAClB,YAAY,QAAQ,cAAc,KAAK;EACvC,GAAI,QAAQ,eAAe,EAAE,cAAc,QAAQ,aAAa,IAAI,CAAC;EACrE,GAAI,QAAQ,SAAS,EAAE,QAAQ,QAAQ,OAAO,IAAI,CAAC;EACnD,GAAI,QAAQ,WAAW,EAAE,UAAU,QAAQ,SAAS,IAAI,CAAC;CAC3D,CAAC;AACH;;;;;;;AAQA,SAAS,8BACP,MACA,gBACmB;CACnB,MAAM,aAAa,2BAA2B,IAAI;CAClD,IAAI,WAAW,SAAS,cAAc,OAAO;CAC7C,IAAI,KAAK,UAAU,GAAG,OAAO;EAAE,MAAM;EAAe,KAAK;EAAM,oBAAoB,KAAK;CAAQ;CAChG,OAAO,mBAAmB,KAAA,IAAY,aAAa;EAAE,MAAM;EAAa,KAAK;CAAe;AAC9F;;;;;;AAOA,SAAgB,2BACd,MACgB;CAChB,IAAI,CAAC,OAAO,SAAS,KAAK,OAAO,KAAK,KAAK,UAAU,GACnD,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,sBAAsB;CAEtE,MAAM,aAAa,KAAK;CACxB,IAAI,CAAC,cAAc,OAAO,eAAe,UACvC,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,wBAAwB;CAExE,IAAI,WAAW,SAAS,cAAc;EACpC,IAAI,WAAW,QAAQ,MACrB,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,wCAAwC;EAExF,OAAO;GAAE,MAAM;GAAc,KAAK;EAAK;CACzC;CACA,IACG,WAAW,SAAS,cAAc,WAAW,SAAS,eACvD,CAAC,OAAO,SAAS,WAAW,GAAG,KAC/B,WAAW,MAAM,GAEjB,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,6BAA6B;CAE7E,IAAI,WAAW,QAAQ,KAAK,SAC1B,MAAM,IAAI,MAAM,kBAAkB,KAAK,OAAO,+CAA+C;CAE/F,OAAO;EAAE,MAAM,WAAW;EAAM,KAAK,WAAW;CAAI;AACtD;AAEA,SAAgB,8BACd,MAC+B;CAC/B,IAAI,KAAK,eAAe,YACtB,OAAO;EACL,iBAAiB;EACjB,qBAAqB;EACrB,GAAI,KAAK,QAAQ,EAAE,uBAAuB,KAAK,MAAM,IAAI,CAAC;CAC5D;CAEF,IAAI,KAAK,eAAe,SACtB,OAAO;EACL,iBAAiB;EACjB,qBAAqB;EACrB,iBAAiB;CACnB;CAEF,IAAI,CAAC,KAAK,OACR,OAAO;EAAE,iBAAiB;EAAa,qBAAqB;CAAE;CAEhE,OAAO;EACL,iBAAiB;EACjB,wBAAwB;CAC1B;AACF;;;;;;;;AASA,SAAgB,2BACd,MAC+B;CAC/B,IAAI,KAAK,eAAe,YACtB,OAAO;EAAE,uBAAuB,CAAC;EAAG,cAAc,CAAC;EAAG,KAAK,CAAC;CAAE;CAGhE,MAAM,WAAmD,CAAC;CAC1D,MAAM,wBAAoD,CAAC;CAC3D,MAAM,kCAAkB,IAAI,IAAsB;CAClD,MAAM,aAAuB,CAAC;CAC9B,MAAM,QAAkB,CAAC;CACzB,MAAM,eAAe,IAAI,IACvB,KAAK,eAAe,UAAU,CAAC,KAAK,cAAc,eAAe,IAAI,CAAC,CACxE;CACA,MAAM,MAA8B,CAAC;CAErC,KAAK,MAAM,CAAC,WAAW,UAAU,OAAO,QAAQ,KAAK,WAAW,GAAG;EACjE,MAAM,kBAAmB,MAAmC;EAC5D,IACE,OAAO,oBAAoB,YAC3B,oBAAoB,QACpB,MAAM,QAAQ,eAAe,GAE7B,MAAM,IAAI,6BAA6B,SAAS;EAElD,MAAM,mBAAmB,OAAO,OAAO,MAAM,UAAU,CAAC,CAAC,MAAM,OAAO,QAAQ;EAC9E,IAAI,MAAM,UAAU,CAAC,OAAO,SAAS,MAAM,SAAS,KAAK,CAAC,kBAAkB;GAC1E,aAAa,IAAI,SAAS;GAC1B;EACF;EAEA,WAAW,KAAK,MAAM,SAAS;EAC/B,sBAAsB,aAAa;EACnC,MAAM,aAAa,EAAE,GAAG,MAAM,WAAW;EACzC,SAAS,aAAa;EACtB,KAAK,MAAM,CAAC,WAAW,UAAU,OAAO,QAAQ,UAAU,GAAG;GAC3D,IAAI,GAAG,UAAU,GAAG,eAAe;GACnC,MAAM,SAAS,gBAAgB,IAAI,SAAS,KAAK,CAAC;GAClD,OAAO,KAAK,KAAK;GACjB,gBAAgB,IAAI,WAAW,MAAM;EACvC;EACA,IAAI,MAAM,OAAO,MAAM,KAAK,GAAG,UAAU,IAAI,MAAM,OAAO;EAC1D,KAAK,MAAM,eAAe,MAAM,gBAAgB,CAAC,GAC/C,aAAa,IAAI,GAAG,UAAU,GAAG,aAAa;CAElD;CAEA,IAAI,aAAa,OAAO,GAAG,IAAI,oBAAoB,aAAa;CAChE,MAAM,qBAAqB,CAAC,GAAG,YAAY,CAAC,CAAC,KAAK;CAClD,IAAI,WAAW,WAAW,GACxB,OAAO;EACL;EACA,cAAc;EACd;CACF;CAGF,MAAM,YAAY,KAAK,UAAU;CACjC,MAAM,aAAa,OAAO,YACxB,CAAC,GAAG,gBAAgB,QAAQ,CAAC,CAAC,CAAC,KAAK,CAAC,WAAW,YAAY,CAAC,WAAW,KAAK,MAAM,CAAC,CAAC,CACvF;CACA,MAAM,WACJ,KAAK,UAAU,KAAA,KAAa,KAAK,eAAe,KAAA,KAAa,aAAa,SAAS;CACrF,IAAI,UAAU,IAAI,YAAY;CAE9B,OAAO;EACL,GAAI,WAAW,EAAE,OAAO,UAAU,IAAI,CAAC;EACvC;EACA;EACA,cAAc;EACd,aAAa;GACX;GACA;GACA;GACA,GAAI,mBAAmB,SAAS,IAAI,EAAE,cAAc,mBAAmB,IAAI,CAAC;GAC5E,GAAI,MAAM,SAAS,IAAI,EAAE,OAAO,MAAM,KAAK,KAAK,EAAE,IAAI,CAAC;EACzD;CACF;AACF;AAEA,SAAS,cAAc,SAAqE;CAC1F,MAAM,SAAiC,CAAC;CACxC,KAAK,MAAM,CAAC,KAAK,UAAU,OAAO,QAAQ,WAAW,CAAC,CAAC,GACrD,IAAI,OAAO,SAAS,KAAK,GAAG,OAAO,OAAO;CAE5C,OAAO;AACT;AAEA,SAAS,KAAK,QAA0B;CACtC,OAAO,OAAO,QAAQ,KAAK,UAAU,MAAM,OAAO,CAAC,IAAI,OAAO;AAChE"}
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { c as ValidationError } from "./errors-DEE6u6ot.js";
|
|
2
2
|
import { p as CostProvenance } from "./cost-ledger-DbQdN3nO.js";
|
|
3
|
-
import { r as AgentProfileCell } from "./agent-profile-cell-
|
|
3
|
+
import { r as AgentProfileCell } from "./agent-profile-cell-BNp-Js9F.js";
|
|
4
4
|
import { o as FailureClass } from "./schema-2g7trUwH.js";
|
|
5
5
|
//#region src/run-record.d.ts
|
|
6
6
|
/** Search/dev/holdout split tag. 'search' is the paper-grade alias for the
|
|
@@ -74,8 +74,6 @@ interface RunTraceRef {
|
|
|
74
74
|
* ran the same scenario shares.
|
|
75
75
|
*/
|
|
76
76
|
declare function searchCellRunId(search: Pick<RunSearchCoordinates, 'cellId' | 'attempt'>): string;
|
|
77
|
-
/** The run's proven spend: its total when known, its floor when only a floor is proven, else 0. */
|
|
78
|
-
declare function runCostFloorUsd(record: Pick<RunRecord, 'costUsd' | 'costProvenance'>): number;
|
|
79
77
|
interface RunJudgeMetadata {
|
|
80
78
|
model: string;
|
|
81
79
|
promptVersion: string;
|
|
@@ -285,8 +283,6 @@ declare function parseRunRecordSafe(input: unknown): {
|
|
|
285
283
|
ok: false;
|
|
286
284
|
error: RunRecordValidationError;
|
|
287
285
|
};
|
|
288
|
-
/** Round-trip helper — `JSON.parse(JSON.stringify(record))` then validate. */
|
|
289
|
-
declare function roundTripRunRecord(record: RunRecord): RunRecord;
|
|
290
286
|
/**
|
|
291
287
|
* Snapshot check for provider model identifiers. Accepts ISO and compact
|
|
292
288
|
* dates, Router's `-MMDD` snapshots, one opaque `@token`, and Vertex-style
|
|
@@ -295,5 +291,5 @@ declare function roundTripRunRecord(record: RunRecord): RunRecord;
|
|
|
295
291
|
*/
|
|
296
292
|
declare function modelHasSnapshot(model: string): boolean;
|
|
297
293
|
//#endregion
|
|
298
|
-
export {
|
|
299
|
-
//# sourceMappingURL=run-record-
|
|
294
|
+
export { runTaskScore as _, RunRecord as a, RunSplitTag as c, RunTokenUsage as d, RunTraceRef as f, parseRunRecordSafe as g, modelHasSnapshot as h, RunOutcome as i, RunTaskFailure as l, isRunRecord as m, RunCostProvenance as n, RunRecordValidationError as o, UNKNOWN_MODEL as p, RunJudgeMetadata as r, RunSearchCoordinates as s, JudgeScoresRecord as t, RunTerminalOutcome as u, searchCellRunId as v, validateRunRecord as y };
|
|
295
|
+
//# sourceMappingURL=run-record-p2AYTrCn.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"run-record-p2AYTrCn.d.ts","names":[],"sources":["../src/run-record.ts"],"mappings":";;;;;;;KAsCY;;;;;;;KAQA;;cAGC;UAEI;EACf;;EAEA;;;EAGA;;EAEA;;EAEA;;EAEA;;;;;;;;;;;;;;;KAgBU,oBACR;EACE;EAAqB;EAAW;;;;;;;UAOrB;EACf;EACA;;;EAGA;;EAEA;;;UAIe;;EAEf;;EAEA;;;;;;;iBAQc,gBAAgB,QAAQ,KAAK;UAI5B;EACf;EACA;;;EAGA;;;EAGA;;;;;;;;;;;;;;;;;;;;;UAsBe;;EAEf,UAAU,eAAe;;EAEzB,YAAY;;;EAGZ;;;;EAIA;;;EAGA;;UAGe;;;EAGf;;;EAGA;;;;EAIA,KAAK;;;;;;EAML,cAAc;;;;;;;EAOd;IAAa;IAAe;IAAgB;;;;;;;;;;;;;;;;;;;;;;UAsB7B;;EAEf;;;EAGA;;;EAGA;;;EAGA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA,gBAAgB;;EAEhB,YAAY;;EAEZ,iBAAiB;;;EAGjB;;EAEA,gBAAgB;;EAEhB,SAAS;;;;;EAKT,eAAe;;;EAGf;;EAEA,UAAU;;;;;EAKV;;;;;;;;EAQA,eAAe;;;EAGf,SAAS;;;EAGT,WAAW;;;;;;;;;KAUD;EACN;EAA0B;;EAC1B;EAAyB;;EAEzB,cAAc,QAAQ;EACtB;;;;;;;;;;iBAWU,aAAa,QAAQ;cAmCxB,iCAAiC;WACnC;EACT,YAAY,iBAAiB;;;;;;;iBAWf,kBAAkB,iBAAiB;;iBA2SnC,YAAY,iBAAiB,SAAS;;iBAUtC,mBACd;EACG;EAAU,OAAO;;EAAgB;EAAW,OAAO;;;;;;;;iBA2FxC,iBAAiB"}
|
package/dist/search/index.d.ts
CHANGED
|
@@ -1,12 +1,8 @@
|
|
|
1
1
|
import { A as SearchLedgerHash, C as SearchEdgeOperator, G as SearchSplit, H as SearchProposerKind, I as SearchNodeStatus, L as SearchOpenedEvent, T as SearchEstimateMethod, W as SearchSourceRef, i as SearchArtifactRef, j as SearchModelIdentity, l as SearchCandidateSurface, m as SearchClaim, t as NodeEstimate, y as SearchCloseReason } from "../search-ledger-types-p2xKfiDJ.js";
|
|
2
|
-
import { $ as SearchNode, it as SearchStateView } from "../types-
|
|
3
|
-
import { B as SearchPolicyView, N as SearchUnitExtension, R as SearchPolicy, at as OperatorYieldData, ct as OperatorYieldSignal, dt as LensSignal, ft as SampleSummary, it as OperatorOutcomeCounts, lt as operatorYield, mt as summarizeSamples, nt as ExpansionOperator, ot as OperatorYieldOptions, p as SearchRunResult, pt as rankingSplit, r as SearchArtifactCodec, rt as MIN_OUTCOMES_FOR_WEIGHT, s as SearchLane, st as OperatorYieldRow, t as RunSearchOptions, tt as EXPANSION_OPERATORS, ut as LensResult } from "../search-kernel-
|
|
2
|
+
import { $ as SearchNode, it as SearchStateView } from "../types-Dv9sYwBB.js";
|
|
3
|
+
import { B as SearchPolicyView, N as SearchUnitExtension, R as SearchPolicy, at as OperatorYieldData, ct as OperatorYieldSignal, dt as LensSignal, ft as SampleSummary, it as OperatorOutcomeCounts, lt as operatorYield, mt as summarizeSamples, nt as ExpansionOperator, ot as OperatorYieldOptions, p as SearchRunResult, pt as rankingSplit, r as SearchArtifactCodec, rt as MIN_OUTCOMES_FOR_WEIGHT, s as SearchLane, st as OperatorYieldRow, t as RunSearchOptions, tt as EXPANSION_OPERATORS, ut as LensResult } from "../search-kernel-ah2GESe3.js";
|
|
4
4
|
import { AgentProfileResourceRef } from "@tangle-network/agent-interface";
|
|
5
5
|
//#region src/search/lenses/edit-credit.d.ts
|
|
6
|
-
/** Signal name a policy reads. */
|
|
7
|
-
export declare const EDIT_CREDIT_SIGNAL = "reusableHunks";
|
|
8
|
-
/** The estimator every gene credit names. */
|
|
9
|
-
export declare const EDIT_CREDIT_ESTIMATOR: SearchSourceRef;
|
|
10
6
|
interface EditCreditOptions {
|
|
11
7
|
/**
|
|
12
8
|
* The parsed content a node's artifact reference names, or undefined when
|
|
@@ -411,9 +407,6 @@ type LandscapeEmbedding = {
|
|
|
411
407
|
};
|
|
412
408
|
/** The default: edits recorded between two profiles along the lineage. */
|
|
413
409
|
export declare function lineageEdits(): LandscapeEmbedding;
|
|
414
|
-
/** Surfaces whose content digests differ: the edit distance between two
|
|
415
|
-
* profiles at the granularity of their declared surfaces. */
|
|
416
|
-
export declare function surfaceDigestEdits(): LandscapeEmbedding;
|
|
417
410
|
/**
|
|
418
411
|
* Line edits between the surfaces' texts: for each declared surface, the
|
|
419
412
|
* fewest line insertions and deletions that turn one text into the other
|
|
@@ -425,13 +418,6 @@ export declare function surfaceDigestEdits(): LandscapeEmbedding;
|
|
|
425
418
|
export declare function surfaceTextEdits(read: (surface: SearchCandidateSurface) => unknown): LandscapeEmbedding;
|
|
426
419
|
/** A caller's vector per node, for example a model embedding of its profile. */
|
|
427
420
|
export declare function vectorEmbedding(name: string, vector: (node: SearchNode) => readonly number[] | null): LandscapeEmbedding;
|
|
428
|
-
/**
|
|
429
|
-
* JSON as lines a diff can count: one line per scalar leaf, `path=value`, and
|
|
430
|
-
* one line per line of a string, so an edit inside a long prompt counts the
|
|
431
|
-
* lines it changed rather than the whole prompt. Object keys are visited in
|
|
432
|
-
* code-unit order.
|
|
433
|
-
*/
|
|
434
|
-
export declare function profileTextLines(value: unknown, path?: string): string[];
|
|
435
421
|
/** Fewest line insertions plus deletions turning `a` into `b` (Myers' O(ND)
|
|
436
422
|
* greedy algorithm, which is fast when the texts are close). */
|
|
437
423
|
export declare function lineEditDistance(a: readonly string[], b: readonly string[]): number;
|
|
@@ -581,12 +567,6 @@ interface LandscapeData {
|
|
|
581
567
|
* signal. Pure: it reads the state and calls `embed`.
|
|
582
568
|
*/
|
|
583
569
|
export declare function landscape(state: SearchStateView, embed?: LandscapeEmbedding, options?: LandscapeOptions): GeometryLensResult<LandscapeData>;
|
|
584
|
-
/**
|
|
585
|
-
* The symmetric k-nearest-neighbour graph of points in any dimension: each
|
|
586
|
-
* point joins its k nearest (ties by index) and every edge is kept in both
|
|
587
|
-
* directions. Brute force, so callers keep n in the low thousands.
|
|
588
|
-
*/
|
|
589
|
-
export declare function nearestNeighbours(points: readonly (readonly number[])[], k: number): number[][];
|
|
590
570
|
/** The landscape as text for a terminal or a proposer's context: the same
|
|
591
571
|
* numbers the view draws, with a coarse shaded map. */
|
|
592
572
|
export declare function formatLandscape(lens: GeometryLensResult<LandscapeData>): string;
|
|
@@ -816,8 +796,6 @@ interface MetaSearchOptions {
|
|
|
816
796
|
* null until one is named. */
|
|
817
797
|
objective?: string;
|
|
818
798
|
}
|
|
819
|
-
/** The signal's name, as a policy or a view reads it. */
|
|
820
|
-
export declare const META_SEARCH_SIGNAL = "metaSearch.bestPolicyConfiguration";
|
|
821
799
|
/** The per-search score `metaSearchScore` computes. An outer search names it
|
|
822
800
|
* as its objective's judge. */
|
|
823
801
|
export declare const META_SEARCH_SCORE_SOURCE: SearchSourceRef;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/search/lenses/edit-credit.ts","../../src/search/lenses/front.ts","../../src/search/lenses/geometry.ts","../../src/search/lenses/plateau.ts","../../src/search/lenses/landscape.ts","../../src/search/lenses/meta-search.ts","../../src/search/lenses/skill-manifold.ts","../../src/search/lenses/task-matrix.ts","../../src/search/lenses/tree.ts","../../src/search/nested-search.ts"],"mappings":";;;;;;
|
|
1
|
+
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/search/lenses/edit-credit.ts","../../src/search/lenses/front.ts","../../src/search/lenses/geometry.ts","../../src/search/lenses/plateau.ts","../../src/search/lenses/landscape.ts","../../src/search/lenses/meta-search.ts","../../src/search/lenses/skill-manifold.ts","../../src/search/lenses/task-matrix.ts","../../src/search/lenses/tree.ts","../../src/search/nested-search.ts"],"mappings":";;;;;UA+FiB;;;;;;EAMf,aAAa,KAAK;;;EAGlB,QAAQ;;;EAGR;;KAGU;UAEK;EACf;EACA;EACA,UAAU;EACV;;EAEA;;UAGe;EACf;EACA;;;EAGA;;EAEA;;EAEA;;;;EAIA,YAAY;;EAEZ;;;EAGA;;EAEA;;;EAGA;EACA;;IAEE;;IAEA;;IAEA;;IAEA;;;;EAIF,QAAQ;EACR,SAAS;;;;EAIT;;UAGe;EACf;;EAEA;IACE;IACA;IACA;IACA;;;EAGF;EACA,QAAQ;;;;EAIR;;;EAGA;EACA;;EAEA;;EAEA;EACA;;;EAGA;;UAGe;;;;EAIf,UAAU,QAAQ;IAA2B;;EAC7C;IAAkB;IAAmB;MAAU;;;;EAE/C;EACA;EACA,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;EACA,UAAU;EACV,QAAQ;;EAER;EACA;;UAGe;EACf,OAAO;EACP;EACA;;IAEE;;IAEA;;IAEA,SAAS;;;;EAIX;IAAS;IAAe;;EACxB;IAAS;IAAe;IAAwB;;;EAEhD,OAAO;EACP,QAAQ,OAAO;;;EAGf,YAAY,OAAO;EACnB;IACE;;IAEA;;IAEA;IACA;IACA;;IAEA,OAAO;;EAET,iBAAiB;;;EAGjB;;IAEE;;;IAGA;IACA;;;;EAIF,SAAS;EACT;IACE,WAAW;IACX;IACA;IACA;IACA;;IAEA;;;;;UAMa,yBAAyB;;EAExC;;;EAGA;;KAGU,mBAAmB,WAAW;EACxC,QAAQ;;;;;;wBA0CM,WAAW,OAAO,iBAAiB,SAAS,oBAAoB;;;wBAs8BhE,eAAe,QAAQ,kBAAkB;EAAW;;;;UCtvCnD;EACf;EACA;;;;EAIA,MAAM,OAAO,iBAAiB;;UAGf;EACf;;;EAGA;;EAEA;EACA,QAAQ;;;EAGR;;;EAGA;;;EAGA;;;EAGA;;EAEA,UAAU;;KAGA;UAEK;EACf;EACA;EACA,MAAM;EACN;;EAEA;;EAEA,YAAY,OAAO;;UAGJ;EACf;EACA,OAAO;;;;;KAMG,cAAc;wBAEV,MACd,OAAO,iBACP,UAAS;EACN,MAAM;EAAW;IAAU;IAAc,OAAO;;;;;;;UClEpC,uBAAuB;;;EAGtC;;EAEA;;EAEA;;EAEA;;UAGe,mBAAmB,eAAe,WAAW;;EAE5D;EACA;;EAEA;EACA,QAAQ;;;;;;;;;wBAaM,cACd,OAAO,iBACP;EAAW;;;;;KCtCD,oBAAoB,KAC9B;UAIe;;EAEf;;UAGe;;;EAGf;EACA;;;;EAIA;;EAEA;;;EAGA;;;EAGA;EACA;IAAQ;IAAgB;IAAc;;;;EAGtC;EACA;EACA;EACA;;;;;;;wBAUc,cACd,MAAM,mBACN,UAAS,uBACR;UA6Ec;;;EAGf;;;EAGA;;;;;;;;;;;;;;;;;;;;;;wBAuBc,eACd,MAAM,cACN,UAAS,wBACR;;;;;;;;;;;;;;KCzGS;EACN;EAAiB;EAAc;;EAE/B;EACA;EACA;EACA,SAAS,GAAG,YAAY,GAAG;;EAG3B;EACA;EACA;EACA,OAAO,MAAM;;;wBAIH,gBAAgB;;;;;;;;;wBAiBhB,iBACd,OAAO,SAAS,qCACf;;wBAoDa,gBACd,cACA,SAAS,MAAM,0CACd;;;wBAsCa,iBAAiB,sBAAsB;UA2BtC;;EAEf;;EAEA;;EAEA;;;EAGA;;EAEA;;EAEA;;UAGe;EACf;EACA;EACA;EACA,QAAQ;;EAER;;EAEA;EACA;EACA;;;EAGA;;;EAGA;EACA;EACA,QAAQ;;;EAGR;;UAGe;EACf;EACA;EACA;EACA;;;;EAIA;;EAEA;EACA;;;;EAIA;IACE;IACA;IACA;;IAEA;;;;IAIA;;IAEA;;EAEF;;UAGe;;EAEf;IAAQ;IAAgB;IAAW;IAAW;IAAe;;;;;EAI7D;;;EAGA;EACA;EACA;;UAGe;EACf;EACA;EACA;IACE;IACA;IACA;;;IAGA;MAAoB;MAAgB;;;IAEpC;;IAEA;;;IAGA;IACA;IACA;;EAEF,OAAO;;EAEP,SAAS;IAAQ;IAAgB;IAAe;;EAChD,MAAM;EACN;EACA;IACE;;IAEA;;IAEA;;;IAGA;;IAEA;;;IAGA;IACA;IACA;IACA;IACA,OAAO;;EAET,SAAS;;;;;;;wBAmBK,UACd,OAAO,iBACP,QAAO,oBACP,UAAS,mBACR,mBAAmB;;;wBAi5BN,gBAAgB,MAAM,mBAAmB;;;;UCrsCxC;;;;EAIf;;EAEA;EACA;IACE;IACA;IACA;IACA;IACA;;;;EAIF,WACI;IAAQ,MAAM;IAAoB;IAAc,QAAQ;;IAEtD;;;;EAIN,mBACI;IACM;IAAe,OAAO;;IACtB;IAAuB,QAAQ;;IAEnC;;;;KAII;UAQK;EACf;;EAEA;;EAEA;;;;KAKU;EAEN;EACA;;EAEA;;EAGA;;;;EAIA;EACA;EACA;;EAEA;EAAmB;;KAEb;EAEN;EACA;;EAEA;;EAEA,UAAU;;EAEV;;EAEA;EACA,OAAO;EACP,YAAY;;EAGZ;EACA,UAAU;EACV,QAAQ;EACR;EACA,OAAO;;;KAID;EAEN;EACA;;EAEA;;EAEA;;EAGA;EACA;EACA;EACA;;;EAGA;EACA;;UAGW;EACf;EACA;EACA;IAAa;IAAgB;;;EAE7B;EACA;EACA,aAAa;EACb,QAAQ;EACR,cAAc;EACd;EACA;EACA;EACA;IAAS;IAAiB;;EAC1B;IACE,UAAU;IACV;IACA;IACA;;EAEF,OAAO;EACP,QAAQ;EACR;;EAEA;;;EAGA;IAAS,OAAO;IAA+C;;;;UAIhD;;;EAGf;EACA;EACA;EACA,QAAQ;EACR;;EAEA;;UAGe;EACf,cAAc;EACd;EACA,QAAQ;EACR;EACA;IAAa;IAAc;IAAc;IAA+B;;;EAExE;;EAEA;EACA,UAAU,QAAQ,OAAO;EACzB,UAAU;;;;EAIV,YAAY;IACV;IACA;IACA;IACA,oBAAoB;IACpB,eAAe;;;UAIF;EACf,UAAU;EACV;;EAEA,gBAAgB;EAChB;;EAEA;;;;UAKe;;EAEf;EACA;EACA,cAAc;EACd,QAAQ;EACR;EACA,QAAQ;EACR,gBAAgB;EAChB,OAAO;EACP;;;EAGA;EACA;;;;EAIA;IAAa;IAAkB;;;EAE/B;;UAGe;;EAEf;;;;EAIA;;;;qBA+BW,0BAA0B;;;;;;wBAUvB,mBAAmB,OAAO,kBAAkB;;;;;;wBAkD5C,gBAAgB,OAAO,kBAAkB;;wBAgFzC,WACd,mBAAmB,mBACnB,UAAS,oBACR,WAAW,gBAAgB;;wBA8Fd,aAAa,QAAQ,YAAY;UA+QhC;;;EAGf;;;;;;;;wBASc,qBACd,QAAQ,WAAW,gBAAgB,0BACnC,UAAS;;;UC9wBM;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA;;;;;;EAMA;;;;EAIA;;;EAGA,cAAc;;;;EAId;;;;EAIA;IAAU;IAAe;;;;UAIV;;EAEf;;;EAGA,MAAM;;;EAGN;EACA;EACA;EACA;;;EAGA;;EAEA,OAAO;IAAQ;IAAgB;IAAc;;;KAGnC;UAEK;EACf;EACA;EACA,QAAQ;;EAER;;;EAGA;;EAEA;EACA;;UAGe;EACf;;EAEA;;;EAGA;;;EAGA;;;EAGA;;EAEA;;UAGe;EACf;;;EAGA;;EAEA;;UAGe;EACf;EACA;;EAEA;EACA;EACA;IACE;IACA;;IAEA;IACA;;IAEA;;IAEA;IACA;;EAEF;IACE;IACA,MAAM;IACN;IACA;IACA;IACA;IACA;IACA;;;EAGF;IACE,QAAQ;MAAQ;MAAc;;;IAE9B;IACA;;EAEF;IACE;IACA;;IAEA;IACA;IACA,OAAO;MAAQ;MAAc;MAAe;;IAC5C;;;EAGF;EACA;EACA,OAAO;EACP,OAAO;EACP,SAAS;IAAQ;IAAgB;IAAuB;;EACxD,WAAW;;;EAGX;EACA;;;;;;;wBA8Cc,cACd,OAAO,iBACP,qBACA,UAAS,uBACR,mBAAmB;;;;;;wBA0SN,iBACd,MAAM,mBAAmB,qBACxB;;;;;;;;;wBAg3Ba,kBACd,aAAa,kBACb;EAAW;IACV;;wBAsEa,oBAAoB,MAAM,mBAAmB;;;UCn8C5C;EACf;EACA;;;EAGA;;UAGe;;EAEf;EACA;;UAGe;EACf;EACA;;;;;EAKA;;EAEA;;EAEA;;EAEA;;UAGe;EACf;EACA;;EAEA;;EAEA;IACE;IACA;IACA;;EAEF;EACA;EACA,cAAc;EACd,cAAc;EACd,OAAO;EACP,gBAAgB;;UAGD;EACf;;;;KAKU,mBAAmB;wBAEf,WACd,OAAO,iBACP,UAAS;EACN,MAAM;EAAgB;IAAU;IAAc,OAAO;;;;;UClFzC;EACf;EACA,QAAQ;EACR;;;EAGA;EACA;;;EAGA;EACA;EACA;;;EAGA,UAAU;EACV,UAAU;;UAGK;;;;;EAKf,OAAO;EACP;EACA;;wBAGc,KAAK,OAAO;EAC1B,MAAM;EACN;IAAU;IAAc;;;;;;KCId,0BAA0B,YAAY;;UAGjC,qBAAqB;EACpC,QAAQ;;EAER,aAAa;;;;EAIb;EACA;IAAQ;IAAgB;IAAgB,OAAO;IAAa;;EAC5D,QAAQ;;UAGO,uBAAuB,iBAC9B,KAAK,iBAAiB;;;EAG9B,QAAQ,oBAAoB;;;EAG5B,MAAM;;;EAGN,SAAS,OAAO,qBAAqB,WAAW,QAAQ;;;EAGxD,YAAY,OAAO,qBAAqB,WAAW,QAAQ;;;wBAI7C,eAAe,aAAa;;;;;wBAS5B,kBAAkB,YAAY,oBAAoB;;;;;;;wBA6B5C,gBAAgB,SACpC,SAAS,uBAAuB,WAC/B,QAAQ"}
|
package/dist/search/index.js
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
import { a as hashCanonical } from "../canonical-BsZVgBhW.js";
|
|
2
|
-
import { g as operatorYield, h as MIN_OUTCOMES_FOR_WEIGHT, m as EXPANSION_OPERATORS, n as runSearch, v as rankingSplit, y as summarizeSamples } from "../search-kernel-
|
|
3
|
-
import {
|
|
2
|
+
import { g as operatorYield, h as MIN_OUTCOMES_FOR_WEIGHT, m as EXPANSION_OPERATORS, n as runSearch, v as rankingSplit, y as summarizeSamples } from "../search-kernel-BIXWM7YV.js";
|
|
3
|
+
import { C as editCredit, S as front, _ as surfaceTextEdits, a as skillCalibration, b as searchPlateau, c as metaSearch, d as renderMetaSearchText, f as searchPolicyGenome, g as lineageEdits, h as lineEditDistance, i as nextUnitExtension, l as metaSearchScore, m as landscape, n as taskMatrix, o as skillManifold, p as formatLandscape, r as formatSkillManifold, s as META_SEARCH_SCORE_SOURCE, t as tree, u as objectiveKey, v as vectorEmbedding, w as editCreditText, x as screenedNodes, y as draftOnPlateau } from "../tree-kc1d80yF.js";
|
|
4
4
|
//#region src/search/nested-search.ts
|
|
5
5
|
/** The deterministic id of the inner search one outer cell attempt runs. */
|
|
6
6
|
function nestedSearchId(containment) {
|
|
@@ -182,6 +182,6 @@ function asRecord(value) {
|
|
|
182
182
|
return value !== null && typeof value === "object" && !Array.isArray(value) ? value : { value };
|
|
183
183
|
}
|
|
184
184
|
//#endregion
|
|
185
|
-
export {
|
|
185
|
+
export { EXPANSION_OPERATORS, META_SEARCH_SCORE_SOURCE, MIN_OUTCOMES_FOR_WEIGHT, draftOnPlateau, editCredit, editCreditText, formatLandscape, formatSkillManifold, front, landscape, lineEditDistance, lineageEdits, metaSearch, metaSearchScore, nestedSearchId, nextUnitExtension, objectiveKey, operatorYield, rankingSplit, renderMetaSearchText, runNestedSearch, screenedNodes, searchConfigCodec, searchPlateau, searchPolicyGenome, skillCalibration, skillManifold, summarizeSamples, surfaceTextEdits, taskMatrix, tree, vectorEmbedding };
|
|
186
186
|
|
|
187
187
|
//# sourceMappingURL=index.js.map
|
|
@@ -1,12 +1,12 @@
|
|
|
1
1
|
import { a as hashCanonical, i as compareCodeUnits, r as canonicalString } from "./canonical-BsZVgBhW.js";
|
|
2
2
|
import { t as mulberry32 } from "./random-Dn5fPWkt.js";
|
|
3
3
|
import { t as confidenceInterval } from "./descriptive-CnU8Yq29.js";
|
|
4
|
-
import { c as wilson } from "./paired-binary-
|
|
4
|
+
import { c as wilson } from "./paired-binary-Q36bhn2R.js";
|
|
5
5
|
import { n as DECISION_PAIRED_DELTA_STATISTIC } from "./paired-tests-jbwPdSz5.js";
|
|
6
|
-
import { i as pairedDeltaTest, r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-
|
|
6
|
+
import { i as pairedDeltaTest, r as minimumPairsForPairedDeltaTest, t as decidePairedPromotion } from "./paired-promotion-decision-BN6IQTiu.js";
|
|
7
7
|
import { a as SearchLedgerError } from "./search-ledger-file-DY0AXX3W.js";
|
|
8
8
|
import { f as redactText } from "./redact-Dk1ZdJEm.js";
|
|
9
|
-
import { f as searchUnitScores, l as searchCellId, t as FileSearchLedger } from "./search-ledger-
|
|
9
|
+
import { f as searchUnitScores, l as searchCellId, t as FileSearchLedger } from "./search-ledger-aMTNQn_m.js";
|
|
10
10
|
import { t as acquireSingleRunLock } from "./single-run-lock-fVu3MZWH.js";
|
|
11
11
|
//#region src/pareto.ts
|
|
12
12
|
/** Does candidate A weakly dominate B — strictly better on at least one objective and no worse on any? */
|
|
@@ -2514,4 +2514,4 @@ function errorMessage(error) {
|
|
|
2514
2514
|
//#endregion
|
|
2515
2515
|
export { searchCellSetDigest as C, dominates as D, seedFromDigest as E, paretoFrontier as O, estimateNodeFromCells as S, searchPosterior as T, INSUFFICIENT_FROM as _, searchPolicyView as a, SEARCH_ESTIMATOR as b, planSearchClaim as c, verifySearchClaim as d, pairedPromotionPower as f, operatorYield as g, MIN_OUTCOMES_FOR_WEIGHT as h, searchExpansionIndex as i, paretoFrontierWithCrowding as k, searchClaimDecision as l, EXPANSION_OPERATORS as m, runSearch as n, SEARCH_CLAIM_RULE as o, requiredPairsForPairedPromotion as p, searchDivergence as r, decideSearchClaim as s, SEARCH_KERNEL_SOURCE as t, searchClaimReserveUsd as u, rankingSplit as v, searchEstimateMethod as w, estimateNode as x, summarizeSamples as y };
|
|
2516
2516
|
|
|
2517
|
-
//# sourceMappingURL=search-kernel-
|
|
2517
|
+
//# sourceMappingURL=search-kernel-BIXWM7YV.js.map
|