@tangle-network/agent-eval 0.187.2 → 0.189.0
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 +95 -0
- package/dist/adapters/http.d.ts +2 -2
- package/dist/{agent-profile-uAIQK3Nr.d.ts → agent-profile-DNV_U6QR.d.ts} +4 -4
- package/dist/{agent-profile-uAIQK3Nr.d.ts.map → agent-profile-DNV_U6QR.d.ts.map} +1 -1
- package/dist/analyst/index.d.ts +12 -12
- package/dist/analyst/index.d.ts.map +1 -1
- package/dist/analyst/index.js +7 -7
- package/dist/{analyze-runs-DapYZz02.d.ts → analyze-runs-BvK4-wSs.d.ts} +5 -5
- package/dist/{analyze-runs-DapYZz02.d.ts.map → analyze-runs-BvK4-wSs.d.ts.map} +1 -1
- package/dist/{analyze-runs-DCY7LMt4.js → analyze-runs-SCSWYpy0.js} +17 -9
- package/dist/analyze-runs-SCSWYpy0.js.map +1 -0
- package/dist/{benchmark-command-DGokT8zd.js → benchmark-command-BhhAYstU.js} +14 -14
- package/dist/benchmark-command-BhhAYstU.js.map +1 -0
- package/dist/benchmarks/index.d.ts +4 -4
- package/dist/benchmarks/index.js +3 -3
- package/dist/builder-eval/index.d.ts +3 -3
- package/dist/builder-eval/index.js +1 -1
- package/dist/campaign/index.d.ts +5 -5
- package/dist/campaign/index.js +7 -7
- package/dist/{campaign-DlxdTOvw.js → campaign-BWoSQPSU.js} +10 -10
- package/dist/{campaign-DlxdTOvw.js.map → campaign-BWoSQPSU.js.map} +1 -1
- package/dist/{campaign-evidence-IF1p0oDa.js → campaign-evidence-CBlGJ3kC.js} +5 -5
- package/dist/campaign-evidence-CBlGJ3kC.js.map +1 -0
- package/dist/{chat-client-DqPqWzVV.js → chat-client-C4dg7Ng_.js} +2 -2
- package/dist/{chat-client-DqPqWzVV.js.map → chat-client-C4dg7Ng_.js.map} +1 -1
- package/dist/{chat-json-call-C-DnDxEy.js → chat-json-call-JVGbfx95.js} +2 -2
- package/dist/{chat-json-call-C-DnDxEy.js.map → chat-json-call-JVGbfx95.js.map} +1 -1
- package/dist/cli.js +4 -4
- package/dist/{client-DCc7a8ZC.d.ts → client-BhEyAdfe.d.ts} +4 -4
- package/dist/{client-DCc7a8ZC.d.ts.map → client-BhEyAdfe.d.ts.map} +1 -1
- package/dist/{client-CXE-U1SA.js → client-QwJpsNYU.js} +5 -1
- package/dist/client-QwJpsNYU.js.map +1 -0
- package/dist/contract/index.d.ts +11 -11
- package/dist/contract/index.js +22 -13
- package/dist/contract/index.js.map +1 -1
- package/dist/{counterfactual-CLgrwhkY.d.ts → counterfactual-CPaWN99-.d.ts} +4 -4
- package/dist/{counterfactual-CLgrwhkY.d.ts.map → counterfactual-CPaWN99-.d.ts.map} +1 -1
- package/dist/{counterfactual-Bjq1mlUu.js → counterfactual-DtMWvn7D.js} +2 -2
- package/dist/{counterfactual-Bjq1mlUu.js.map → counterfactual-DtMWvn7D.js.map} +1 -1
- package/dist/deep-freeze-CurcOQg3.js +16 -0
- package/dist/deep-freeze-CurcOQg3.js.map +1 -0
- package/dist/{default-registry-DMxeb4GD.d.ts → default-registry-B5HrwrPM.d.ts} +6 -6
- package/dist/{default-registry-DMxeb4GD.d.ts.map → default-registry-B5HrwrPM.d.ts.map} +1 -1
- package/dist/{default-registry-Ca2IWH1X.js → default-registry-m2jhbG6Q.js} +4 -3
- package/dist/default-registry-m2jhbG6Q.js.map +1 -0
- package/dist/{define-agent-eval-C0EqJ9Gz.js → define-agent-eval-98sgUaC1.js} +8 -8
- package/dist/{define-agent-eval-C0EqJ9Gz.js.map → define-agent-eval-98sgUaC1.js.map} +1 -1
- package/dist/{define-agent-eval-CGBwPPaq.d.ts → define-agent-eval-q6890HU4.d.ts} +7 -7
- package/dist/{define-agent-eval-CGBwPPaq.d.ts.map → define-agent-eval-q6890HU4.d.ts.map} +1 -1
- package/dist/diagnosis/index.d.ts +38 -96
- package/dist/diagnosis/index.d.ts.map +1 -1
- package/dist/diagnosis/index.js +84 -135
- package/dist/diagnosis/index.js.map +1 -1
- package/dist/{dspy-rlm-engine-BE9lCzzL.js → dspy-rlm-engine-DLYttKBm.js} +3 -3
- package/dist/{dspy-rlm-engine-BE9lCzzL.js.map → dspy-rlm-engine-DLYttKBm.js.map} +1 -1
- package/dist/emitter-DbMXUIVP.js +334 -0
- package/dist/emitter-DbMXUIVP.js.map +1 -0
- package/dist/{emitter-Cs0egaFd.d.ts → emitter-HcCmFQl_.d.ts} +37 -6
- package/dist/emitter-HcCmFQl_.d.ts.map +1 -0
- package/dist/{engine-DNlKhIm0.d.ts → engine-BkLCnZQX.d.ts} +15 -5
- package/dist/engine-BkLCnZQX.d.ts.map +1 -0
- package/dist/{eval-campaign-XFHEpN86.js → eval-campaign-C0Q1Eeo_.js} +8 -5
- package/dist/{eval-campaign-XFHEpN86.js.map → eval-campaign-C0Q1Eeo_.js.map} +1 -1
- package/dist/evaluation.d.ts +2 -2
- package/dist/{exact-types-BEHN7tnq.d.ts → exact-types-CfpP2A4Z.d.ts} +2 -2
- package/dist/{exact-types-BEHN7tnq.d.ts.map → exact-types-CfpP2A4Z.d.ts.map} +1 -1
- package/dist/experiment/index.d.ts +3 -3
- package/dist/experiment/index.js +2 -3
- package/dist/experiment/index.js.map +1 -1
- package/dist/{external-optimizer-process-BbhvXTFx.js → external-optimizer-process-C3quNeKV.js} +3 -3
- package/dist/{external-optimizer-process-BbhvXTFx.js.map → external-optimizer-process-C3quNeKV.js.map} +1 -1
- package/dist/{external-optimizer-subprocess-DMgmqTB6.js → external-optimizer-subprocess-BDwgKLO1.js} +3 -3
- package/dist/{external-optimizer-subprocess-DMgmqTB6.js.map → external-optimizer-subprocess-BDwgKLO1.js.map} +1 -1
- package/dist/{failure-cluster-OldNRoAt.d.ts → failure-cluster-CbgEDcFJ.d.ts} +3 -3
- package/dist/{failure-cluster-OldNRoAt.d.ts.map → failure-cluster-CbgEDcFJ.d.ts.map} +1 -1
- package/dist/failure-taxonomy-CjabTATj.js +538 -0
- package/dist/failure-taxonomy-CjabTATj.js.map +1 -0
- package/dist/{feedback-trajectory-CytTR7u4.d.ts → feedback-trajectory-B3ZSkfAq.d.ts} +6 -6
- package/dist/{feedback-trajectory-CytTR7u4.d.ts.map → feedback-trajectory-B3ZSkfAq.d.ts.map} +1 -1
- package/dist/hosted/index.d.ts +2 -2
- package/dist/hosted/index.d.ts.map +1 -1
- package/dist/hosted/index.js +1 -1
- package/dist/{index-vYNHCV2G.d.ts → index-2e_MTznj.d.ts} +8 -8
- package/dist/{index-vYNHCV2G.d.ts.map → index-2e_MTznj.d.ts.map} +1 -1
- package/dist/{index-CKZa_rnt.d.ts → index-Bcc9PkOz.d.ts} +3 -3
- package/dist/{index-CKZa_rnt.d.ts.map → index-Bcc9PkOz.d.ts.map} +1 -1
- package/dist/{index-u0d1Jp4F.d.ts → index-BqDJ6fUK.d.ts} +45 -24
- package/dist/index-BqDJ6fUK.d.ts.map +1 -0
- package/dist/{index-BRsrfSHT.d.ts → index-CnJOPR_E.d.ts} +10 -10
- package/dist/{index-BRsrfSHT.d.ts.map → index-CnJOPR_E.d.ts.map} +1 -1
- package/dist/{index-YZrbnBSV.d.ts → index-Dk1dOUfm.d.ts} +11 -11
- package/dist/{index-YZrbnBSV.d.ts.map → index-Dk1dOUfm.d.ts.map} +1 -1
- package/dist/index.d.ts +384 -69
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +1492 -217
- package/dist/index.js.map +1 -1
- package/dist/{insight-report-DpOo-3iL.d.ts → insight-report-CvPkEIcC.d.ts} +14 -7
- package/dist/insight-report-CvPkEIcC.d.ts.map +1 -0
- package/dist/{integrity-Cy9WHAtb.js → integrity-BFP4kjGj.js} +6 -1
- package/dist/integrity-BFP4kjGj.js.map +1 -0
- package/dist/{integrity-rGOfSUle.d.ts → integrity-CaeNeusS.d.ts} +4 -4
- package/dist/integrity-CaeNeusS.d.ts.map +1 -0
- package/dist/jev.d.ts +2 -2
- package/dist/jev.js +2 -1
- package/dist/jev.js.map +1 -1
- package/dist/{journal-BuRLm6ho.js → journal-CDP0Vhx_.js} +262 -115
- package/dist/journal-CDP0Vhx_.js.map +1 -0
- package/dist/{kind-factory-V79Hw0AU.js → kind-factory-BGwN0nE6.js} +17 -5
- package/dist/{kind-factory-V79Hw0AU.js.map → kind-factory-BGwN0nE6.js.map} +1 -1
- package/dist/ledger-core/index.d.ts +2 -2
- package/dist/ledger-core/index.js +1 -1
- package/dist/{llm-client-DDhZ29Vm.js → llm-client-DF9bPNYA.js} +2 -2
- package/dist/{llm-client-DDhZ29Vm.js.map → llm-client-DF9bPNYA.js.map} +1 -1
- package/dist/{llm-judge-BVso5x-u.js → llm-judge-M_l_WXw2.js} +26 -24
- package/dist/llm-judge-M_l_WXw2.js.map +1 -0
- package/dist/{matrix-CPKsRY8Y.d.ts → matrix-CNypzP-v.d.ts} +2 -2
- package/dist/{matrix-CPKsRY8Y.d.ts.map → matrix-CNypzP-v.d.ts.map} +1 -1
- package/dist/meta-eval/index.d.ts +4 -4
- package/dist/{mint-BI14xehJ.d.ts → mint-DsS38v5c.d.ts} +25 -5
- package/dist/{mint-BI14xehJ.d.ts.map → mint-DsS38v5c.d.ts.map} +1 -1
- package/dist/{mint-B4zgSAJW.js → mint-iCD-W43s.js} +30 -12
- package/dist/mint-iCD-W43s.js.map +1 -0
- 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/pipelines/index.d.ts +67 -12
- package/dist/pipelines/index.d.ts.map +1 -1
- package/dist/pipelines/index.js +178 -41
- package/dist/pipelines/index.js.map +1 -1
- package/dist/{prime-protocol-CDBcL9NM.d.ts → prime-protocol-Fwvctmhm.d.ts} +2 -2
- package/dist/{prime-protocol-CDBcL9NM.d.ts.map → prime-protocol-Fwvctmhm.d.ts.map} +1 -1
- package/dist/{produced-state-Cn9ECr4v.js → produced-state-BOY98IJO.js} +4 -4
- package/dist/{produced-state-Cn9ECr4v.js.map → produced-state-BOY98IJO.js.map} +1 -1
- package/dist/{query-D6W6MaGx.d.ts → query-DBUQhehH.d.ts} +3 -3
- package/dist/{query-D6W6MaGx.d.ts.map → query-DBUQhehH.d.ts.map} +1 -1
- package/dist/random-DGEyr3zG.d.ts +83 -0
- package/dist/random-DGEyr3zG.d.ts.map +1 -0
- package/dist/{raw-provider-sink-BQd7mzyT.js → raw-provider-sink-C96CcqeG.js} +19 -56
- package/dist/raw-provider-sink-C96CcqeG.js.map +1 -0
- package/dist/redact-DPoH1-OW.d.ts +158 -0
- package/dist/redact-DPoH1-OW.d.ts.map +1 -0
- package/dist/redact-DxwlUAvh.js +824 -0
- package/dist/redact-DxwlUAvh.js.map +1 -0
- package/dist/{registry-D5Ly3GRV.d.ts → registry-mAQVtuLx.d.ts} +4 -4
- package/dist/{registry-D5Ly3GRV.d.ts.map → registry-mAQVtuLx.d.ts.map} +1 -1
- package/dist/{release-confidence-CSp3MsU4.d.ts → release-confidence-CkoXFx9I.d.ts} +4 -4
- package/dist/{release-confidence-CSp3MsU4.d.ts.map → release-confidence-CkoXFx9I.d.ts.map} +1 -1
- package/dist/{release-confidence-RanRUXRW.js → release-confidence-XQm_xjhR.js} +3 -3
- package/dist/release-confidence-XQm_xjhR.js.map +1 -0
- package/dist/{report-command-CSTjzFGY.js → report-command-DKJ_yAIj.js} +21 -12
- package/dist/{report-command-CSTjzFGY.js.map → report-command-DKJ_yAIj.js.map} +1 -1
- package/dist/reporting.d.ts +3 -3
- package/dist/reporting.js +1 -1
- package/dist/{researcher-D7E92STT.d.ts → researcher-DojEMb7E.d.ts} +7 -7
- package/dist/{researcher-D7E92STT.d.ts.map → researcher-DojEMb7E.d.ts.map} +1 -1
- package/dist/{reward-hacking-BIQ2BWLG.d.ts → reward-hacking-Cyk-QHk9.d.ts} +2 -2
- package/dist/{reward-hacking-BIQ2BWLG.d.ts.map → reward-hacking-Cyk-QHk9.d.ts.map} +1 -1
- package/dist/rl.d.ts +11 -10
- package/dist/rl.d.ts.map +1 -1
- package/dist/rl.js +7 -6
- package/dist/rl.js.map +1 -1
- package/dist/rollout/index.d.ts +4 -4
- package/dist/rollout/index.js +2 -2
- package/dist/{rollout-C0Ifk3vK.js → rollout-B-XBReTC.js} +2 -2
- package/dist/{rollout-C0Ifk3vK.js.map → rollout-B-XBReTC.js.map} +1 -1
- package/dist/{rubric-predictive-validity-CTU8y8Cv.d.ts → rubric-predictive-validity-DIUwJJ9Y.d.ts} +2 -2
- package/dist/{rubric-predictive-validity-CTU8y8Cv.d.ts.map → rubric-predictive-validity-DIUwJJ9Y.d.ts.map} +1 -1
- package/dist/{run-record-tg_5H2IK.js → run-record-BRlM99rX.js} +41 -3
- package/dist/run-record-BRlM99rX.js.map +1 -0
- package/dist/{run-record-D9wS35fK.d.ts → run-record-DF-fuu31.d.ts} +59 -7
- package/dist/run-record-DF-fuu31.d.ts.map +1 -0
- package/dist/{run-record-CNNzWGpB.js → run-record-DnMFyg_g.js} +26 -13
- package/dist/run-record-DnMFyg_g.js.map +1 -0
- package/dist/{schema-CR5cpjQ3.d.ts → schema-2g7trUwH.d.ts} +17 -2
- package/dist/schema-2g7trUwH.d.ts.map +1 -0
- package/dist/{schema-BzWDXhOR.d.ts → schema-B5LtYvn9.d.ts} +7 -5
- package/dist/{schema-BzWDXhOR.d.ts.map → schema-B5LtYvn9.d.ts.map} +1 -1
- package/dist/schema-C1aaAxTf.js.map +1 -1
- package/dist/schema-CSf6qWgZ.js.map +1 -1
- package/dist/{semantic-concept-judge-Cnt0TWp0.js → semantic-concept-judge-Cm2K2wAo.js} +6 -6
- package/dist/{semantic-concept-judge-Cnt0TWp0.js.map → semantic-concept-judge-Cm2K2wAo.js.map} +1 -1
- package/dist/{series-convergence-BnMs_uAr.d.ts → series-convergence-DY7ha0kx.d.ts} +2 -2
- package/dist/{series-convergence-BnMs_uAr.d.ts.map → series-convergence-DY7ha0kx.d.ts.map} +1 -1
- package/dist/{server-DmpcFiVK.js → server-C8XvqCg8.js} +2 -2
- package/dist/{server-DmpcFiVK.js.map → server-C8XvqCg8.js.map} +1 -1
- package/dist/{skillopt-optimization-method-Cc9SVLqi.js → skillopt-optimization-method-BVSB676o.js} +7 -7
- package/dist/{skillopt-optimization-method-Cc9SVLqi.js.map → skillopt-optimization-method-BVSB676o.js.map} +1 -1
- package/dist/spans-W0o8pbEg.d.ts +60 -0
- package/dist/spans-W0o8pbEg.d.ts.map +1 -0
- package/dist/{statistical-heldout-DTmhTW8P.d.ts → statistical-heldout-plAHG82H.d.ts} +3 -3
- package/dist/{statistical-heldout-DTmhTW8P.d.ts.map → statistical-heldout-plAHG82H.d.ts.map} +1 -1
- package/dist/{store-BErPvYBr.d.ts → store-C4avFjIi.d.ts} +2 -2
- package/dist/{store-BErPvYBr.d.ts.map → store-C4avFjIi.d.ts.map} +1 -1
- package/dist/{store-otlp-2gYCy-v0.js → store-otlp-D0AV4cvB.js} +2 -2
- package/dist/{store-otlp-2gYCy-v0.js.map → store-otlp-D0AV4cvB.js.map} +1 -1
- package/dist/{store-tool-spans-B3bX9yqn.d.ts → store-tool-spans-8PAyhRqb.d.ts} +9 -54
- package/dist/store-tool-spans-8PAyhRqb.d.ts.map +1 -0
- package/dist/{store-tool-spans-B1e6Auc2.js → store-tool-spans-B_0mXvYl.js} +4 -4
- package/dist/{store-tool-spans-B1e6Auc2.js.map → store-tool-spans-B_0mXvYl.js.map} +1 -1
- package/dist/storyboard/index.d.ts +1 -1
- package/dist/{summary-report-Iy6kXo8x.d.ts → summary-report-DvvesgR_.d.ts} +6 -5
- package/dist/summary-report-DvvesgR_.d.ts.map +1 -0
- package/dist/supervisor-run/index.d.ts +2 -2
- package/dist/supervisor-run/index.d.ts.map +1 -1
- package/dist/supervisor-run/index.js +1 -1
- package/dist/{tool-groups-veQYPiTy.d.ts → tool-groups-W7MJobX_.d.ts} +3 -3
- package/dist/tool-groups-W7MJobX_.d.ts.map +1 -0
- package/dist/{tool-waste-nbHPAxei.js → tool-waste-CJDfxH45.js} +3 -537
- package/dist/tool-waste-CJDfxH45.js.map +1 -0
- package/dist/{tool-waste-D23I0zWm.d.ts → tool-waste-CP5DO383.d.ts} +4 -4
- package/dist/{tool-waste-D23I0zWm.d.ts.map → tool-waste-CP5DO383.d.ts.map} +1 -1
- package/dist/trace-repair/index.d.ts +2 -2
- package/dist/traces.d.ts +12 -11
- package/dist/traces.d.ts.map +1 -1
- package/dist/traces.js +9 -9
- package/dist/{trajectory-D7qrNvaN.d.ts → trajectory-Y1fmdlw7.d.ts} +3 -3
- package/dist/{trajectory-D7qrNvaN.d.ts.map → trajectory-Y1fmdlw7.d.ts.map} +1 -1
- package/dist/trajectory-replay/index.d.ts +3 -3
- package/dist/trajectory-replay/index.js +2 -2
- package/dist/{types-BjB2wZM9.d.ts → types-Bay1dh2o.d.ts} +6 -6
- package/dist/{types-BjB2wZM9.d.ts.map → types-Bay1dh2o.d.ts.map} +1 -1
- package/dist/{types-BTRcIqFr.d.ts → types-BvmWNjjs.d.ts} +4 -4
- package/dist/{types-BTRcIqFr.d.ts.map → types-BvmWNjjs.d.ts.map} +1 -1
- package/dist/{types-CBbLtr2J.d.ts → types-C0bWoSKC.d.ts} +10 -11
- package/dist/types-C0bWoSKC.d.ts.map +1 -0
- package/dist/{types-C8VBbiw6.js → types-CW5b90JH.js} +2 -15
- package/dist/types-CW5b90JH.js.map +1 -0
- package/dist/{types--18oHwVF.d.ts → types-Cvqy7asY.d.ts} +3 -3
- package/dist/{types--18oHwVF.d.ts.map → types-Cvqy7asY.d.ts.map} +1 -1
- package/dist/{types-Tp5WhYL9.d.ts → types-DRNlLbBR.d.ts} +2 -2
- package/dist/{types-Tp5WhYL9.d.ts.map → types-DRNlLbBR.d.ts.map} +1 -1
- package/dist/wire/index.d.ts +3 -3
- package/dist/wire/index.js +1 -1
- package/docs/code-agent-intake.md +1 -1
- package/docs/diagnosis.md +2 -3
- package/docs/eval-surface-map.md +1 -1
- package/docs/experiment.md +0 -1
- package/docs/insight-report.md +3 -2
- package/docs/public-api.md +2 -2
- package/docs/redaction.md +59 -0
- package/docs/rollout.md +5 -5
- package/docs/search-history-receipts.md +1 -0
- package/docs/trace-contracts.md +102 -0
- package/docs/verdicts.md +1 -1
- package/package.json +3 -2
- package/dist/analyze-runs-DCY7LMt4.js.map +0 -1
- package/dist/benchmark-command-DGokT8zd.js.map +0 -1
- package/dist/campaign-evidence-IF1p0oDa.js.map +0 -1
- package/dist/client-CXE-U1SA.js.map +0 -1
- package/dist/default-registry-Ca2IWH1X.js.map +0 -1
- package/dist/emitter-Cs0egaFd.d.ts.map +0 -1
- package/dist/emitter-DeQHiDMm.js +0 -257
- package/dist/emitter-DeQHiDMm.js.map +0 -1
- package/dist/engine-DNlKhIm0.d.ts.map +0 -1
- package/dist/experiment-tracker-BKEumQug.js +0 -280
- package/dist/experiment-tracker-BKEumQug.js.map +0 -1
- package/dist/experiment-tracker-URRGP9ek.d.ts +0 -237
- package/dist/experiment-tracker-URRGP9ek.d.ts.map +0 -1
- package/dist/index-u0d1Jp4F.d.ts.map +0 -1
- package/dist/insight-report-DpOo-3iL.d.ts.map +0 -1
- package/dist/integrity-Cy9WHAtb.js.map +0 -1
- package/dist/integrity-rGOfSUle.d.ts.map +0 -1
- package/dist/journal-BuRLm6ho.js.map +0 -1
- package/dist/llm-judge-BVso5x-u.js.map +0 -1
- package/dist/mint-B4zgSAJW.js.map +0 -1
- package/dist/raw-provider-sink-BQd7mzyT.js.map +0 -1
- package/dist/redact-7Aq1ukl-.js +0 -107
- package/dist/redact-7Aq1ukl-.js.map +0 -1
- package/dist/release-confidence-RanRUXRW.js.map +0 -1
- package/dist/run-record-CNNzWGpB.js.map +0 -1
- package/dist/run-record-D9wS35fK.d.ts.map +0 -1
- package/dist/run-record-tg_5H2IK.js.map +0 -1
- package/dist/schema-CR5cpjQ3.d.ts.map +0 -1
- package/dist/store-tool-spans-B3bX9yqn.d.ts.map +0 -1
- package/dist/summary-report-Iy6kXo8x.d.ts.map +0 -1
- package/dist/tool-groups-veQYPiTy.d.ts.map +0 -1
- package/dist/tool-waste-nbHPAxei.js.map +0 -1
- package/dist/types-C8VBbiw6.js.map +0 -1
- package/dist/types-CBbLtr2J.d.ts.map +0 -1
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"tool-waste-
|
|
1
|
+
{"version":3,"file":"tool-waste-CP5DO383.d.ts","names":[],"sources":["../src/baseline.ts","../src/pipelines/budget-breach.ts","../src/pipelines/judge-agreement.ts","../src/tool-use-metrics.ts","../src/pipelines/tool-waste.ts"],"mappings":";;;;UA2BiB;EACf;EACA;EACA;EACA;;;EAGA;EACA;EACA;EACA;EACA;;EAEA;EACA;;UAGe;EACf,SAAS;;EAET;;EAEA;;UAGe;;EAEf;;EAEA;;EAEA;;;iBAiHc,IAAI;;;UCjKH;EACf;EACA;EACA;EACA,iBAAiB;EACjB;EACA;EACA;EACA;;UAGe;EACf,UAAU;EACV,aAAa;EACb,YAAY;EACZ,WAAW;EACX;EACA;;iBAGoB,iBACpB,OAAO,YACP;EAAW;EAAqB;IAC/B,QAAQ;;;UCnBM;EACf;EACA;EACA;;EAEA;EACA;EACA;;UAGe;EACf,OAAO;EACP;EACA;;iBAGoB,mBAAmB,OAAO,aAAa,QAAQ;;;UCjBpD;EACf;EACA;;EAEA;EACA,QAAQ,eAAe;EACvB;;EAEA;;EAEA;;EAEA;;UAGe;EACf;EACA;EACA;EACA;EACA;;UAGe;;EAEf,kBAAkB;;iBAGE,sBACpB,OAAO,YACP,eACA,UAAS,iBACR,QAAQ;;;UC5BM;EACf;EACA;EACA;EACA;;UAGe;EACf,OAAO;EACP;;UAGe;EACf;EACA,eAAe,MAAM,UAAU;IAAS,KAAK,QAAQ,kBAAkB;;;iBAGnD,cACpB,OAAO,YACP,UAAS,mBACR,QAAQ"}
|
|
@@ -1,8 +1,8 @@
|
|
|
1
1
|
import { r as CaptureIntegrityError } from "../errors-DEE6u6ot.js";
|
|
2
2
|
import { b as CustomTokenPricing, c as CostLedgerHandle, p as CostProvenance } from "../cost-ledger-DbQdN3nO.js";
|
|
3
|
-
import {
|
|
3
|
+
import { d as RunTokenUsage } from "../run-record-DF-fuu31.js";
|
|
4
4
|
import { t as DefaultVerdict } from "../verdict-E4eRNf7-.js";
|
|
5
|
-
import { a as TraceAnalystLimits, n as TraceAnalysisEngine } from "../engine-
|
|
5
|
+
import { a as TraceAnalystLimits, n as TraceAnalysisEngine } from "../engine-BkLCnZQX.js";
|
|
6
6
|
import { t as PrimeBridgeTransport } from "../prime-bridge-transport-6feEglLf.js";
|
|
7
7
|
import { a as RecordedTrajectoryStep, h as isRecordedTimeout } from "../steps-Bg5-KVx1.js";
|
|
8
8
|
//#region src/trace-repair/mini-swe-scaffold.d.ts
|
package/dist/traces.d.ts
CHANGED
|
@@ -1,15 +1,16 @@
|
|
|
1
1
|
import { c as ValidationError, o as LimitExceededError, r as CaptureIntegrityError, s as NotFoundError, t as AgentEvalError } from "./errors-DEE6u6ot.js";
|
|
2
|
-
import { C as
|
|
3
|
-
import { a as RunRecord,
|
|
4
|
-
import { A as DatasetOverview, B as TraceAnalystSpanKind, C as ReadSpanSourceInput, D as TraceAnalysisStore, E as TRACE_ANALYSIS_LIMITS, F as SpanMatchRecord, G as ViewTraceResult, H as TraceAnalystTraceSummary, I as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, L as TraceAnalystByteBudgets, M as QueryTracesPage, N as SearchSpanResult, O as TraceAnalysisStoreContext, P as SearchTraceResult, R as TraceAnalystFilters, S as BoundedTraceAnalysisStoreOptions, T as SpanSourceReader, U as ViewSpansResult, V as TraceAnalystSpanStatus, W as ViewTraceOversized, j as ErrorCluster, k as DEFAULT_TRACE_ANALYST_BUDGETS, w as ReadSpanSourceResult, z as TraceAnalystSpan } from "./types-
|
|
5
|
-
import {
|
|
6
|
-
import { $ as NoopRawProviderSink, Q as InMemoryRawProviderSinkOptions, X as FileSystemRawProviderSinkOptions, Y as FileSystemRawProviderSink, Z as InMemoryRawProviderSink, at as defaultProviderRedactor, et as ProviderRedactor, it as RawProviderSinkFilter, nt as RawProviderEvent, ot as providerFromBaseUrl, rt as RawProviderSink, tt as RawProviderDirection } from "./types-
|
|
7
|
-
import { a as RunFilter, i as InMemoryTraceStore, n as FileSystemTraceStore, o as SpanFilter, r as FileSystemTraceStoreOptions, s as TraceStore, t as EventFilter } from "./store-
|
|
8
|
-
import { _ as
|
|
9
|
-
import { a as
|
|
2
|
+
import { C as ToolSpan, D as isToolSpan, E as isLlmSpan, S as TRACE_SCHEMA_VERSION, T as isJudgeSpan, _ as SandboxSpan, a as FAILURE_CLASSES, b as SpanKind, c as JudgeSpan, d as RetrievalSpan, f as Run, g as RunStatus, h as RunOutcome, i as EventKind, l as LlmSpan, m as RunLayer, n as BudgetLedgerEntry, o as FailureClass, p as RunCaptureReport, r as BudgetSpec, s as GenericSpan, t as Artifact, u as Message, v as Span, w as TraceEvent, x as SpanStatus, y as SpanBase } from "./schema-2g7trUwH.js";
|
|
3
|
+
import { a as RunRecord, c as RunSplitTag, d as RunTokenUsage, u as RunTerminalOutcome } from "./run-record-DF-fuu31.js";
|
|
4
|
+
import { A as DatasetOverview, B as TraceAnalystSpanKind, C as ReadSpanSourceInput, D as TraceAnalysisStore, E as TRACE_ANALYSIS_LIMITS, F as SpanMatchRecord, G as ViewTraceResult, H as TraceAnalystTraceSummary, I as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, L as TraceAnalystByteBudgets, M as QueryTracesPage, N as SearchSpanResult, O as TraceAnalysisStoreContext, P as SearchTraceResult, R as TraceAnalystFilters, S as BoundedTraceAnalysisStoreOptions, T as SpanSourceReader, U as ViewSpansResult, V as TraceAnalystSpanStatus, W as ViewTraceOversized, j as ErrorCluster, k as DEFAULT_TRACE_ANALYST_BUDGETS, w as ReadSpanSourceResult, z as TraceAnalystSpan } from "./types-Bay1dh2o.js";
|
|
5
|
+
import { A as AnalyzeTracesInput, B as CaptureFetchContext, C as inferDomainKeywords, D as planTraceInsightQuestions, E as isTraceInsightEvidence, F as OTEL_AGENT_EVAL_SCOPE, G as SseUsageMode, H as captureFetchToRawSink, I as OtlpExport, J as extractUsageFromSse, K as extractUsage, L as OtlpResourceSpans, M as AnalyzeTracesResult, N as analyzeTraces, O as scoreTraceInsightReadiness, P as OtlpFlatLine, R as OtlpSpan, S as domainEvidencePattern, T as isTraceInsightAttribution, U as ExtractUsageFromSseOptions, V as CaptureFetchOptions, W as ExtractedUsage, Y as createBoundedTraceAnalysisStore, _ as TraceInsightTask, a as ToolSpansToTraceAnalysisStoreOptions, b as defaultTraceInsightPanel, c as TraceInsightContext, d as TraceInsightPanelRole, f as TraceInsightPromptInput, g as TraceInsightSuite, h as TraceInsightReadiness, i as OtlpFileTraceStoreOptions, j as AnalyzeTracesOptions, k as tokenizeDomainWords, l as TraceInsightEvidence, m as TraceInsightQuestion, n as toolSpansToTraceAnalysisStore, o as otlpTextToTraceAnalysisStore, p as TraceInsightQualityGate, q as extractUsageFromResponse, r as OtlpFileTraceStore, s as TraceInsightAttribution, t as ToolTraceMissingError, u as TraceInsightFinding, v as buildTraceInsightContext, w as isTraceInsightArtifact, x as describeTraceInsightScope, y as buildTraceInsightPrompt, z as exportRunAsOtlp } from "./store-tool-spans-8PAyhRqb.js";
|
|
6
|
+
import { $ as NoopRawProviderSink, Q as InMemoryRawProviderSinkOptions, X as FileSystemRawProviderSinkOptions, Y as FileSystemRawProviderSink, Z as InMemoryRawProviderSink, at as defaultProviderRedactor, et as ProviderRedactor, it as RawProviderSinkFilter, nt as RawProviderEvent, ot as providerFromBaseUrl, rt as RawProviderSink, tt as RawProviderDirection } from "./types-C0bWoSKC.js";
|
|
7
|
+
import { a as RunFilter, i as InMemoryTraceStore, n as FileSystemTraceStore, o as SpanFilter, r as FileSystemTraceStoreOptions, s as TraceStore, t as EventFilter } from "./store-C4avFjIi.js";
|
|
8
|
+
import { _ as buildTraceAnalysisToolDescriptors, g as UNTRUSTED_TRACE_TEXT, h as TraceAnalysisToolDescriptor, m as TRACE_ANALYST_TOOL_NAMESPACE, p as BuildTraceAnalysisToolsOptions, v as traceAnalystFunctionGroup } from "./engine-BkLCnZQX.js";
|
|
9
|
+
import { a as SpanHandle, i as RunCompleteHookContext, n as CaptureWrite, o as TraceEmitter, r as RunCompleteHook, s as TraceEmitterOptions, t as CaptureErrorHandler } from "./emitter-HcCmFQl_.js";
|
|
10
|
+
import { S as shareAllowed, _ as emptyRedactionReport, a as RedactionFinding, b as redactForShare, c as SafetyCategory, d as ShareSafetyVerdict, f as assessShareSafety, g as detectPersonalData, h as detectCredential, i as RedactOptions, l as ShareSafetyFinding, m as combineVerdicts, n as REDACTION_PROFILES, o as RedactionProfile, p as classifyKey, r as REDACTION_VERSION, s as RedactionReport, t as AssessOptions, u as ShareSafetyStatus, v as normalizeKey, x as redactText, y as redact } from "./redact-DPoH1-OW.js";
|
|
10
11
|
import { C as TOOL_LATENCY_MS, D as asNumber, E as applyLlmSpanOtlpAttributes, O as contextInputTokens, S as TOOL_ARGS_CAPTURED, T as TOOL_NAME_ATTR_KEYS, _ as LlmSpanOtlpInput, a as LLM_CACHE_WRITE_TOKEN_ATTR_KEYS, b as RUN_COST_ATTR_KEYS, c as LLM_COST_USD, d as LLM_MODEL_ATTR_KEYS, f as LLM_MODEL_NAME, g as LLM_REASONING_TOKEN_ATTR_KEYS, h as LLM_REASONING_TOKENS, i as LLM_CACHE_WRITE_TOKENS, k as firstNumberAttr, l as LLM_INPUT_TOKENS, m as LLM_OUTPUT_TOKEN_ATTR_KEYS, n as LLM_CACHED_TOKENS, o as LLM_CONTEXT_TOKENS, p as LLM_OUTPUT_TOKENS, r as LLM_CACHED_TOKEN_ATTR_KEYS, s as LLM_COST_ATTR_KEYS, t as INPUT_VALUE, u as LLM_INPUT_TOKEN_ATTR_KEYS, v as OPENINFERENCE_SPAN_KIND, w as TOOL_NAME, x as SPAN_KIND_ATTR_KEYS, y as OUTPUT_VALUE } from "./attribute-vocabulary-DLJ6303h.js";
|
|
11
|
-
import { a as RunIntegrityReport, i as RunIntegrityIssueCode, n as RunIntegrityExpectations, o as assertRunCaptured, r as RunIntegrityIssue, s as throwIfRunIncomplete, t as RunIntegrityError } from "./integrity-
|
|
12
|
-
import { a as groupBy, c as judgeSpans, d as runMetricExtractor, f as runsForScenario, i as argHash, l as llmSpans, n as RunMetric, o as hasCapturedToolArgs, p as toolSpans, r as aggregateLlm, s as isRunMetric, t as RUN_METRICS, u as runFailureClass } from "./query-
|
|
12
|
+
import { a as RunIntegrityReport, i as RunIntegrityIssueCode, n as RunIntegrityExpectations, o as assertRunCaptured, r as RunIntegrityIssue, s as throwIfRunIncomplete, t as RunIntegrityError } from "./integrity-CaeNeusS.js";
|
|
13
|
+
import { a as groupBy, c as judgeSpans, d as runMetricExtractor, f as runsForScenario, i as argHash, l as llmSpans, n as RunMetric, o as hasCapturedToolArgs, p as toolSpans, r as aggregateLlm, s as isRunMetric, t as RUN_METRICS, u as runFailureClass } from "./query-DBUQhehH.js";
|
|
13
14
|
//#region src/trace-analyst/errors.d.ts
|
|
14
15
|
/** Invalid trace-tool arguments, including malformed filters and regexes. */
|
|
15
16
|
declare class TraceAnalysisValidationError extends ValidationError {
|
|
@@ -395,5 +396,5 @@ declare function otlpRowsToTraceRunRecords(rows: Iterable<object>, opts: OtlpToR
|
|
|
395
396
|
declare const TRACE_ANALYST_ACTOR_DESCRIPTION = "Answer the question by inspecting the OTLP trace dataset with the available tools.\n\n1. Call getDatasetOverview first. Use its real trace ids and dataset size to plan the investigation.\n2. Narrow with queryTraces and countTraces before scanning large payloads.\n3. For a small trace, use viewTrace. For a large trace, use searchTrace and then viewSpans or searchSpan.\n4. Never invent a trace id, span id, tool result, error, frequency, or final outcome.\n5. When a search reports has_more, refine the query before drawing a conclusion.\n6. Use llm_query only over evidence already loaded. A recursive query cannot inspect traces itself.\n7. Cite exact evidence URIs returned by the tools. Include a short exact excerpt when it supports the claim.\n8. Return no finding when the available evidence cannot support one.\n\nThe prose answer must directly answer the question and state important uncertainty.\nThe findings array contains only actionable or decision-relevant claims supported by inspected evidence.";
|
|
396
397
|
declare const TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION = "trace-analyst-research-v1-2026-07-30";
|
|
397
398
|
//#endregion
|
|
398
|
-
export { type AnalyzeTracesInput, type AnalyzeTracesOptions, type AnalyzeTracesResult, Artifact, type BoundedTraceAnalysisStoreOptions, BudgetLedgerEntry, BudgetSpec, type BuildTraceAnalysisToolsOptions, CaptureFetchContext, CaptureFetchOptions,
|
|
399
|
+
export { type AnalyzeTracesInput, type AnalyzeTracesOptions, type AnalyzeTracesResult, Artifact, AssessOptions, type BoundedTraceAnalysisStoreOptions, BudgetLedgerEntry, BudgetSpec, type BuildTraceAnalysisToolsOptions, CaptureErrorHandler, CaptureFetchContext, CaptureFetchOptions, CaptureWrite, DEFAULT_TRACE_ANALYST_BUDGETS, type DatasetOverview, type ErrorCluster, EventFilter, EventKind, ExportableSpan, ExtractUsageFromSseOptions, ExtractedUsage, FAILURE_CLASSES, FailureClass, FileSystemRawProviderSink, FileSystemRawProviderSinkOptions, FileSystemTraceStore, FileSystemTraceStoreOptions, type FlattenOtlpOptions, GenericSpan, INPUT_VALUE, InMemoryRawProviderSink, InMemoryRawProviderSinkOptions, InMemoryTraceStore, JudgeSpan, LLM_CACHED_TOKENS, LLM_CACHED_TOKEN_ATTR_KEYS, LLM_CACHE_WRITE_TOKENS, LLM_CACHE_WRITE_TOKEN_ATTR_KEYS, LLM_CONTEXT_TOKENS, LLM_COST_ATTR_KEYS, LLM_COST_USD, LLM_INPUT_TOKENS, LLM_INPUT_TOKEN_ATTR_KEYS, LLM_MODEL_ATTR_KEYS, LLM_MODEL_NAME, LLM_OUTPUT_TOKENS, LLM_OUTPUT_TOKEN_ATTR_KEYS, LLM_REASONING_TOKENS, LLM_REASONING_TOKEN_ATTR_KEYS, LlmSpan, LlmSpanOtlpInput, Message, NoopRawProviderSink, OPENINFERENCE_SPAN_KIND, OTEL_AGENT_EVAL_SCOPE, OUTPUT_VALUE, OtelExportConfig, OtelExporter, OtlpExport, OtlpFileTraceStore, type OtlpFileTraceStoreOptions, type OtlpFlatLine, OtlpResourceSpans, OtlpSpan, OtlpSpanRole, OtlpSpanRoleInput, type OtlpToRunRecordsOptions, type OtlpTraceRunRecord, type ProjectedOtlpSpan, ProviderRedactor, type QueryTracesPage, REDACTION_PROFILES, REDACTION_VERSION, RUN_COST_ATTR_KEYS, RUN_METRICS, RawProviderDirection, RawProviderEvent, RawProviderSink, RawProviderSinkFilter, type ReadSpanSourceInput, type ReadSpanSourceResult, RedactOptions, RedactionFinding, RedactionProfile, RedactionReport, RetrievalSpan, Run, RunCaptureReport, RunCompleteHook, RunCompleteHookContext, RunFilter, RunIntegrityError, RunIntegrityExpectations, RunIntegrityIssue, RunIntegrityIssueCode, RunIntegrityReport, RunLayer, RunMetric, RunOutcome, RunStatus, SPAN_KIND_ATTR_KEYS, SafetyCategory, SandboxSpan, type SearchSpanResult, type SearchTraceResult, ShareSafetyFinding, ShareSafetyStatus, ShareSafetyVerdict, Span, SpanBase, SpanFilter, SpanHandle, SpanKind, type SpanMatchRecord, SpanNotFoundError, type SpanSourceReader, SpanStatus, SseUsageMode, TOOL_ARGS_CAPTURED, TOOL_LATENCY_MS, TOOL_NAME, TOOL_NAME_ATTR_KEYS, TRACE_ANALYSIS_LIMITS, TRACE_ANALYST_ACTOR_DESCRIPTION, TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, TRACE_ANALYST_TOOL_NAMESPACE, TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, TRACE_SCHEMA_VERSION, ToolSpan, ToolSpanOtlpInput, type ToolSpansToTraceAnalysisStoreOptions, ToolTraceMissingError, type TraceAggregate, TraceAnalysisLimitError, type TraceAnalysisStore, type TraceAnalysisStoreContext, TraceAnalysisStoreContractError, type TraceAnalysisToolDescriptor, TraceAnalysisValidationError, type TraceAnalystByteBudgets, type TraceAnalystFilters, type TraceAnalystHookOptions, type TraceAnalystSpan, type TraceAnalystSpanKind, type TraceAnalystSpanStatus, type TraceAnalystTraceSummary, TraceEmitter, TraceEmitterOptions, TraceEvent, TraceFileMalformedError, TraceFileMissingError, TraceFileTooLargeError, type TraceInsightAttribution, type TraceInsightContext, type TraceInsightEvidence, type TraceInsightFinding, type TraceInsightPanelRole, type TraceInsightPromptInput, type TraceInsightQualityGate, type TraceInsightQuestion, type TraceInsightReadiness, type TraceInsightSuite, type TraceInsightTask, TraceNotFoundError, TraceStore, TraceStoreSource, TraceStoreToOtlpOptions, TracesToOtlpResult, UNTRUSTED_TRACE_TEXT, type ViewSpansResult, type ViewTraceOversized, type ViewTraceResult, aggregateLlm, analyzeTraces, applyLlmSpanOtlpAttributes, applyToolSpanOtlpAttributes, argHash, asNumber, asString, assertRunCaptured, assessShareSafety, buildTraceAnalysisToolDescriptors, buildTraceInsightContext, buildTraceInsightPrompt, captureFetchToRawSink, classifyKey, classifyOtlpSpanRole, combineVerdicts, contextInputTokens, convertTraceStoresToOtlp, createBoundedTraceAnalysisStore, createOtelExporter, createOtelTracingStore, defaultProviderRedactor, defaultTraceInsightPanel, describeTraceInsightScope, detectCredential, detectPersonalData, domainEvidencePattern, emptyRedactionReport, exportRunAsOtlp, extractOtlpAttributes, extractUsage, extractUsageFromResponse, extractUsageFromSse, firstNumberAttr, firstStringAttr, flattenOtlpExportToNdjson, groupBy, hasCapturedToolArgs, inferDomainKeywords, isJudgeSpan, isLlmSpan, isOtlpModelCall, isRunMetric, isToolSpan, isTraceInsightArtifact, isTraceInsightAttribution, isTraceInsightEvidence, judgeSpans, llmSpans, normalizeKey, otlpRowsToRunRecords, otlpRowsToTraceRunRecords, otlpTextToTraceAnalysisStore, otlpToRunRecords, otlpToTraceRunRecords, planTraceInsightQuestions, projectOtlpFlatLine, providerFromBaseUrl, redact, redactForShare, redactText, runFailureClass, runMetricExtractor, runsForScenario, scoreTraceInsightReadiness, shareAllowed, stringField, throwIfRunIncomplete, tokenizeDomainWords, toolSpans, toolSpansToTraceAnalysisStore, traceAnalystFunctionGroup, traceAnalystOnRunComplete, traceSpanKindToOpenInferenceKind };
|
|
399
400
|
//# sourceMappingURL=traces.d.ts.map
|
package/dist/traces.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"traces.d.ts","names":[],"sources":["../src/trace-analyst/errors.ts","../src/trace/otlp-attributes.ts","../src/trace/otel-export.ts","../src/trace/otel-bridge.ts","../src/trace/store-to-otlp.ts","../src/trace-analyst/hook.ts","../src/trace-analyst/otlp-flatten.ts","../src/trace-analyst/otlp-span.ts","../src/trace-analyst/otlp-to-run-records.ts","../src/trace-analyst/prompts.ts"],"mappings":"
|
|
1
|
+
{"version":3,"file":"traces.d.ts","names":[],"sources":["../src/trace-analyst/errors.ts","../src/trace/otlp-attributes.ts","../src/trace/otel-export.ts","../src/trace/otel-bridge.ts","../src/trace/store-to-otlp.ts","../src/trace-analyst/hook.ts","../src/trace-analyst/otlp-flatten.ts","../src/trace-analyst/otlp-span.ts","../src/trace-analyst/otlp-to-run-records.ts","../src/trace-analyst/prompts.ts"],"mappings":";;;;;;;;;;;;;;;cASa,qCAAqC;WAC/B;;;cAIN,gCAAgC;WAC1B;WACR;WACA;WACA;EAET,YAAY,mBAAmB,gBAAgB,eAAe;;;cASnD,wCAAwC;WAClC;WACR;EAET,YAAY,mBAAmB,iBAAiB;IAAY;;;cAMjD,8BAA8B;WACxB;WACR;EAET,YAAY;;cAMD,+BAA+B;WACjC;WACA;WACA;EAET,YAAY,cAAc,oBAAoB;;cAcnC,gCAAgC;WAClC;WACA;WACA;EAET,YAAY,cAAc,qBAAqB,qBAAqB;;cAQzD,2BAA2B;WACrB;WACR;EAET,YAAY;;cAMD,0BAA0B;WACpB;WACR;WACA;EAET,YAAY,kBAAkB;;;;KCxEpB,oBAAoB,KAC9B;KAIU;UAUK;EACf;EACA,YAAY;EACZ;;;;;;;iBAkBc,qBAAqB,OAAO,oBAAoB;iBA2BhD,gBAAgB,OAAO;iBAkBvB,4BACd,YAAY,yBACZ,MAAM;iBAMQ,iCAAiC;;;UC7FhC;;EAEf;;EAEA,UAAU;;EAEV;;EAEA;;EAEA,qBAAqB;;EAErB;;UAGe;;EAEf,WAAW,MAAM;;EAEjB,SAAS;;EAET,YAAY;;UAGG;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA,OAAO;EACP,aAAa;;;;;;iBAOC,mBAAmB,SAAS,mBAAmB;;;;;;;;;;;iBC9C/C,uBACd,OAAO,YACP,UAAU,cACV,kBACC;;;UCUc;;EAEf;;EAEA;;EAEA;;;EAGA;;;;;;;;;;;;UAae;;EAEf;;EAEA;;EAEA;;;;UAKe;;;EAGf;;;EAGA,sBAAsB,KAAK,QAAQ;;EAEnC,iBAAiB,KAAK,QAAQ;;;;;;;;;;;iBAYhB,yBACd,iBAAiB,4BAA4B,oBAC7C,iBACA,OAAM,0BACL;;;UCxEc;;;;;;EAMf,SAAS,KAAK;IAAoC,SAAS;;;;;;;EAM3D;;;;;;EAMA,QAAQ,QAAQ,qBAAqB,KAAK,2BAA2B;;;;;EAKrE,aAAa,KAAK;;;;;;EAMlB,UAAU,QAAQ,qBAAqB,KAAK;;iBAM9B,0BAA0B,MAAM,0BAA0B;;;UCnCzD;;;;EAIf;;EAEA,UAAU,QAAQ;;iBA+DJ,0BACd,YAAY,YACZ,OAAM,qBACL;;;;;;;;;UC1Dc;EACf;EACA;EACA;EACA;EACA,MAAM;EACN;EACA;EACA;EACA,QAAQ;EACR;EACA;EACA;EACA;EACA;;EAEA,YAAY;;;;;;iBAOE,oBAAoB,KAAK,0BAA0B;;;;;;iBA8GnD,sBAAsB,KAAK,0BAA0B;iBAoBrD,YAAY,KAAK,yBAAyB;iBAK1C,SAAS;;iBAKT,gBACd,OAAO,yBACP;;;UCnIe;;EAEf;;;EAGA;;;EAGA,WAAW;;EAEX;;EAEA;;EAEA;;EAEA;;;;;;;EAOA;;;;;;EAMA;;;;;;;;;EASA,wBAAwB;;;;;;;;;EASxB,iBAAiB,eAAe,MAAM;;;;;EAKtC,yBAAyB,kBAAkB;;;;;;UAO5B;EACf,QAAQ;;EAER;;EAEA;;;UAIe;EACf;;EAEA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA,YAAY;;EAEZ;EACA;EACA;EACA;EACA;;EAEA,iBAAiB;;;;;;;iBAeH,iBAAiB,mBAAmB,MAAM,0BAA0B;;;;;;iBASpE,qBACd,MAAM,kBACN,MAAM,0BACL;;iBAKa,sBACd,mBACA,MAAM,0BACL;;iBAQa,0BACd,MAAM,kBACN,MAAM,0BACL;;;;cC/MU;cAcA"}
|
package/dist/traces.js
CHANGED
|
@@ -1,17 +1,17 @@
|
|
|
1
1
|
import { a as isToolSpan, i as isLlmSpan, n as TRACE_SCHEMA_VERSION, r as isJudgeSpan, t as FAILURE_CLASSES } from "./schema-CSf6qWgZ.js";
|
|
2
2
|
import { a as hasCapturedToolArgs, c as llmSpans, d as runsForScenario, f as toolSpans, i as groupBy, l as runFailureClass, n as aggregateLlm, o as isRunMetric, r as argHash, s as judgeSpans, t as RUN_METRICS, u as runMetricExtractor } from "./query-DlV0pdx4.js";
|
|
3
|
-
import {
|
|
4
|
-
import { i as
|
|
5
|
-
import { t as TraceEmitter } from "./emitter-
|
|
6
|
-
import { A as
|
|
3
|
+
import { i as modelHasSnapshot, u as validateRunRecord } from "./run-record-BRlM99rX.js";
|
|
4
|
+
import { a as combineVerdicts, c as emptyRedactionReport, d as redactForShare, f as redactText, i as classifyKey, l as normalizeKey, n as REDACTION_VERSION, o as detectCredential, p as shareAllowed, r as assessShareSafety, s as detectPersonalData, t as REDACTION_PROFILES, u as redact } from "./redact-DxwlUAvh.js";
|
|
5
|
+
import { t as TraceEmitter } from "./emitter-DbMXUIVP.js";
|
|
6
|
+
import { A as spanEpochMillis, C as TraceFileTooLargeError, D as extractOtlpAttributes, E as compareSpanTime, O as firstStringAttr, S as TraceFileMissingError, T as asString, _ as SpanNotFoundError, a as TRACE_ANALYST_TOOL_NAMESPACE, b as TraceAnalysisValidationError, c as traceAnalystFunctionGroup, g as TRACE_ANALYSIS_LIMITS, h as TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, j as stringField, k as projectOtlpFlatLine, m as DEFAULT_TRACE_ANALYST_BUDGETS, o as UNTRUSTED_TRACE_TEXT, s as buildTraceAnalysisToolDescriptors, u as createBoundedTraceAnalysisStore, v as TraceAnalysisLimitError, w as TraceNotFoundError, x as TraceFileMalformedError, y as TraceAnalysisStoreContractError } from "./kind-factory-BGwN0nE6.js";
|
|
7
7
|
import { INPUT_VALUE, LLM_CACHED_TOKENS, LLM_CACHED_TOKEN_ATTR_KEYS, LLM_CACHE_WRITE_TOKENS, LLM_CACHE_WRITE_TOKEN_ATTR_KEYS, LLM_CONTEXT_TOKENS, LLM_COST_ATTR_KEYS, LLM_COST_USD, LLM_INPUT_TOKENS, LLM_INPUT_TOKEN_ATTR_KEYS, LLM_MODEL_ATTR_KEYS, LLM_MODEL_NAME, LLM_OUTPUT_TOKENS, LLM_OUTPUT_TOKEN_ATTR_KEYS, LLM_REASONING_TOKENS, LLM_REASONING_TOKEN_ATTR_KEYS, OPENINFERENCE_SPAN_KIND, OUTPUT_VALUE, RUN_COST_ATTR_KEYS, SPAN_KIND_ATTR_KEYS, TOOL_ARGS_CAPTURED, TOOL_LATENCY_MS, TOOL_NAME, TOOL_NAME_ATTR_KEYS, applyLlmSpanOtlpAttributes, asNumber, contextInputTokens, firstNumberAttr } from "./trace-attributes.js";
|
|
8
8
|
import { n as extractUsageFromResponse, r as extractUsageFromSse, t as extractUsage } from "./extract-usage-BrQ8mCLX.js";
|
|
9
|
-
import { a as providerFromBaseUrl, i as defaultProviderRedactor, n as InMemoryRawProviderSink, r as NoopRawProviderSink, t as FileSystemRawProviderSink } from "./raw-provider-sink-
|
|
10
|
-
import { C as spanIdForWire, D as captureFetchToRawSink, E as toOtlpAttributes, S as exportRunAsOtlp, T as msToUnixNano, _ as tokenizeDomainWords, a as spanStatusToOtlp, b as TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, c as defaultTraceInsightPanel, d as inferDomainKeywords, f as isTraceInsightArtifact, g as scoreTraceInsightReadiness, h as planTraceInsightQuestions, i as epochMillisToIso, l as describeTraceInsightScope, m as isTraceInsightEvidence, n as toolSpansToTraceAnalysisStore, o as buildTraceInsightContext, p as isTraceInsightAttribution, r as createOtlpFlatLine, s as buildTraceInsightPrompt, t as ToolTraceMissingError, u as domainEvidencePattern, v as analyzeTraces, w as traceIdForWire, x as OTEL_AGENT_EVAL_SCOPE, y as TRACE_ANALYST_ACTOR_DESCRIPTION } from "./store-tool-spans-
|
|
11
|
-
import { n as assertRunCaptured, r as throwIfRunIncomplete, t as RunIntegrityError } from "./integrity-
|
|
9
|
+
import { a as providerFromBaseUrl, i as defaultProviderRedactor, n as InMemoryRawProviderSink, r as NoopRawProviderSink, t as FileSystemRawProviderSink } from "./raw-provider-sink-C96CcqeG.js";
|
|
10
|
+
import { C as spanIdForWire, D as captureFetchToRawSink, E as toOtlpAttributes, S as exportRunAsOtlp, T as msToUnixNano, _ as tokenizeDomainWords, a as spanStatusToOtlp, b as TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, c as defaultTraceInsightPanel, d as inferDomainKeywords, f as isTraceInsightArtifact, g as scoreTraceInsightReadiness, h as planTraceInsightQuestions, i as epochMillisToIso, l as describeTraceInsightScope, m as isTraceInsightEvidence, n as toolSpansToTraceAnalysisStore, o as buildTraceInsightContext, p as isTraceInsightAttribution, r as createOtlpFlatLine, s as buildTraceInsightPrompt, t as ToolTraceMissingError, u as domainEvidencePattern, v as analyzeTraces, w as traceIdForWire, x as OTEL_AGENT_EVAL_SCOPE, y as TRACE_ANALYST_ACTOR_DESCRIPTION } from "./store-tool-spans-B_0mXvYl.js";
|
|
11
|
+
import { n as assertRunCaptured, r as throwIfRunIncomplete, t as RunIntegrityError } from "./integrity-BFP4kjGj.js";
|
|
12
12
|
import { i as traceSpanKindToOpenInferenceKind, n as classifyOtlpSpanRole, r as isOtlpModelCall, t as applyToolSpanOtlpAttributes } from "./otlp-attributes-BpC_blLF.js";
|
|
13
13
|
import { n as InMemoryTraceStore, t as FileSystemTraceStore } from "./store-DNe_Uv1Q.js";
|
|
14
|
-
import { i as otlpTextToTraceAnalysisStore, n as OtlpFileTraceStore } from "./store-otlp-
|
|
14
|
+
import { i as otlpTextToTraceAnalysisStore, n as OtlpFileTraceStore } from "./store-otlp-D0AV4cvB.js";
|
|
15
15
|
import { i as summarizeTraceErrors, n as recordAggregateMeasurements, r as summarizeExecutionMeasurements, t as readTaskFailureLabels } from "./task-failure-attributes-B34TiMBe.js";
|
|
16
16
|
import { readFileSync, readdirSync, statSync, writeFileSync } from "node:fs";
|
|
17
17
|
import { join } from "node:path";
|
|
@@ -990,6 +990,6 @@ function ensureSnapshot(model, fallbackModel) {
|
|
|
990
990
|
return `${model}${fallbackModel.includes("@") ? fallbackModel.slice(fallbackModel.indexOf("@")) : "@otlp"}`;
|
|
991
991
|
}
|
|
992
992
|
//#endregion
|
|
993
|
-
export {
|
|
993
|
+
export { DEFAULT_TRACE_ANALYST_BUDGETS, FAILURE_CLASSES, FileSystemRawProviderSink, FileSystemTraceStore, INPUT_VALUE, InMemoryRawProviderSink, InMemoryTraceStore, LLM_CACHED_TOKENS, LLM_CACHED_TOKEN_ATTR_KEYS, LLM_CACHE_WRITE_TOKENS, LLM_CACHE_WRITE_TOKEN_ATTR_KEYS, LLM_CONTEXT_TOKENS, LLM_COST_ATTR_KEYS, LLM_COST_USD, LLM_INPUT_TOKENS, LLM_INPUT_TOKEN_ATTR_KEYS, LLM_MODEL_ATTR_KEYS, LLM_MODEL_NAME, LLM_OUTPUT_TOKENS, LLM_OUTPUT_TOKEN_ATTR_KEYS, LLM_REASONING_TOKENS, LLM_REASONING_TOKEN_ATTR_KEYS, NoopRawProviderSink, OPENINFERENCE_SPAN_KIND, OTEL_AGENT_EVAL_SCOPE, OUTPUT_VALUE, OtlpFileTraceStore, REDACTION_PROFILES, REDACTION_VERSION, RUN_COST_ATTR_KEYS, RUN_METRICS, RunIntegrityError, SPAN_KIND_ATTR_KEYS, SpanNotFoundError, TOOL_ARGS_CAPTURED, TOOL_LATENCY_MS, TOOL_NAME, TOOL_NAME_ATTR_KEYS, TRACE_ANALYSIS_LIMITS, TRACE_ANALYST_ACTOR_DESCRIPTION, TRACE_ANALYST_ACTOR_DESCRIPTION_VERSION, TRACE_ANALYST_TOOL_NAMESPACE, TRACE_ANALYST_TRUNCATION_MARKER_PREFIX, TRACE_SCHEMA_VERSION, ToolTraceMissingError, TraceAnalysisLimitError, TraceAnalysisStoreContractError, TraceAnalysisValidationError, TraceEmitter, TraceFileMalformedError, TraceFileMissingError, TraceFileTooLargeError, TraceNotFoundError, UNTRUSTED_TRACE_TEXT, aggregateLlm, analyzeTraces, applyLlmSpanOtlpAttributes, applyToolSpanOtlpAttributes, argHash, asNumber, asString, assertRunCaptured, assessShareSafety, buildTraceAnalysisToolDescriptors, buildTraceInsightContext, buildTraceInsightPrompt, captureFetchToRawSink, classifyKey, classifyOtlpSpanRole, combineVerdicts, contextInputTokens, convertTraceStoresToOtlp, createBoundedTraceAnalysisStore, createOtelExporter, createOtelTracingStore, defaultProviderRedactor, defaultTraceInsightPanel, describeTraceInsightScope, detectCredential, detectPersonalData, domainEvidencePattern, emptyRedactionReport, exportRunAsOtlp, extractOtlpAttributes, extractUsage, extractUsageFromResponse, extractUsageFromSse, firstNumberAttr, firstStringAttr, flattenOtlpExportToNdjson, groupBy, hasCapturedToolArgs, inferDomainKeywords, isJudgeSpan, isLlmSpan, isOtlpModelCall, isRunMetric, isToolSpan, isTraceInsightArtifact, isTraceInsightAttribution, isTraceInsightEvidence, judgeSpans, llmSpans, normalizeKey, otlpRowsToRunRecords, otlpRowsToTraceRunRecords, otlpTextToTraceAnalysisStore, otlpToRunRecords, otlpToTraceRunRecords, planTraceInsightQuestions, projectOtlpFlatLine, providerFromBaseUrl, redact, redactForShare, redactText, runFailureClass, runMetricExtractor, runsForScenario, scoreTraceInsightReadiness, shareAllowed, stringField, throwIfRunIncomplete, tokenizeDomainWords, toolSpans, toolSpansToTraceAnalysisStore, traceAnalystFunctionGroup, traceAnalystOnRunComplete, traceSpanKindToOpenInferenceKind };
|
|
994
994
|
|
|
995
995
|
//# sourceMappingURL=traces.js.map
|
|
@@ -1,5 +1,5 @@
|
|
|
1
|
-
import {
|
|
2
|
-
import { s as TraceStore } from "./store-
|
|
1
|
+
import { v as Span, w as TraceEvent } from "./schema-2g7trUwH.js";
|
|
2
|
+
import { s as TraceStore } from "./store-C4avFjIi.js";
|
|
3
3
|
//#region src/trajectory.d.ts
|
|
4
4
|
interface TrajectoryStep {
|
|
5
5
|
index: number;
|
|
@@ -21,4 +21,4 @@ interface Trajectory {
|
|
|
21
21
|
declare function buildTrajectory(store: TraceStore, runId: string): Promise<Trajectory>;
|
|
22
22
|
//#endregion
|
|
23
23
|
export { TrajectoryStep as n, buildTrajectory as r, Trajectory as t };
|
|
24
|
-
//# sourceMappingURL=trajectory-
|
|
24
|
+
//# sourceMappingURL=trajectory-Y1fmdlw7.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"trajectory-
|
|
1
|
+
{"version":3,"file":"trajectory-Y1fmdlw7.d.ts","names":[],"sources":["../src/trajectory.ts"],"mappings":";;;UAaiB;EACf;EACA,MAAM;;EAEN;;EAEA,QAAQ;;UAGO;EACf;EACA,OAAO;EACP;EACA;EACA;EACA;EACA;;iBAGoB,gBAAgB,OAAO,YAAY,gBAAgB,QAAQ"}
|
|
@@ -1,7 +1,7 @@
|
|
|
1
|
-
import { s as TraceStore } from "../store-
|
|
1
|
+
import { s as TraceStore } from "../store-C4avFjIi.js";
|
|
2
2
|
import { t as DefaultVerdict } from "../verdict-E4eRNf7-.js";
|
|
3
|
-
import {
|
|
4
|
-
import { i as CounterfactualRunner, t as CounterfactualContext } from "../counterfactual-
|
|
3
|
+
import { o as TraceEmitter } from "../emitter-HcCmFQl_.js";
|
|
4
|
+
import { i as CounterfactualRunner, t as CounterfactualContext } from "../counterfactual-CPaWN99-.js";
|
|
5
5
|
import { _ as isSubmitOnlyAction, a as RecordedTrajectoryStep, b as unreadableExitCount, c as TIMEOUT_OBSERVATION_MARKER, d as decodeRecordedTurns, f as deriveFailureSignature, g as isSubmitAction, h as isRecordedTimeout, i as RecordedObservationKind, l as assertReplayableTrajectory, m as isElidedField, n as FORMAT_ERROR_OBSERVATION_PREFIX, o as RecordedTrajectoryTurn, p as finalRecordedOutcome, r as RecordedFinalOutcome, s as SUBMIT_ACTION_SIGNATURE, t as DecodedTrajectory, u as classifyObservation, v as parseObservationOutput, y as parseRecordedReturncode } from "../steps-Bg5-KVx1.js";
|
|
6
6
|
//#region src/trajectory-replay/corpus.d.ts
|
|
7
7
|
interface CorpusSpec {
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import { t as mulberry32 } from "../random-Dn5fPWkt.js";
|
|
2
2
|
import { t as certificationEvidenceDigest } from "../verdict-DhPoXApk.js";
|
|
3
|
-
import { t as TraceEmitter } from "../emitter-
|
|
4
|
-
import { t as runCounterfactual } from "../counterfactual-
|
|
3
|
+
import { t as TraceEmitter } from "../emitter-DbMXUIVP.js";
|
|
4
|
+
import { t as runCounterfactual } from "../counterfactual-DtMWvn7D.js";
|
|
5
5
|
import { n as InMemoryTraceStore } from "../store-DNe_Uv1Q.js";
|
|
6
6
|
import { t as packageVersion } from "../package-version-D7lQHt_-.js";
|
|
7
7
|
import { a as assertReplayableTrajectory, c as deriveFailureSignature, d as isRecordedTimeout, f as isSubmitAction, g as unreadableExitCount, h as parseRecordedReturncode, i as TIMEOUT_OBSERVATION_MARKER, l as finalRecordedOutcome, m as parseObservationOutput, n as FORMAT_ERROR_OBSERVATION_PREFIX, o as classifyObservation, p as isSubmitOnlyAction, r as SUBMIT_ACTION_SIGNATURE, s as decodeRecordedTurns, t as wrapActionForExec, u as isElidedField } from "../exec-ChJ6d5if.js";
|
|
@@ -1,6 +1,6 @@
|
|
|
1
|
-
import { c as CostLedgerHandle } from "./cost-ledger-DbQdN3nO.js";
|
|
2
|
-
import { a as RunRecord,
|
|
3
|
-
import { c as JudgeInput, x as ChatClient } from "./types-
|
|
1
|
+
import { c as CostLedgerHandle, p as CostProvenance } from "./cost-ledger-DbQdN3nO.js";
|
|
2
|
+
import { a as RunRecord, d as RunTokenUsage } from "./run-record-DF-fuu31.js";
|
|
3
|
+
import { c as JudgeInput, x as ChatClient } from "./types-C0bWoSKC.js";
|
|
4
4
|
import { RE2JS } from "re2js";
|
|
5
5
|
//#region src/trace-analyst/types.d.ts
|
|
6
6
|
/**
|
|
@@ -516,7 +516,7 @@ interface AnalystUsageReceipt {
|
|
|
516
516
|
/** Null when the provider did not return token accounting. */
|
|
517
517
|
tokens: RunTokenUsage | null;
|
|
518
518
|
/** Observed, estimated, or explicitly uncaptured dollar cost. */
|
|
519
|
-
cost:
|
|
519
|
+
cost: CostProvenance;
|
|
520
520
|
/** Known lower bound when one or more calls have uncaptured cost. */
|
|
521
521
|
knownCostUsd?: number;
|
|
522
522
|
/**
|
|
@@ -595,7 +595,7 @@ interface AnalystRunResult {
|
|
|
595
595
|
* Provenance for `total_cost_usd`. When uncaptured, the numeric field is only
|
|
596
596
|
* the known subtotal and must not be treated as the run's total spend.
|
|
597
597
|
*/
|
|
598
|
-
total_cost_provenance?:
|
|
598
|
+
total_cost_provenance?: CostProvenance;
|
|
599
599
|
}
|
|
600
600
|
/**
|
|
601
601
|
* Events emitted by `AnalystRegistry.runStream(...)` in real time as
|
|
@@ -635,4 +635,4 @@ type AnalystRunEvent = {
|
|
|
635
635
|
};
|
|
636
636
|
//#endregion
|
|
637
637
|
export { DatasetOverview as A, TraceAnalystSpanKind as B, ReadSpanSourceInput as C, TraceAnalysisStore as D, TRACE_ANALYSIS_LIMITS as E, SpanMatchRecord as F, ViewTraceResult as G, TraceAnalystTraceSummary as H, TRACE_ANALYST_TRUNCATION_MARKER_PREFIX as I, TraceAnalystByteBudgets as L, QueryTracesPage as M, SearchSpanResult as N, TraceAnalysisStoreContext as O, SearchTraceResult as P, TraceAnalystFilters as R, BoundedTraceAnalysisStoreOptions as S, SpanSourceReader as T, ViewSpansResult as U, TraceAnalystSpanStatus as V, ViewTraceOversized as W, ProposalFinding as _, AnalystInputKind as a, makeFinding as b, AnalystRunInputs as c, AnalystSeverity as d, AnalystUsageReceipt as f, ExecutionProbeRequest as g, ExecutionProbeOutcome as h, AnalystFinding as i, ErrorCluster as j, DEFAULT_TRACE_ANALYST_BUDGETS as k, AnalystRunResult as l, ExecutionProbe as m, AnalystContext as n, AnalystRequirements as o, EvidenceRef as p, AnalystCost as r, AnalystRunEvent as s, Analyst as t, AnalystRunSummary as u, ProposalFindingOrigin as v, ReadSpanSourceResult as w, makeProposalFinding as x, computeFindingId as y, TraceAnalystSpan as z };
|
|
638
|
-
//# sourceMappingURL=types-
|
|
638
|
+
//# sourceMappingURL=types-Bay1dh2o.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types-
|
|
1
|
+
{"version":3,"file":"types-Bay1dh2o.d.ts","names":[],"sources":["../src/trace-analyst/types.ts","../src/trace-analyst/store-contract.ts","../src/analyst/types.ts"],"mappings":";;;;;;;;;;;;;;;;;;;;KAgBY;KAUA;;;;UAKK;EACf;EACA;EACA;EACA;EACA,MAAM;EACN;EACA;EACA;EACA,QAAQ;EACR;EACA;EACA;EACA;EACA;;;EAGA,YAAY;;UAGG;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA;;;;;;;UAQe;;EAEf;;EAEA;;EAEA;;EAEA;EACA;EACA;;EAEA;;EAEA;;EAEA;;UAGe;EACf;EACA;EACA;EACA;EACA;EACA;;EAEA;EACA;IAAU;IAAqB;;;;;;EAK/B,gBAAgB;EAChB;IAAc;IAAkB;;;UAGjB;EACf,QAAQ;EACR;EACA;;;;;UAMe;EACf;EACA,QAAQ;EACR,YAAY;;UAGG;EACf;;EAEA,gBAAgB;;EAEhB;EACA;;UAGe;EACf;EACA,OAAO;;EAEP;;EAEA;;EAEA;;EAEA;;UAGe;EACf;EACA;EACA;EACA,WAAW;;;EAGX;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA,MAAM;EACN;;UAGe;EACf;EACA;EACA,MAAM;EACN;;;UAIe;;;EAGf;;;EAGA;;;EAGA;;;EAGA;;cAGW,+BAA+B;;;cAS/B;;;cC7MA;WACX;WACA;WACA;WACA;WACA;WACA;WACA;WACA;;UAGe;EACf,SAAS;;;UAIM;EACf;EACA;EACA;EACA;EACA;EACA;;KAGU;EAEN;EACA;EACA;EACA;EACA;EACA;;EAGA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;IACE;IACA;IACA;IACA;IACA;;;;KAKI,oBACV,OAAO,qBACP,UAAU,8BACP,QAAQ;;;;;;;;UASI;;EAEf,iBAAiB;EAEjB,SAAS,kBAAkB,UAAU,4BAA4B;EAEjE,SACE;IAAS;IAAkB;KAC3B,UAAU,4BACT;EAEH,YACE,UAAU,qBACV,UAAU,4BACT,QAAQ;EAEX,YACE;IAAS,UAAU;IAAqB;IAAe;KACvD,UAAU,4BACT,QAAQ;EAEX,YAAY,UAAU,qBAAqB,UAAU,4BAA4B;EAEjF,UACE;IACE;IACA;KAEF,UAAU,4BACT,QAAQ;EAEX,UACE;IACE;IACA;IACA;KAEF,UAAU,4BACT,QAAQ;EAEX,YACE;IACE;IACA;IACA;KAEF,UAAU,4BACT,QAAQ;EAEX,WACE;IACE;IACA;IACA;IACA;KAEF,UAAU,4BACT,QAAQ;;UAGI;EACf,UAAU,QAAQ;;;;;;;;UC9GH;EACf;;;;;;;EAOA;EACA;EACA;EACA,UAAU;;;;;;;EAOV;EACA;EACA;EACA,eAAe;EACf;EACA;;EAEA;;;;;;EAMA;;;;EAIA;;EAEA,WAAW;;;;cAKA;KAED,0BAA0B;;KAG1B;;KAGA,kBAAkB;WACnB,iBAAiB;;UAGX;;;;;;;EAOf;EACA;EACA;;;;;;;;KAWU;UAOK;;EAEf;;EAEA;;EAEA;;EAEA;;UAGe;;EAEf;;EAEA;;;;;;UAOe;EACf,aAAa;EACb;EACA,YAAY;EACZ,aAAa;;EAEb,SAAS;;UAGM;EACf;;EAEA;;EAEA;;EAEA;;EAEA,aAAa;;EAEb;;;;;;;EAOA,OAAO;;;;;;;;;;EAUP,gBAAgB,cAAc;;;;;;;EAO9B,mBAAmB,cAAc;;;;;EAKjC,eAAe,SAAS;;EAExB,OAAO;;EAEP,OAAO,aAAa,SAAS;;EAE7B,SAAS;;;;;;;;EAQT,QAAQ;;;UAMO;EACf;;EAEA;;EAEA;;EAEA;EACA,SAAS;;;;;;;KAQC;EAEN;EACA;EACA;EACA;EACA;;EAEA;;EAEA;EAAkB;IAAS;IAAe;;;;;;;;UAO/B;;WAEN;EACT,QAAQ,SAAS,wBAAwB,QAAQ;;;;;;;;UASlC,QAAQ;;WAEd;;WAEA;WACA,WAAW;WACX,MAAM;WACN,WAAW;;WAEX;EACT,QAAQ,OAAO,QAAQ,KAAK,iBAAiB,QAAQ;;;UAItC;;EAEf;;EAEA,QAAQ;;EAER,MAAM;;EAEN;;;;;;;;EAQA;IAAkB;IAAsB;;;;;;;;EAOxC;;;;;;;;;;iBAac,iBAAiB;EAC/B;EACA;EACA;EACA;;EAEA;;;;;;iBAyBc,YACd,MAAM,KAAK;EACT;EACA;IAED;;iBAiBa,oBACd,MAAM,KAAK;EACT;EACA;IAED;UAOc;EACf;EACA;;EAEA;EACA;EACA;;EAEA,OAAO;;EAEP;IAAU;IAAe;;;UAGV;EACf;EACA;EACA;EACA;EACA,UAAU;EACV,aAAa;;EAEb;;;;;EAKA,wBAAwB;;;;;;;;;;;;;;;KAkBd;EAEN;EACA;EACA;EACA;;EAEA,aAAa;;EAGb;EACA,SAAS;;EAGT;EACA;EACA;;EAGA;;EAEA,SAAS;EACT,UAAU,cAAc;;EAGxB;EACA,QAAQ"}
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import { S as PaidCallResult, T as RunPaidCallInput, a as CostChannel, c as CostLedgerHandle, f as CostLedgerSummary, p as CostProvenance } from "./cost-ledger-DbQdN3nO.js";
|
|
2
|
-
import {
|
|
3
|
-
import { _ as ProposalFinding } from "./types-
|
|
4
|
-
import { M as LlmCallMetadata } from "./types-
|
|
2
|
+
import { d as RunTokenUsage } from "./run-record-DF-fuu31.js";
|
|
3
|
+
import { _ as ProposalFinding } from "./types-Bay1dh2o.js";
|
|
4
|
+
import { M as LlmCallMetadata } from "./types-C0bWoSKC.js";
|
|
5
5
|
import { t as SeriesDistribution } from "./descriptive-B2iPaT9J.js";
|
|
6
6
|
//#region src/campaign/types.d.ts
|
|
7
7
|
/** Stable identifier + kind tag for any scenario. Consumers
|
|
@@ -670,4 +670,4 @@ interface CampaignResult<TArtifact = unknown, TScenario extends Scenario = Scena
|
|
|
670
670
|
}
|
|
671
671
|
//#endregion
|
|
672
672
|
export { LabeledScenarioWrite as A, ScoredSurfaceOutcome as B, JudgeDimension as C, LabeledScenarioSampleArgs as D, LabeledScenarioRecord as E, ProposeContext as F, labelTrustRank as G, SurfaceProposer as H, ProposedCandidate as I, RedactionStatus as L, OptimizerConfig as M, ParetoParent as N, LabeledScenarioSource as O, ProposalTrackContext as P, Scenario as R, JudgeConfig as S, LabelTrust as T, TraceSpan as U, SessionScript as V, isProposedCandidate as W, GateDecision as _, CampaignResult as a, GenerationRecord as b, CampaignTraceWriter as c, DispatchContext as d, DispatchFn as f, GateContribution as g, GateContext as h, CampaignCostMeter as i, MutableSurface as j, LabeledScenarioStore as k, CodeSurface as l, GateCheckStatus as m, CampaignArtifactWriter as n, CampaignScenarioIdentity as o, Gate as p, CampaignCellResult as r, CampaignTokenUsage as s, CampaignAggregates as t, ComponentSurface as u, GateResult as v, JudgeScore as w, JudgeAggregate as x, GenerationCandidate as y, ScenarioAggregate as z };
|
|
673
|
-
//# sourceMappingURL=types-
|
|
673
|
+
//# sourceMappingURL=types-BvmWNjjs.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types-
|
|
1
|
+
{"version":3,"file":"types-BvmWNjjs.d.ts","names":[],"sources":["../src/campaign/types.ts"],"mappings":";;;;;;;;UAiCiB;EACf;EACA;EACA;;;;;EAKA;;;UAIe,iCAAiC,KAAK;EACrD;;;;;UAMe;EACf;;EAEA;EACA;EACA;EACA;EACA,QAAQ;EACR,OAAO;EACP,WAAW;EACX,MAAM;;EAEN;;EAEA;;;;;;;;EAQA;;;;KAKU,WAAW,kBAAkB,UAAU,cACjD,UAAU,WACV,KAAK,oBACF,QAAQ;;;;UAOI,cAAc,WAAW;EACxC;EACA;EACA;;EAEA;;;EAGA,sBAAsB,UAAU,WAAW,sBAAsB,UAAU,cAAc;;UAK1E;;EAEf;;EAEA;;;;;;;;;UAUe,YAAY,WAAW,kBAAkB,WAAW;EACnE;EACA,YAAY;;;;EAIZ;;;EAGA,MAAM;IACJ,UAAU;IACV,UAAU;IACV,QAAQ;;IAER,aAAa;IACb;IACA,WAAW;MACT,aAAa,QAAQ;EACzB,aAAa,UAAU;;;;;;;;;;;UAYR;EACf,YAAY;EACZ;EACA;;EAEA,UAAU;;;;;;;EAOV;;;;EAIA,eAAe,eAAe;IAAQ;IAAe;;;;;EAIrD;;;EAGA;;EAEA;;EAEA,WAAW,eAAe;;;;;;;UAUX;WACN;;;WAGA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;;WAGA;aACE;aACA;aACA;;;WAGF;;;UAIM;WACN;WACA,YAAY,SAAS;;;;;;;;;;;KAYpB,0BAA0B,mBAAmB;;;;;;;UAQxC;EACf,SAAS;;EAET;;;;EAIA;;;;;;EAMA,cAAc,SAAS;;;;iBAKT,oBACd,OAAO,iBAAiB,oBACvB,SAAS;;;;;;;;;UAkBK;EACf,SAAS;EACT;;;EAGA,YAAY;;;EAGZ;;EAEA;EACA;EACA;;;UAIe;EACf;EACA;EACA;EACA;EACA;;;;;UAMe;;;EAGf;;EAEA;EACA;EACA;EACA,YAAY;;;;;EAKZ,WAAW;IAAQ;IAAoB;IAAmB;IAAgB;;EAC1E;IACE;IACA;;;;;UAMa,eAAe,YAAY;;;WAGjC,gBAAgB;WAChB,SAAS,cAAc;WACvB,UAAU,cAAc;;WAExB;WACA;WACA,QAAQ;;WAER,QAAQ;;;WAGR,kBAAkB;;;WAGlB,mBAAmB;;;;WAInB,gBAAgB;;;;WAIhB;;;;;;;WAOA,gBAAgB,cAAc;;WAE9B,aAAa;WACb;;;;;;;;;;;;;;UAeM,gBAAgB,YAAY;EAC3C;;;;;EAKA,QAAQ,KAAK,eAAe,aAAa,QAAQ,MAAM,iBAAiB;;;EAGxE,QAAQ;IAAQ,SAAS,cAAc;;IAAwB;IAAe;;;UAG/D;EACf;EACA;EACA,mBAAmB,qBAAqB;;UAGzB,wBAAwB;EACvC,UAAU;;;KAMA;;KAGA;UAEK;EACf;EACA,QAAQ;EACR;;UAGe,YAAY,WAAW,kBAAkB;EACxD,oBAAoB,YAAY;EAChC,oBAAoB,YAAY;;EAEhC,aAAa,YAAY,eAAe;;;;;EAKxC,sBAAsB,YAAY,eAAe;;;;;;;EAOjD,yBAAyB,YAAY,eAAe;;;EAGpD,uBAAuB,YAAY;EACnC,WAAW;EACX;IAAQ;IAAmB;;;EAE3B,aAAa;EACb;EACA,QAAQ;;UAGO;EACf,UAAU;EACV;EACA,mBAAmB;EACnB;;;UAIe,KAAK,qBAAqB,kBAAkB,WAAW;EACtE;EACA,OAAO,KAAK,YAAY,WAAW,aAAa,QAAQ;;;;UAOzC;EACf,KAAK,cAAc,aAAa,0BAA0B;EAC1D,SAAS;;UAGM;EACf,IAAI,aAAa;EACjB,aAAa,aAAa;;;;UAKX;EACf,MAAM,cAAc,kBAAkB,aAAa;EACnD,UAAU,cAAc,iBAAiB;;;;;;KAO/B,qBAAqB;;;;;UAMhB;;EAEf,YAAY,GACV,OAAO,KAAK,iBAAiB;IAC3B,UAAU;MAEX,QAAQ,eAAe;;;;;KAQhB;KAOA;;;;;;;;;;;;;;;KAgBA;;iBASI,eAAe,OAAO;;;;UAOrB,qBAAqB,kBAAkB,WAAW,UAAU;EAC3E,UAAU;EACV,UAAU;EACV,aAAa,eAAe;EAC5B,QAAQ;EACR;EACA;EACA,iBAAiB;;;;;EAKjB,aAAa;;EAEb;;UAGe,sBAAsB,kBAAkB,WAAW,UAAU,6BACpE,qBAAqB,WAAW;;EAExC;;;EAGA;;UAGe;EACf;;EAEA;;;;EAIA;EACA;IACE;IACA,SAAS,wBAAwB;IACjC;IACA;;;;;IAKA,WAAW;;;UAIE;EACf,QAAQ,OAAO,uBAAuB;EACtC,OAAO,MAAM,4BAA4B,QAAQ;EACjD,QAAQ;IACN;IACA;IACA,UAAU;;;IAGV,SAAS,OAAO;;;UAMH,mBAAmB;;;EAGlC;EACA;EACA;EACA;EACA;EACA,UAAU;EACV,aAAa,eAAe;;;EAG5B;;EAEA,gBAAgB;;EAEhB;;;EAGA,YAAY;;;EAGZ;;;EAGA;EACA;EACA;EACA;;;;EAIA;;EAEA;;EAEA;EACA;;UAGe;EACf;EACA;EACA;EACA;;;;;;EAMA,cAAc;;UAGC;EACf;EACA;EACA;;;EAGA,cAAc;;UAGC;EACf;EACA,YAAY;EACZ;;;;;;UAOe;EACf;;EAEA;;EAEA;;EAEA;;EAEA;;;EAGA;;;;EAIA;;;;EAIA;IACE;IACA;IACA,iBAAiB;MAAQ;MAAgB;;;;;EAI3C,YAAY;;;;;;;;;;EAUZ,WAAW;IAAQ;IAAoB;IAAmB;IAAgB;;;;EAG1E;;;;EAIA;;;;EAIA,cAAc,SAAS;;UAGR;EACf,SAAS,eAAe;EACxB,YAAY,eAAe;;EAE3B,MAAM;;EAEN;EACA;EACA;;EAEA;;EAEA;;EAEA;;EAEA;;UAGe,eAAe,qBAAqB,kBAAkB,WAAW;;EAEhF;;EAEA;EACA;;EAEA;EACA;EACA;EACA;EACA,OAAO,MAAM,mBAAmB;EAChC,YAAY;EACZ;IACE,aAAa;IACb;;EAEF,OAAO;EACP;EACA;EACA,iBAAiB;;;EAGjB,WAAW,MAAM,2BAA2B,KAAK"}
|
|
@@ -175,12 +175,12 @@ declare function assertCrossFamilyServed(pairs: ReadonlyArray<{
|
|
|
175
175
|
* request, response, and error is recorded — including retries, with the
|
|
176
176
|
* attempt index attached so a flaky call's full event chain is recoverable.
|
|
177
177
|
*
|
|
178
|
-
* Redaction is enforced at sink time. The default redactor
|
|
179
|
-
*
|
|
180
|
-
*
|
|
181
|
-
*
|
|
182
|
-
*
|
|
183
|
-
*
|
|
178
|
+
* Redaction is enforced at sink time. The default redactor runs the redaction
|
|
179
|
+
* core (`./redact`, default profile) over headers, bodies and the error
|
|
180
|
+
* message: credential headers and fields are replaced by name, credential and
|
|
181
|
+
* personal-data values by shape. Override via the sink constructor or the
|
|
182
|
+
* per-call `redactor`. The `redactedFields` array on the persisted event lets
|
|
183
|
+
* a reviewer see what was replaced without exposing the values.
|
|
184
184
|
*/
|
|
185
185
|
type RawProviderDirection = 'request' | 'response' | 'error';
|
|
186
186
|
interface RawProviderEvent {
|
|
@@ -214,7 +214,7 @@ interface RawProviderEvent {
|
|
|
214
214
|
responseBody?: unknown;
|
|
215
215
|
/** Set on `direction: 'error'` events. */
|
|
216
216
|
errorMessage?: string;
|
|
217
|
-
/**
|
|
217
|
+
/** JSON Pointers of the fields the redactor replaced ('/requestHeaders/Authorization', '/requestBody/apiKey', …). */
|
|
218
218
|
redactedFields: string[];
|
|
219
219
|
}
|
|
220
220
|
interface RawProviderSinkFilter {
|
|
@@ -232,9 +232,8 @@ interface RawProviderSink {
|
|
|
232
232
|
}
|
|
233
233
|
type ProviderRedactor = (event: RawProviderEvent) => RawProviderEvent;
|
|
234
234
|
/**
|
|
235
|
-
*
|
|
236
|
-
*
|
|
237
|
-
* `event.redactedFields` so a downstream reviewer can see what was removed.
|
|
235
|
+
* The default redactor: the redaction core over the parts of an event that
|
|
236
|
+
* carry provider traffic. Every replaced path is added to `redactedFields`.
|
|
238
237
|
*/
|
|
239
238
|
declare function defaultProviderRedactor(event: RawProviderEvent): RawProviderEvent;
|
|
240
239
|
interface InMemoryRawProviderSinkOptions {
|
|
@@ -863,4 +862,4 @@ interface CheckResult {
|
|
|
863
862
|
}
|
|
864
863
|
//#endregion
|
|
865
864
|
export { NoopRawProviderSink as $, createChatClient as A, LlmToolCall as B, ChatResponse as C, judgeFamily as Ct, MockTransportOpts as D, CustomTransportOpts as E, LlmChargeBounds as F, costReceiptFromLlmError as G, LlmToolDefinition as H, LlmMessage as I, stripFencedJson as J, isTransientLlmError as K, LlmResponseError as L, LlmCallMetadata as M, LlmCallRequest as N, OpenAiCompatibleTransportOpts as O, LlmCallResult as P, InMemoryRawProviderSinkOptions as Q, LlmThinkingMode as R, ChatRequest as S, assertCrossFamily as St, CreateChatClientOpts as T, LlmUsage as U, LlmToolChoice as V, costReceiptFromLlm as W, FileSystemRawProviderSinkOptions as X, FileSystemRawProviderSink as Y, InMemoryRawProviderSink as Z, ScenarioResult as _, checkServedModel as _t, CompletionCriterion as a, defaultProviderRedactor as at, ChatCallOpts as b, CrossFamilyError as bt, JudgeInput as c, AssertServedModelOptions as ct, PersonaConfig as d, ServedModelCheck as dt, ProviderRedactor as et, ProductClientConfig as f, ServedModelPolicy as ft, ScenarioFile as g, assertServedModels as gt, Scenario as h, assertServedModel as ht, CollectedArtifacts as i, RawProviderSinkFilter as it, LlmCallError as j, SandboxSdkTransportOpts as k, JudgeRubric as l, ModelSubstitutionError as lt, RubricDimension as m, assertCrossFamilyServed as mt, ArtifactResult as n, RawProviderEvent as nt, DriverState as o, providerFromBaseUrl as ot, RouteMap as p, ServedModelVerdict as pt, maximumChargeForLlmRequest as q, CheckResult as r, RawProviderSink as rt, JudgeFn as s, AssertCrossFamilyServedOptions as st, ArtifactCheck as t, RawProviderDirection as tt, JudgeScore as u, ServedCrossFamilyError as ut, Turn as v, servedModelAcceptable as vt, ChatTransport as w, ChatClient as x, JudgeFamily as xt, TurnResult as y, AssertCrossFamilyOptions as yt, LlmTokenLogprob as z };
|
|
866
|
-
//# sourceMappingURL=types-
|
|
865
|
+
//# sourceMappingURL=types-C0bWoSKC.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types-C0bWoSKC.d.ts","names":[],"sources":["../src/judge-families.ts","../src/integrity/served-model.ts","../src/trace/raw-provider-sink.ts","../src/llm-client.ts","../src/analyst/chat-client.ts","../src/types.ts"],"mappings":";;;;;;;;;;;;;KAWY;;;;;;;iBAsEI,YAAY,kBAAkB;UAc7B;;EAEf;;;;EAIA;;cAGW,yBAAyB;WAGlB,UAAU;WACV;EAHlB,YACE,iBACgB,UAAU,eACV;;;;;;;;;;;;;;;iBAoBJ,kBACd,kBACA,OAAM,2BACL;;;;KCvFS;;;;;;;;;;;UAYK;;EAEf;;EAEA;EACA,iBAAiB;;EAEjB,cAAc;EACd,SAAS;;EAET;;;;;;;;;;;;iBA4Cc,iBACd,mBACA,oCACC;cA4CU,+BAA+B;WAGxB,QAAQ,cAAc;EAFxC,YACE,iBACgB,QAAQ,cAAc;;;;;;;;;KAc9B;UAWK;;;;;;EAMf;;;;;EAKA;;EAEA;;;;;;;iBAQc,sBACd,OAAO,kBACP,OAAM;;;;;;iBA6BQ,kBACd,mBACA,mCACA,OAAM,2BACL;;;;;iBAea,mBACd,OAAO;EAAgB;EAAmB;IAC1C,OAAM,2BACL;UAec,uCAAuC;;EAEtD;;EAEA;;cAGW,+BAA+B;WAGxB,UAAU;WACV,QAAQ,cAAc;EAHxC,YACE,iBACgB,UAAU,eACV,QAAQ,cAAc;;;;;;;;;;;iBAgB1B,wBACd,OAAO;EAAgB;EAAmB;IAC1C,OAAM,iCACL;;;;;;;;;;;;;;;;;;;;;;;;;;;;;KCjRS;UAEK;;EAEf;;EAEA;EACA;;;;;;EAMA;EACA;;EAEA;;EAEA;;EAEA;EACA,WAAW;;EAEX;;EAEA;EACA;EACA,iBAAiB;EACjB;EACA,kBAAkB;EAClB;;EAEA;;EAEA;;UAGe;EACf;EACA;EACA,YAAY;EACZ;;UAGe;EACf,OAAO,OAAO,mBAAmB;;EAEjC,MAAM,SAAS,wBAAwB,QAAQ;;EAE/C,UAAU;;KAGA,oBAAoB,OAAO,qBAAqB;;;;;iBAM5C,wBAAwB,OAAO,mBAAmB;UAqBjD;EACf,WAAW;;cAGA,mCAAmC;UACtC;UACA;EAER,YAAY,OAAM;EAIZ,OAAO,OAAO,mBAAmB;EAIjC,KAAK,SAAQ,wBAA6B,QAAQ;EAUxD;;cAKW,+BAA+B;EACpC,UAAU;;;;;;;EASV,QAAQ,QAAQ;;UAOP;;EAEf;;EAEA;;EAEA;EACA,WAAW;;cAGA,qCAAqC;UACxC;UACA;UACA;UACA;UACA;UACA;UACA;EAER,YAAY,MAAM;UAOJ;UAON;EAKF,OAAO,OAAO,mBAAmB;EAYjC,KAAK,SAAQ,wBAA6B,QAAQ;;;;;;iBAkC1C,oBAAoB;;;UCpLnB;EACf;;;;;EAKA,kBAEI;IACM;IAAc;;IACd;IAAmB;MAAa;MAAa;;;;EAGvD,YAAY;;EAEZ;;KAGU;;UAGK;EACf;EACA;IACE;IACA;IACA,YAAY;;;;KAKJ;EAIN;EAAkB;IAAY;;;;UAGnB;EACf;EACA;;EAEA;;UAGe;EACf;EACA,UAAU;;EAEV;;EAEA;IAAe;IAAc,QAAQ;;;EAErC,QAAQ;;EAER,aAAa;EACb;EACA;;;;;EAKA;IAAa;;;EAEb,WAAW;;EAEX;;;;;;UAOe;;EAEf;;EAEA;;EAEA,WAAW;;EAEX,qBAAqB;;iBAGP,2BACd,SAAS,KACP,0GAGF,UAAS,kBACR;UAgCc;EACf;EACA;EACA;;EAEA;;EAEA;;EAEA;;;;UAKe;EACf;EACA;EACA,KAAK;IAAgB;IAAe;;;UAGrB;;EAEf;;EAEA,YAAY;EACZ,OAAO;;;;;EAKP;;;;;;;;;EASA;;;;;;;;;;;;EAYA;;EAEA;;;;;;;;;EASA;;;;;;;EAOA;;;;;;;;EAQA,WAAW,cAAc;;EAEzB,KAAK;;KAGK,kBAAkB,KAAK;;;;;;iBAOnB,mBACd,QAAQ,eACR,qBAAqB,qBACpB;;iBA6Ba,wBACd,OAAO,OACP,qBAAqB,qBACpB;cAMU,qBAAqB;WAGd;WACA;WACA;EAJlB,YACE,iBACgB,gBACA,cACA;;;;;cASP,yBAAyB;WAGlB,QAAQ;EAF1B,YACE,iBACgB,QAAQ,eACxB;IAAY;;;UAMC,yBAAyB;;EAExC;;EAEA;EACA;;EAEA;IAAe;IAAc;;;EAE7B;;EAEA;;;;;;;EAOA,SAAS;;;;;;;;EAQT;;;;;;EAMA;;EAEA,eAAe;;;;;;;;EAQf,UAAU;;;;;EAKV;;EAEA;IAAiB;IAAgB;;;EAEjC,WAAW;;;;;;;;;;EAUX,8BAA8B;;;;;;;;;;;;;iBAwEhB,oBAAoB;;;;;;iBA6PpB,gBAAgB;;;;;;UC3qBf;;WAEN,WAAW;;WAEX;;WAEA;;EAGT,KAAK,KAAK,aAAa,OAAO,eAAe,QAAQ;;KAG3C;UAMK,oBAAoB,KAAK;;EAExC;;KAGU,eAAe;UAEV;;EAEf,SAAS;;EAET;;EAEA;;EAEA;;KAKU,uBACR,0BACA,sBACA,gCACA;UAEM;EACR;;EAEA;;;;;;UAOe,gCAAgC;EAC/C;EACA,OAAO,KAAK,aAAa,OAAO,iBAAiB,QAAQ;;;UAI1C,4BAA4B;EAC3C;EACA,OAAO,KAAK,aAAa,OAAO,iBAAiB,QAAQ;;;;;;;;;;;;;;;;;;;;UAqB1C,sCACP,mBACN,KACE;EAaJ;;EAEA;;EAEA;;EAEA;;EAEA;IAAe;IAAc;;;;;;;UAOd,0BAA0B;EACzC;EACA,UAAU,KAAK,aAAa,OAAO,iBAAiB,QAAQ;;;;;;iBAO9C,iBAAiB,MAAM,uBAAuB;;;UC/J7C;EACf;EACA;EACA;EACA;EACA;EACA,OAAO;EACP,gBAAgB;EAChB;;UAGe;EACf;EACA;EACA;EACA;;UAKe;EACf;EAQA;EACA;EACA;EACA;;UAKe;EACf;EACA;EACA,YAAY;;UAGG;EACf;EACA;EACA;EACA;EACA;;UAKe;EACf;EACA;EACA,OAAO;EACP,iBAAiB;EACjB,aAAa;EACb;EACA;EACA;EACA,WAAW;;EAEX,OAAO;;UAGQ;EACf;EACA;EACA;EACA;EACA;IAAmB;IAAc;;EACjC;EACA;;UAGe;EACf,OAAO;EACP;EACA;;UAGe;EACf;EACA;EACA;EACA;EACA;;UAGe;EACf;IAAc;IAAc;;EAC5B;IAAmB;IAAc,QAAQ;;EACzC;IAAc;IAAkB;;EAChC;;UAKe;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;GACC;;UAGc;EACf;EACA,QAAQ;;EAER;;UAKe;EACf;EACA;EACA;EACA;EACA;EACA;EACA;IACE;MACE;MACA;MACA;;;EAGJ,OAAO;EACP,gBAAgB;;UAKD;EACf;EACA,QAAQ,OAAO;EACf,YAAY,OAAO;;;;;;;;;;KAWT;UAEK;EACf;EACA;EACA;EACA,oBAAoB;EACpB;;EAEA,QAAQ;;;;;;;EAOR;;;;;;;EAOA;;;;;;EAMA;;UAGe;EACf;EACA;EACA;IAAa;IAAiB;IAAkB;;EAChD;EACA;EACA;;UAGe;EACf,UAAU;EACV,OAAO;EACP,WAAW;;EAEX,aAAa;EACb;EACA,WAAW;EACX,SAAS;;KAGC,WAAW,MAAM,YAAY,OAAO,eAAe,QAAQ;UAYtD;EACf;EACA;EACA;EACA"}
|
|
@@ -1,17 +1,4 @@
|
|
|
1
1
|
import { createHash } from "node:crypto";
|
|
2
|
-
//#region src/ledger-core/deep-freeze.ts
|
|
3
|
-
/** Freeze a detached canonical-JSON graph. Canonicalization has already ruled out cycles.
|
|
4
|
-
*
|
|
5
|
-
* Lives outside canonical.ts so the analyst-benchmark implementation digest,
|
|
6
|
-
* which covers canonical.ts, stays bound to the published benchmark evidence. */
|
|
7
|
-
function deepFreezeCanonicalJson(value) {
|
|
8
|
-
if (value && typeof value === "object" && !Object.isFrozen(value)) {
|
|
9
|
-
Object.freeze(value);
|
|
10
|
-
for (const nested of Object.values(value)) deepFreezeCanonicalJson(nested);
|
|
11
|
-
}
|
|
12
|
-
return value;
|
|
13
|
-
}
|
|
14
|
-
//#endregion
|
|
15
2
|
//#region src/analyst/types.ts
|
|
16
3
|
/**
|
|
17
4
|
* Analyst contract — the missing orchestration layer over agent-eval's
|
|
@@ -86,6 +73,6 @@ function makeProposalFinding(init) {
|
|
|
86
73
|
};
|
|
87
74
|
}
|
|
88
75
|
//#endregion
|
|
89
|
-
export {
|
|
76
|
+
export { makeProposalFinding as i, computeFindingId as n, makeFinding as r, ANALYST_SEVERITIES as t };
|
|
90
77
|
|
|
91
|
-
//# sourceMappingURL=types-
|
|
78
|
+
//# sourceMappingURL=types-CW5b90JH.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types-CW5b90JH.js","names":[],"sources":["../src/analyst/types.ts"],"sourcesContent":["/**\n * Analyst contract — the missing orchestration layer over agent-eval's\n * existing analyzers (analyzeTraces, MultiLayerVerifier,\n * SemanticConceptJudge, JudgeFn, ...).\n *\n * Each existing primitive returns its own output shape. The Analyst\n * contract is the single envelope every primitive lifts into, so a\n * registry can run N analysts against a run and a single renderer can\n * compose findings without knowing which analyzer produced them.\n *\n * The contract is intentionally domain-agnostic: nothing here knows\n * about code, voice, RAG, or any particular agent stack. Analysts\n * declare what INPUT KIND they need (a trace store, an artifact dir,\n * a RunRecord, a JudgeInput, or `custom`), and the registry routes\n * the matching input from `AnalystRunInputs`.\n */\n\nimport { createHash } from 'node:crypto'\nimport type { CostLedgerHandle, CostProvenance } from '../cost-ledger'\nimport type { RunRecord, RunTokenUsage } from '../run-record'\nimport type { TraceAnalysisStore } from '../trace-analyst/store'\nimport type { JudgeInput } from '../types'\nimport type { ChatClient } from './chat-client'\n\n/**\n * Unified envelope every analyst emits. Schema-versioned so renderers\n * and time-series diffs survive future field additions.\n */\nexport interface AnalystFinding {\n schema_version: '1.0.0'\n /**\n * Stable hash over identity-defining fields (analyst_id + canonical\n * claim + area + optional subject). Two findings from two runs that\n * \"are the same finding\" share this id — that's what `diffFindings`\n * uses to compute appeared/disappeared sets across runs.\n */\n finding_id: string\n analyst_id: string\n produced_at: string\n severity: AnalystSeverity\n /**\n * Coarse classification. Renderers group by this. Free-form so\n * domain-specific analysts can introduce categories without a\n * schema change ('agent-reasoning', 'verification', 'cost',\n * 'tool-use', 'safety', 'latency', 'data-quality', ...).\n */\n area: string\n claim: string\n rationale?: string\n evidence_refs: EvidenceRef[]\n recommended_action?: string\n validation_plan?: string\n /** 0..1 — the analyst's own confidence. Not calibrated across analysts. */\n confidence: number\n /**\n * Optional subject the finding is about — leaf id, agent id, request\n * id. Included in finding_id when present so per-subject findings\n * diff cleanly across runs.\n */\n subject?: string\n /** True when this finding was lifted from a judge result rather than observed\n * directly in a trace or artifact. Descriptive only: proposal access is\n * controlled by `ProposalFinding.proposal_origin`. */\n derived_from_judge?: boolean\n /** Analyst-private extras; renderers ignore unless they know the analyst. */\n metadata?: Record<string, unknown>\n}\n\n/** Finding severity. `AnalystSeverity` derives from this array, so the type\n * and the schema that validates a finding cannot name different levels. */\nexport const ANALYST_SEVERITIES = ['critical', 'high', 'medium', 'low', 'info'] as const\n\nexport type AnalystSeverity = (typeof ANALYST_SEVERITIES)[number]\n\n/** Data sources that candidate generation may intentionally learn from. */\nexport type ProposalFindingOrigin = 'search' | 'production'\n\n/** A finding explicitly admitted as candidate-generation input. */\nexport type ProposalFinding = AnalystFinding & {\n readonly proposal_origin: ProposalFindingOrigin\n}\n\nexport interface EvidenceRef {\n /**\n * Where the evidence lives. `span` and `event` refer to OTLP trace\n * elements; `artifact` to a file inside the run's artifact tree;\n * `finding` to another AnalystFinding (cross-analyst chaining);\n * `metric` to a named scalar reading the renderer knows how to read.\n */\n kind: 'span' | 'event' | 'artifact' | 'finding' | 'metric'\n uri: string\n excerpt?: string\n}\n\n// ── Analyst contract ─────────────────────────────────────────────────\n\n/**\n * The discriminator the registry uses to pass the right input.\n * `custom` is the escape hatch — analysts that need something else\n * (e.g. an embedding cache, a partner SDK handle) read it from\n * `AnalystRunInputs.custom[<analyst id>]`.\n */\nexport type AnalystInputKind =\n | 'trace-store'\n | 'artifact-dir'\n | 'run-record'\n | 'judge-input'\n | 'custom'\n\nexport interface AnalystCost {\n /** `deterministic` analysts MUST NOT call the LLM. */\n kind: 'deterministic' | 'llm'\n /** Optional declared upper bound; the registry can enforce a budget. */\n est_usd_per_run?: number\n /** Models the analyst expects to use (informational). */\n models?: string[]\n /** Maximum post-cancellation wait for provider usage. Model analysts default to 5 seconds. */\n settlement_timeout_ms?: number\n}\n\nexport interface AnalystRequirements {\n /** Min number of shots / samples the analyst needs to produce signal. */\n min_shots?: number\n /** Capabilities the runtime must supply (e.g. ['network', 'gpu']). */\n capabilities?: string[]\n}\n\n/**\n * What's passed to every analyst call. The registry resolves which\n * field the analyst's `inputKind` selects and asserts it's present.\n */\nexport interface AnalystRunInputs {\n traceStore?: TraceAnalysisStore\n artifactDir?: string\n runRecord?: RunRecord\n judgeInput?: JudgeInput\n /** Keyed by analyst id; populated by callers that registered custom analysts. */\n custom?: Record<string, unknown>\n}\n\nexport interface AnalystContext {\n runId: string\n /** Stable correlation id so logs from a single registry.run() share a tag. */\n correlationId: string\n /** Enforced wall-clock deadline (epoch ms). */\n deadlineMs?: number\n /** Per-analyst USD budget. Analysts MAY check before issuing LLM calls. */\n budgetUsd?: number\n /** Shared paid-call account when the analyst runs inside a larger campaign. */\n costLedger?: CostLedgerHandle\n /** Attribution phase used when writing to the shared paid-call account. */\n costPhase?: string\n /**\n * Shared chat client. Analysts that call an LLM go through this so\n * the operator picks transport (sandbox-sdk | router | cli-bridge |\n * direct-provider | mock) at the registry boundary without touching\n * analyst code.\n */\n chat?: ChatClient\n /**\n * Findings from a prior run the operator wants the analyst to see as\n * retrieval context. Kinds that take advantage of cross-run memory\n * (failure-mode \"I saw this cluster last run\", knowledge-gap \"the wiki\n * page I asked for is still missing\") render these into the actor's\n * working set. Filtering is the operator's job: pass the slice that\n * matches the analyst's id, or pass everything and let the kind\n * filter. Empty / absent means no cross-run context.\n */\n priorFindings?: ReadonlyArray<AnalystFinding>\n /**\n * Findings emitted by analysts that completed earlier in this registry run.\n * This is separate from `priorFindings`: upstream findings are dependency\n * context for the current pass, while prior findings are cross-run memory.\n * The registry populates this only when `RegistryRunOpts.chainFindings` is on.\n */\n upstreamFindings?: ReadonlyArray<AnalystFinding>\n /**\n * Report metered work independently of findings. This keeps an empty finding\n * set from erasing token/cost telemetry. Multiple receipts are accumulated.\n */\n recordUsage?: (receipt: AnalystUsageReceipt) => void\n /** Free-form runtime tags (env, host, op). Findings can echo these into metadata. */\n tags?: Record<string, string>\n /** Logger callback — analysts SHOULD prefer this over console.* for testability. */\n log?: (msg: string, fields?: Record<string, unknown>) => void\n /** Optional abort signal. Analysts SHOULD pass it through to LLM calls. */\n signal?: AbortSignal\n /**\n * Optional live-execution port. A runtime that owns a sandbox or checkout\n * fills it so an analyst can execute a bounded probe against the run's\n * produced state instead of reasoning about it from the trace alone. This\n * package defines only the port: no field here reaches for an agent loop,\n * and an absent probe means the analyst works from recorded evidence.\n */\n probe?: ExecutionProbe\n}\n\n// ── Live-execution port ─────────────────────────────────────────────\n\n/** One bounded command an analyst asks the probe to run. */\nexport interface ExecutionProbeRequest {\n command: string\n /** Working directory inside the probed environment. */\n cwd?: string\n /** Hard wall-clock deadline for this one execution. */\n timeoutMs: number\n /** Bytes of combined output retained; the prober truncates beyond it. */\n maxOutputBytes?: number\n signal?: AbortSignal\n}\n\n/**\n * Typed outcome of one probe execution. `succeeded: false` is a PROBE failure\n * (the environment could not run the command); a command that ran and exited\n * non-zero is a successful observation with a non-zero `exitCode`.\n */\nexport type ExecutionProbeOutcome =\n | {\n succeeded: true\n exitCode: number\n stdout: string\n stderr: string\n durationMs: number\n /** True when output was cut at `maxOutputBytes`. */\n truncated: boolean\n }\n | { succeeded: false; error: { class: string; message: string } }\n\n/**\n * The seam a runtime fills to let analysts observe produced state live.\n * Implementations own sandboxing, credentials, and cleanup; analysts only\n * submit bounded requests and read typed outcomes.\n */\nexport interface ExecutionProbe {\n /** One plain sentence naming what is being probed (e.g. a sandbox id). */\n readonly description: string\n execute(request: ExecutionProbeRequest): Promise<ExecutionProbeOutcome>\n}\n\n/**\n * The minimal contract. Concrete analysts can refine `TInput` so\n * implementations stay type-safe (e.g. a trace analyst's `TInput` is\n * `TraceAnalysisStore`); the registry passes the right field from\n * `AnalystRunInputs` based on `inputKind`.\n */\nexport interface Analyst<TInput = unknown> {\n /** Stable identifier — appears in finding_id, telemetry, and registry exclusion lists. */\n readonly id: string\n /** Human-readable. One sentence. */\n readonly description: string\n readonly inputKind: AnalystInputKind\n readonly cost: AnalystCost\n readonly requires?: AnalystRequirements\n /** Bump on breaking changes to claim wording or area so old finding_ids don't collide. */\n readonly version: string\n analyze(input: TInput, ctx: AnalystContext): Promise<AnalystFinding[]>\n}\n\n/** Metered work performed by one analyst call. */\nexport interface AnalystUsageReceipt {\n /** Number of model-usage records observed at the provider boundary. */\n calls: number | null\n /** Null when the provider did not return token accounting. */\n tokens: RunTokenUsage | null\n /** Observed, estimated, or explicitly uncaptured dollar cost. */\n cost: CostProvenance\n /** Known lower bound when one or more calls have uncaptured cost. */\n knownCostUsd?: number\n /**\n * Token counts the provider reported on only one side. Present exactly when\n * `tokens` is null and at least one side WAS reported: `RunTokenUsage` has no\n * nullable side, so a one-sided count cannot live in `tokens` without writing\n * a zero nobody measured. Read it as a lower bound, never as a total — the\n * field exists so a null `tokens` cannot hide a real count.\n */\n partialTokens?: { input: number | null; output: number | null }\n /**\n * True when the token counts were DERIVED by the transport (from character\n * lengths, say) rather than measured by the model provider. `cost.kind` is\n * `estimated` both for a rate estimate over exact tokens and for one over\n * derived tokens; this is the field that separates them.\n */\n tokensEstimated?: boolean\n}\n\n// ── finding_id stability ─────────────────────────────────────────────\n\n/**\n * Compute the stable finding_id from the identity-defining fields.\n * Default implementation hashes {analyst_id, area, subject, normalized claim}.\n * Analysts that emit findings whose claim text varies per run (timestamps,\n * counts) SHOULD either: (a) pass an explicit `id_basis` to fix the hash,\n * or (b) move the variable part into `rationale`/`metadata` and keep the\n * `claim` static.\n */\nexport function computeFindingId(input: {\n analyst_id: string\n area: string\n subject?: string\n claim: string\n /** Override the claim for hashing — use when the displayed claim has run-specific bits. */\n id_basis?: string\n}): string {\n const basis = JSON.stringify({\n a: input.analyst_id,\n r: input.area,\n s: input.subject ?? '',\n c: normalizeClaim(input.id_basis ?? input.claim),\n })\n return `f_${createHash('sha256').update(basis).digest('hex').slice(0, 20)}`\n}\n\nfunction normalizeClaim(c: string): string {\n // Lowercase, collapse whitespace, strip trailing punctuation. Goal:\n // \"Leaf X failed install\" and \"Leaf X failed install.\" hash the same.\n return c\n .toLowerCase()\n .replace(/\\s+/g, ' ')\n .replace(/[.!?;:,]+$/g, '')\n .trim()\n}\n\n/**\n * Convenience factory: produce a fully-formed AnalystFinding with the\n * id computed automatically. Analyst code stays terse.\n */\nexport function makeFinding(\n init: Omit<AnalystFinding, 'schema_version' | 'finding_id' | 'produced_at'> & {\n id_basis?: string\n produced_at?: string\n },\n): AnalystFinding {\n const { id_basis, produced_at, ...rest } = init\n return {\n schema_version: '1.0.0',\n finding_id: computeFindingId({\n analyst_id: rest.analyst_id,\n area: rest.area,\n subject: rest.subject,\n claim: rest.claim,\n id_basis,\n }),\n produced_at: produced_at ?? new Date().toISOString(),\n ...rest,\n }\n}\n\n/** Build a finding whose source is explicitly allowed during candidate generation. */\nexport function makeProposalFinding(\n init: Omit<ProposalFinding, 'schema_version' | 'finding_id' | 'produced_at'> & {\n id_basis?: string\n produced_at?: string\n },\n): ProposalFinding {\n const { proposal_origin, ...finding } = init\n return { ...makeFinding(finding), proposal_origin }\n}\n\n// ── Registry result envelope ────────────────────────────────────────\n\nexport interface AnalystRunSummary {\n analyst_id: string\n status: 'ok' | 'skipped' | 'failed'\n /** Why skipped — missing input, budget exceeded, capability unmet. */\n reason?: string\n findings_count: number\n latency_ms: number\n /** Additive model usage and cost provenance for this analyst. */\n usage: AnalystUsageReceipt\n /** When `status='failed'`: the error class + message, never the full stack. */\n error?: { class: string; message: string }\n}\n\nexport interface AnalystRunResult {\n run_id: string\n correlation_id: string\n started_at: string\n ended_at: string\n findings: AnalystFinding[]\n per_analyst: AnalystRunSummary[]\n /** Total LLM cost in USD across all analysts in this registry.run(). */\n total_cost_usd: number\n /**\n * Provenance for `total_cost_usd`. When uncaptured, the numeric field is only\n * the known subtotal and must not be treated as the run's total spend.\n */\n total_cost_provenance?: CostProvenance\n}\n\n// ── Streaming event envelope ────────────────────────────────────────\n\n/**\n * Events emitted by `AnalystRegistry.runStream(...)` in real time as\n * the registry executes. UIs subscribe via `for await (const ev of\n * registry.runStream(...))`; `registry.run(...)` is a thin collector\n * over the same stream, so the two surfaces share their invariants.\n *\n * Per-finding events are intentionally omitted — analyzers are batch\n * operations (a recursive engine returns the full `findings:json[]` at the\n * end of the responder), so streaming inside one analyst would only\n * emit partial JSON consumers can't render. The kind-completion event\n * is the right granularity; subscribers wanting per-finding rendering\n * iterate `event.findings` themselves.\n */\nexport type AnalystRunEvent =\n | {\n type: 'run-started'\n run_id: string\n correlation_id: string\n started_at: string\n /** The ordered list of analyst ids the registry will run. */\n analyst_ids: ReadonlyArray<string>\n }\n | {\n type: 'analyst-skipped'\n summary: AnalystRunSummary\n }\n | {\n type: 'analyst-started'\n analyst_id: string\n started_at: string\n }\n | {\n type: 'analyst-completed'\n /** `summary.status` is `'ok'` for clean completion or `'failed'` for thrown analysts. */\n summary: AnalystRunSummary\n findings: ReadonlyArray<AnalystFinding>\n }\n | {\n type: 'run-completed'\n result: AnalystRunResult\n }\n"],"mappings":";;;;;;;;;;;;;;;;;;;;AAsEA,MAAa,qBAAqB;CAAC;CAAY;CAAQ;CAAU;CAAO;AAAM;;;;;;;;;AAiO9E,SAAgB,iBAAiB,OAOtB;CACT,MAAM,QAAQ,KAAK,UAAU;EAC3B,GAAG,MAAM;EACT,GAAG,MAAM;EACT,GAAG,MAAM,WAAW;EACpB,GAAG,eAAe,MAAM,YAAY,MAAM,KAAK;CACjD,CAAC;CACD,OAAO,KAAK,WAAW,QAAQ,CAAC,CAAC,OAAO,KAAK,CAAC,CAAC,OAAO,KAAK,CAAC,CAAC,MAAM,GAAG,EAAE;AAC1E;AAEA,SAAS,eAAe,GAAmB;CAGzC,OAAO,EACJ,YAAY,CAAC,CACb,QAAQ,QAAQ,GAAG,CAAC,CACpB,QAAQ,eAAe,EAAE,CAAC,CAC1B,KAAK;AACV;;;;;AAMA,SAAgB,YACd,MAIgB;CAChB,MAAM,EAAE,UAAU,aAAa,GAAG,SAAS;CAC3C,OAAO;EACL,gBAAgB;EAChB,YAAY,iBAAiB;GAC3B,YAAY,KAAK;GACjB,MAAM,KAAK;GACX,SAAS,KAAK;GACd,OAAO,KAAK;GACZ;EACF,CAAC;EACD,aAAa,gCAAe,IAAI,KAAK,EAAA,CAAE,YAAY;EACnD,GAAG;CACL;AACF;;AAGA,SAAgB,oBACd,MAIiB;CACjB,MAAM,EAAE,iBAAiB,GAAG,YAAY;CACxC,OAAO;EAAE,GAAG,YAAY,OAAO;EAAG;CAAgB;AACpD"}
|
|
@@ -1,5 +1,5 @@
|
|
|
1
|
-
import {
|
|
2
|
-
import { R as Scenario, d as DispatchContext } from "./types-
|
|
1
|
+
import { c as RunSplitTag } from "./run-record-DF-fuu31.js";
|
|
2
|
+
import { R as Scenario, d as DispatchContext } from "./types-BvmWNjjs.js";
|
|
3
3
|
//#region src/benchmarks/types.d.ts
|
|
4
4
|
type BenchmarkTaskKind = 'retrieval' | 'rag-answer' | 'hallucination' | 'kb-improvement' | 'routing' | 'custom';
|
|
5
5
|
type BenchmarkFamily = 'beir' | 'mteb-retrieval' | 'msmarco' | 'trec-dl' | 'miracl' | 'lotte' | 'bright' | 'crag' | 'hotpotqa' | 'kilt' | 'ragtruth' | 'faithbench' | 'first-party' | 'custom';
|
|
@@ -99,4 +99,4 @@ declare const BENCHMARK_SPLIT_SEED = "agent-eval-v1";
|
|
|
99
99
|
declare function deterministicSplit(itemId: string, seed?: string): RunSplitTag;
|
|
100
100
|
//#endregion
|
|
101
101
|
export { BenchmarkFamily as a, BenchmarkSource as c, BenchmarkEvaluation as i, BenchmarkTaskKind as l, BenchmarkAdapter as n, BenchmarkResponder as o, BenchmarkDatasetItem as r, BenchmarkScenario as s, BENCHMARK_SPLIT_SEED as t, deterministicSplit as u };
|
|
102
|
-
//# sourceMappingURL=types
|
|
102
|
+
//# sourceMappingURL=types-Cvqy7asY.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types
|
|
1
|
+
{"version":3,"file":"types-Cvqy7asY.d.ts","names":[],"sources":["../src/benchmarks/types.ts"],"mappings":";;;KASY;KAQA;UAgBK,qBAAqB;;;EAGpC;;EAEA,SAAS;;EAET,QAAQ;;EAER,SAAS;;EAET,WAAW;;EAEX;;EAEA,SAAS;EACT,WAAW;;UAGI;;;;EAIf;;EAEA;;EAEA,aAAa;;;EAGb,KAAK;EACL;;UAGe;EACf;EACA;EACA;;EAEA;EACA;EACA;;EAEA;;;;;EAKA;;;UAQe,iBAAiB,kBAAkB,oBAAoB;;EAEtE;EACA,SAAS;EACT,WAAW;EACX;EACA,SAAS;EACT;;;;;EAKA,YAAY,OAAO,cAAc,QAAQ,qBAAqB;;EAE9D,SAAS,MAAM,qBAAqB,WAAW,UAAU,YAAY,QAAQ;;;;EAI7E,YAAY,iBAAiB;;UAGd,kBAAkB,4BAA4B;EAC7D;EACA;EACA,QAAQ;EACR,UAAU;EACV,UAAU;EACV,MAAM,qBAAqB;;KAGjB,mBAAmB,oBAAoB,uBAAuB;EACxE,UAAU,kBAAkB;EAC5B,MAAM,qBAAqB;EAC3B,SAAS;MACL,QAAQ,aAAa;;;cAoBd;;;;;;;;;iBAUG,mBACd,gBACA,gBACC"}
|
|
@@ -1,5 +1,5 @@
|
|
|
1
1
|
import { t as SeriesDistribution } from "./descriptive-B2iPaT9J.js";
|
|
2
|
-
import { m as RolloutLine } from "./schema-
|
|
2
|
+
import { m as RolloutLine } from "./schema-B5LtYvn9.js";
|
|
3
3
|
//#region src/supervisor-run/types.d.ts
|
|
4
4
|
/** A metric that could not be computed, with the reason its artifact was missing. */
|
|
5
5
|
interface Unavailable {
|
|
@@ -546,4 +546,4 @@ interface SupervisorRunTreeGap {
|
|
|
546
546
|
}
|
|
547
547
|
//#endregion
|
|
548
548
|
export { WallDistribution as A, SupervisorRunSources as C, SupervisorStatusSource as D, SupervisorRunTreeGapCode as E, showMeasured as F, unavailable as I, WorkerNativeSession as M, WorkerTurnCoverage as N, TerminalFailure as O, isUnavailable as P, SupervisorRunRollup as S, SupervisorRunTreeGap as T, SpendMeasurements as _, NamedResourceReceipt as a, SupervisorRunReader as b, PatchStats as c, RoleSpend as d, RollupCellRow as f, SpendMeasurement as g, SourceLimits as h, NO_SOURCE_LIMITS as i, WorkerLogSource as j, Unavailable as k, PerWorkerRow as l, SUPERVISOR_RUN_SCHEMA as m, EconomicsMetrics as n, OrchestrationMetrics as o, SUPERVISOR_RUN_ROLLUP_SCHEMA as p, Measured as r, OutcomeMetrics as s, DecisionMetrics as t, ResourceSpendRecord as u, SteerBreakdown as v, SupervisorRunTree as w, SupervisorRunReport as x, SupervisorRunNodeRole as y };
|
|
549
|
-
//# sourceMappingURL=types-
|
|
549
|
+
//# sourceMappingURL=types-DRNlLbBR.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types-
|
|
1
|
+
{"version":3,"file":"types-DRNlLbBR.d.ts","names":[],"sources":["../src/supervisor-run/types.ts"],"mappings":";;;;UAkCiB;WACN;;;KAIC,SAAS,KAAK,IAAI;iBAEd,YAAY,iBAAiB;iBAI7B,cAAc,aAAa,KAAK;;iBAKhC,aAAa,GAAG;;KAWpB;;UAGK;WACN;WACA;;;KAIC;WACG;WAA8B;;WAC9B;WAAgC;;;;;;UAM9B;;;;;WAKN;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;;;;;WAMA,QAAQ;;;;;;WAMR,gBAAgB;;WAEhB;;WAEA;WACA;WACA;WACA;;;;;;;;;;;;;UAcM;;WAEN;;WAEA;;WAEA;;WAEA;;WAEA;;;cAIE,kBAAkB;;;;;;;;;;UAiBd;;WAEN;WACA;;WAEA;;WAEA;;;;;;WAMA;;;;;;WAMA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA,kBAAkB;;WAElB;;;;WAIA;;;;;;WAMA;;;;;;;WAOA;;WAEA;;WAEA;;WAEA;;;;;WAKA;IACP;IACA;IACA;IACA;;IAEA;IACA;;WAEO;;WAEA,QAAQ;;;;;WAKR;;;;;WAKA;;;;;;UAOM;;WAEN;EACT,QAAQ,QAAQ;;cAOL;cACA;UAEI;;WAEN;WACA;;WAEA;;WAEA;;UAGM;WACN,gBAAgB;WAChB,gBAAgB;WAChB,kBAAkB;;WAElB,QAAQ;WACR,iBAAiB;WACjB,gBAAgB,kBAAkB;;WAElC,kBAAkB;;;;;;WAMlB,OAAO;WACP,WAAW;WACX,gBAAgB;;WAEhB,UAAU;;WAEV,gBAAgB;;WAEhB,iBAAiB;WACjB,oBAAoB;WACpB,kBAAkB;;;;;;;;;WASlB,sBAAsB;;WAEtB,QAAQ;WACR,SAAS;;WAET,mBAAmB;;UAGb;WACN,iBAAiB,SAAS;WAC1B,iBAAiB,SAAS;;;;;;;WAO1B,UAAU;;WAEV,UAAU;;WAEV,WAAW;;WAEX,oBAAoB;;WAEpB,wBAAwB;;WAExB,eAAe;WACf,qBAAqB;;UAGf;WACN,UAAU;WACV,WAAW;;;;;;WAMX,WAAW;WACX,YAAY;WACZ,KAAK;WACL;;UAGM;;WAEN;WACA;;WAEA,MAAM;;WAEN;;WAEA;;WAEA;;WAEA;;WAEA;WACA;;WAEA;WACA;WACA;WACA;WACA;;WAEA;;;;;;;KAQC,mBAAmB;;UAGd;WACN,KAAK;;WAEL;;;;;;;WAOA;;WAEA;;WAEA;;;;;;;;;;;;;UAcM;WACN,gBAAgB;WAChB,aAAa;;;UAIP;WACN,MAAM;WACN,MAAM;WACN,QAAQ;WACR,OAAO;;;UAID;WACN;WACA;WACA;WACA,WAAW,kBAAkB;;UAGvB;;WAEN,kBAAkB,kBAAkB;;WAEpC,OAAO;;;;;;;;WAQP,kBAAkB;;WAElB,SAAS;;WAET,OAAO;;;;;;;;WAQP,UAAU;;;;;;WAMV;WACA,yBAAyB;WACzB,0BAA0B,SAAS;WACnC,WAAW,kBAAkB;;UAGvB;WACN;WACA;WACA;WACA;;;KAIC;;;;;;UAOK;;;;;WAKN;;WAEA;;WAEA;;WAEA;;;;;;;;WAQA;;UAGM;;;;;;WAMN,WAAW;WACX,iBAAiB,SAAS;;;;;;WAM1B,WAAW;;;;;;;WAOX,SAAS,SAAS;WAClB,YAAY;WACZ,WAAW;WACX,eAAe;WACf,YAAY;WACZ,aAAa;WACb,YAAY;WACZ,YAAY;WACZ,UAAU;WACV,OAAO,SAAS;;WAEhB;;UAGM;WACN,eAAe;;WAEf;WACA;WACA;WACA,cAAc;WACd,yBAAyB;WACzB;WACA,eAAe;WACf,UAAU;WACV,WAAW;WACX,SAAS;;WAET;;WAEA;;UAGM;;WAEN,kBAAkB,kBAAkB;WACpC;WACA;WACA,QAAQ;WACR,OAAO;WACP,aAAa;WACb,SAAS;WACT,UAAU;WACV,KAAK;;UAGC;WACN,eAAe;WACf;WACA,aAAa;WACb,iBAAiB;WACjB;WACA,WAAW;WACX,mBAAmB;WACnB,iBAAiB;WACjB,aAAa;WACb,qBAAqB;WACrB,eAAe;;;;;WAKf,UAAU;;;;;;;WAOV;aACE;eAA2B,OAAO;eAA2B;;aAC7D;eAAwB,OAAO;eAA2B;;;WAE5D,eAAe;WACf,kBAAkB;;;;;;;;UASZ;WACN;WACA,gBAAgB;;WAEhB,eAAe;;;KAId;UASK;WACN,MAAM;WACN;WACA;WACA"}
|
package/dist/wire/index.d.ts
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import { b as CustomTokenPricing, c as CostLedgerHandle } from "../cost-ledger-DbQdN3nO.js";
|
|
2
|
-
import { x as ChatClient } from "../types-
|
|
3
|
-
import { s as TraceStore } from "../store-
|
|
4
|
-
import { m as FeedbackTrajectoryStore } from "../feedback-trajectory-
|
|
2
|
+
import { x as ChatClient } from "../types-C0bWoSKC.js";
|
|
3
|
+
import { s as TraceStore } from "../store-C4avFjIi.js";
|
|
4
|
+
import { m as FeedbackTrajectoryStore } from "../feedback-trajectory-B3ZSkfAq.js";
|
|
5
5
|
import { z } from "zod";
|
|
6
6
|
import { ServerType } from "@hono/node-server";
|
|
7
7
|
import { Hono } from "hono";
|
package/dist/wire/index.js
CHANGED
|
@@ -1,2 +1,2 @@
|
|
|
1
|
-
import { A as WIRE_VERSION, C as RUBRIC_VERSION_SCHEME, D as TracesIngestRequestSchema, E as TraceEventSchema, O as TracesIngestResponseSchema, S as ListRubricsResponseSchema, T as RubricSchema, _ as FeedbackIngestResponseSchema, a as runRpcBatch, b as JudgeRequestSchema, c as WireError, d as handleListRubrics, f as handleTracesIngest, g as ErrorResponseSchema, h as listBuiltinRubrics, i as dispatchRpc, j as hashRubric, k as VersionResponseSchema, l as handleFeedbackIngest, m as getBuiltinRubric, n as startServer, o as runRpcOnce, p as handleVersion, r as startServerAsync, s as buildOpenApi, t as createApp, u as handleJudge, v as FeedbackTrajectorySchema, w as RubricDimensionSchema, x as JudgeResultSchema, y as HealthResponseSchema } from "../server-
|
|
1
|
+
import { A as WIRE_VERSION, C as RUBRIC_VERSION_SCHEME, D as TracesIngestRequestSchema, E as TraceEventSchema, O as TracesIngestResponseSchema, S as ListRubricsResponseSchema, T as RubricSchema, _ as FeedbackIngestResponseSchema, a as runRpcBatch, b as JudgeRequestSchema, c as WireError, d as handleListRubrics, f as handleTracesIngest, g as ErrorResponseSchema, h as listBuiltinRubrics, i as dispatchRpc, j as hashRubric, k as VersionResponseSchema, l as handleFeedbackIngest, m as getBuiltinRubric, n as startServer, o as runRpcOnce, p as handleVersion, r as startServerAsync, s as buildOpenApi, t as createApp, u as handleJudge, v as FeedbackTrajectorySchema, w as RubricDimensionSchema, x as JudgeResultSchema, y as HealthResponseSchema } from "../server-C8XvqCg8.js";
|
|
2
2
|
export { ErrorResponseSchema, FeedbackIngestResponseSchema, FeedbackTrajectorySchema, HealthResponseSchema, JudgeRequestSchema, JudgeResultSchema, ListRubricsResponseSchema, RUBRIC_VERSION_SCHEME, RubricDimensionSchema, RubricSchema, TraceEventSchema, TracesIngestRequestSchema, TracesIngestResponseSchema, VersionResponseSchema, WIRE_VERSION, WireError, buildOpenApi, createApp, dispatchRpc, getBuiltinRubric, handleFeedbackIngest, handleJudge, handleListRubrics, handleTracesIngest, handleVersion, hashRubric, listBuiltinRubrics, runRpcBatch, runRpcOnce, startServer, startServerAsync };
|
|
@@ -24,7 +24,7 @@ const { runs, diagnostics, metrics, observations } = fromCodexSession({
|
|
|
24
24
|
|
|
25
25
|
1. **Read the transcript.** `parseCodeAgentJsonl(text)` for a string you already hold, `parseCodeAgentJsonlFile(path)` for a file. Both return `{ entries, malformedLines }`; a line that is not JSON is counted, never dropped silently.
|
|
26
26
|
2. **Convert it.** One function per harness, below. Each returns `runs`, `diagnostics`, `metrics`, and `observations`.
|
|
27
|
-
3. **Read `diagnostics` before `runs`.** One `CodeAgentSessionDiagnostic` per session records what the transcript actually carried: `malformedLines`, `hasExplicitTerminalSignal`, `hasFinalOutput`, `hasQualityLabel`, `hasTokenUsage`, `hasCost`, the `costKind` that was used, and any `warnings`. A run whose diagnostic says `hasCost: false` has no cost, not a cost of zero — check the flag before you aggregate the field.
|
|
27
|
+
3. **Read `diagnostics` before `runs`.** One `CodeAgentSessionDiagnostic` per session records what the transcript actually carried: `malformedLines`, `hasExplicitTerminalSignal`, `hasFinalOutput`, `hasQualityLabel`, `hasTokenUsage`, `hasCost`, the `costKind` that was used, and any `warnings`. A run whose diagnostic says `hasCost: false` has no known total cost, not a cost of zero — check the flag before you aggregate the field. A `lower-bound` cost states only a floor, in `costProvenance.knownLowerBoundUsd`.
|
|
28
28
|
|
|
29
29
|
## One function per harness
|
|
30
30
|
|
package/docs/diagnosis.md
CHANGED
|
@@ -20,9 +20,8 @@ Only an internal context may carry a `topology` (a run graph); a customer contex
|
|
|
20
20
|
|
|
21
21
|
## Order of operations
|
|
22
22
|
|
|
23
|
-
1. The
|
|
24
|
-
|
|
25
|
-
A match becomes `[REDACTED:<rule>]`, and the document's `coverage.redaction` counts matches by rule.
|
|
23
|
+
1. The redaction core ([redaction.md](./redaction.md), default profile) runs on span attributes, names, status messages, the focus question, and the topology.
|
|
24
|
+
A string that holds a credential becomes `[REDACTED:<detector>]`, and the document's `coverage.redaction` counts replacements by detector.
|
|
26
25
|
2. The deterministic pass reports the trace contract's capability table with the trace's own reasons, execution facts with token and cost accounting, and `observed` findings.
|
|
27
26
|
Each observed finding carries a measure with its denominator and cites the spans it counted.
|
|
28
27
|
3. Model mode reads the runs with the most errors through the prime protocol (`runPrimeExchange`).
|
package/docs/eval-surface-map.md
CHANGED
|
@@ -74,7 +74,7 @@ Use the decision diagnostics and intervals to distinguish insufficient evidence
|
|
|
74
74
|
|
|
75
75
|
| Subpath | Use |
|
|
76
76
|
|---|---|
|
|
77
|
-
| `/traces`, `/trace-attributes` | Store trace evidence, [connect observability exporters](./adapters-observability.md), and use canonical measurement attribute names. |
|
|
77
|
+
| `/traces`, `/trace-attributes` | Store trace evidence, [redact it and check it is safe to share](./redaction.md), [connect observability exporters](./adapters-observability.md), and use canonical measurement attribute names. |
|
|
78
78
|
| `/analyst` | Execute declared analysts against recorded evidence. |
|
|
79
79
|
| `/reporting`, `/pipelines` | Compare runs, render [research reports](./research-report-methodology.md), and extract recorded failure patterns. |
|
|
80
80
|
| `/supervisor-run` | Read recursive run directories and their evidence coverage. |
|
package/docs/experiment.md
CHANGED
|
@@ -269,7 +269,6 @@ The statistical machinery underneath is re-exported from its existing homes; thi
|
|
|
269
269
|
| `paretoSignificanceGate`, `buildEvidenceVector` | `src/campaign/gates/promotion-policy.ts` |
|
|
270
270
|
| `pairArms`, `comparePairedArms`, `pairRunRecords` | `src/paired-arms.ts` |
|
|
271
271
|
| `hashJson`, `manifestContentDigest`, `signManifest`, `verifyManifest`, `HypothesisManifest` | `src/pre-registration.ts` |
|
|
272
|
-
| `ExperimentTracker` (run ledger with KEEP/ITERATE/NOISE/REGRESSION) | `src/experiment-tracker.ts` |
|
|
273
272
|
|
|
274
273
|
`HypothesisManifest` stays as the lightweight single-metric registration; `sealExperiment` is the full-design registration.
|
|
275
274
|
The trace-repair admission machinery (`buildDenominatorChain`, oracle determinism, control policy) keeps its repair vocabulary in `./trace-repair`; this module is the general form new experiments should register against.
|
package/docs/insight-report.md
CHANGED
|
@@ -103,8 +103,9 @@ Store it as `RunRecord.outcome.judgeScores` alongside the relevant search or hol
|
|
|
103
103
|
Different judge means can reflect different coverage, scales, or criteria.
|
|
104
104
|
Compare shared cases before attributing a difference to miscalibration.
|
|
105
105
|
|
|
106
|
-
`costQuality.provenance` separates observed USD, estimated USD, and uncaptured costs.
|
|
107
|
-
|
|
106
|
+
`costQuality.provenance` separates observed USD, estimated USD, lower-bound floors, and uncaptured costs.
|
|
107
|
+
Lower-bound and uncaptured rows are excluded from the cost distribution and Pareto calculation.
|
|
108
|
+
`lowerBound.floorUsd` sums the proven floors of runs whose total is unknown; it is never a total.
|
|
108
109
|
Read `knownFraction` and `costQuality.degraded` before comparing costs.
|
|
109
110
|
A frontier only compares the observed candidate points; it does not identify the best possible system.
|
|
110
111
|
|