@tangle-network/agent-eval 0.150.1 → 0.161.0
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 +178 -1
- package/README.md +7 -3
- package/dist/{active-curriculum-C4mk67HP.js → active-curriculum-CD5TU2yW.js} +3 -13
- package/dist/active-curriculum-CD5TU2yW.js.map +1 -0
- package/dist/{agent-profile-cell-BkcRDikH.d.ts → agent-profile-cell-CTOZJUuE.d.ts} +4 -2
- package/dist/agent-profile-cell-CTOZJUuE.d.ts.map +1 -0
- package/dist/analyst/index.d.ts +19 -36
- package/dist/analyst/index.d.ts.map +1 -1
- package/dist/analyst/index.js +8 -8
- package/dist/analyst/index.js.map +1 -1
- package/dist/{backend-integrity-DOCa_QrR.d.ts → backend-integrity-DxuQCu_A.d.ts} +4 -3
- package/dist/backend-integrity-DxuQCu_A.d.ts.map +1 -0
- package/dist/{benchmark-BtAWA8nT.d.ts → benchmark-CGPp-kDC.d.ts} +3 -3
- package/dist/{benchmark-BtAWA8nT.d.ts.map → benchmark-CGPp-kDC.d.ts.map} +1 -1
- package/dist/{benchmark-command-CAFwbH0L.js → benchmark-command-BVtaq_ve.js} +26 -31
- package/dist/benchmark-command-BVtaq_ve.js.map +1 -0
- package/dist/benchmarks/index.d.ts +6 -19
- package/dist/benchmarks/index.d.ts.map +1 -1
- package/dist/benchmarks/index.js +4 -4
- package/dist/benchmarks/index.js.map +1 -1
- package/dist/builder-eval/index.d.ts +3 -3
- package/dist/builder-eval/index.js +3 -3
- package/dist/campaign/index.d.ts +9 -9
- package/dist/campaign/index.js +7 -7
- package/dist/{campaign-CN_7xJdV.js → campaign-BSmOwskD.js} +77 -795
- package/dist/campaign-BSmOwskD.js.map +1 -0
- package/dist/{canonical-D-XsTQ6_.js → canonical-IL-Bu-14.js} +26 -2
- package/dist/canonical-IL-Bu-14.js.map +1 -0
- package/dist/{capture-fetch-BBVFzhkk.d.ts → capture-fetch-CqwsJkkG.d.ts} +3 -3
- package/dist/{capture-fetch-BBVFzhkk.d.ts.map → capture-fetch-CqwsJkkG.d.ts.map} +1 -1
- package/dist/{chat-client-2bVfrzhN.js → chat-client-DlMlAeYI.js} +5 -55
- package/dist/{chat-client-2bVfrzhN.js.map → chat-client-DlMlAeYI.js.map} +1 -1
- package/dist/chat-json-call-6g5sJobJ.js +53 -0
- package/dist/chat-json-call-6g5sJobJ.js.map +1 -0
- package/dist/cli.js +54 -19
- package/dist/cli.js.map +1 -1
- package/dist/{client-LIuo-KPv.js → client-CX7KqIdB.js} +3 -3
- package/dist/client-CX7KqIdB.js.map +1 -0
- package/dist/{client-kPQYT_56.d.ts → client-L9VVPkim.d.ts} +4 -4
- package/dist/{client-kPQYT_56.d.ts.map → client-L9VVPkim.d.ts.map} +1 -1
- package/dist/contract/index.d.ts +13 -27
- package/dist/contract/index.d.ts.map +1 -1
- package/dist/contract/index.js +14 -17
- package/dist/contract/index.js.map +1 -1
- package/dist/{counterfactual--bpysZF0.d.ts → counterfactual-BaFUWK3H.d.ts} +4 -4
- package/dist/{counterfactual--bpysZF0.d.ts.map → counterfactual-BaFUWK3H.d.ts.map} +1 -1
- package/dist/{counterfactual-lDfCx0Uz.js → counterfactual-D_VWavVm.js} +2 -2
- package/dist/{counterfactual-lDfCx0Uz.js.map → counterfactual-D_VWavVm.js.map} +1 -1
- package/dist/{dataset-CJjKqQfA.d.ts → dataset-DQqhOCPt.d.ts} +5 -4
- package/dist/{dataset-CJjKqQfA.d.ts.map → dataset-DQqhOCPt.d.ts.map} +1 -1
- package/dist/{default-registry-Cw0Ohdoj.d.ts → default-registry-G9CKMNkc.d.ts} +7 -8
- package/dist/{default-registry-Cw0Ohdoj.d.ts.map → default-registry-G9CKMNkc.d.ts.map} +1 -1
- package/dist/{define-agent-eval-CEQWL9Hy.d.ts → define-agent-eval-Dx1JnPEa.d.ts} +26 -6
- package/dist/define-agent-eval-Dx1JnPEa.d.ts.map +1 -0
- package/dist/{define-agent-eval-rqNyVhVV.js → define-agent-eval-h-s-sI-v.js} +17 -11
- package/dist/define-agent-eval-h-s-sI-v.js.map +1 -0
- package/dist/{descriptive-B5MwKfbf.js → descriptive-jDOuI6mz.js} +22 -2
- package/dist/descriptive-jDOuI6mz.js.map +1 -0
- package/dist/{dspy-rlm-engine-DHI0WrUU.js → dspy-rlm-engine-DptEII26.js} +95 -15
- package/dist/dspy-rlm-engine-DptEII26.js.map +1 -0
- package/dist/{emitter-CPBAhxum.js → emitter-BpYFQPj4.js} +2 -18
- package/dist/emitter-BpYFQPj4.js.map +1 -0
- package/dist/{emitter-DGQGoLyj.d.ts → emitter-D_jYSGRd.d.ts} +4 -20
- package/dist/{emitter-DGQGoLyj.d.ts.map → emitter-D_jYSGRd.d.ts.map} +1 -1
- package/dist/{engine-BLzhNzoY.d.ts → engine-Cu5qD5Fc.d.ts} +9 -11
- package/dist/{engine-BLzhNzoY.d.ts.map → engine-Cu5qD5Fc.d.ts.map} +1 -1
- package/dist/{eval-campaign-C4jmuM-b.js → eval-campaign-BsXWL2-2.js} +17 -28
- package/dist/eval-campaign-BsXWL2-2.js.map +1 -0
- package/dist/{exact-types-ccQAyut1.d.ts → exact-types-qnexxJ1Z.d.ts} +2 -2
- package/dist/{exact-types-ccQAyut1.d.ts.map → exact-types-qnexxJ1Z.d.ts.map} +1 -1
- package/dist/{exec-y-DCLqK7.js → exec-D9WpA2p-.js} +2 -2
- package/dist/exec-D9WpA2p-.js.map +1 -0
- package/dist/experiment/index.d.ts +45 -11
- package/dist/experiment/index.d.ts.map +1 -1
- package/dist/experiment/index.js +30 -11
- package/dist/experiment/index.js.map +1 -1
- package/dist/{experiment-tracker-0MhuPArU.d.ts → experiment-tracker-DCO6Cz4s.d.ts} +2 -2
- package/dist/{experiment-tracker-0MhuPArU.d.ts.map → experiment-tracker-DCO6Cz4s.d.ts.map} +1 -1
- package/dist/{exporters-q9iL-2Jf.js → exporters-Df7TgHFv.js} +3 -3
- package/dist/exporters-Df7TgHFv.js.map +1 -0
- package/dist/{external-optimizer-contracts-DbLsm4Po.d.ts → external-optimizer-contracts-szBJ_1vh.d.ts} +2 -2
- package/dist/{external-optimizer-contracts-DbLsm4Po.d.ts.map → external-optimizer-contracts-szBJ_1vh.d.ts.map} +1 -1
- package/dist/{external-optimizer-process-Bhmzngf-.js → external-optimizer-process-WosTBChy.js} +4 -4
- package/dist/{external-optimizer-process-Bhmzngf-.js.map → external-optimizer-process-WosTBChy.js.map} +1 -1
- package/dist/{external-optimizer-subprocess-Dn90UqN2.js → external-optimizer-subprocess-BIWbHpgD.js} +10 -6
- package/dist/external-optimizer-subprocess-BIWbHpgD.js.map +1 -0
- package/dist/{failure-cluster-BLURuWG4.d.ts → failure-cluster-CXL8NbEw.d.ts} +3 -3
- package/dist/{failure-cluster-BLURuWG4.d.ts.map → failure-cluster-CXL8NbEw.d.ts.map} +1 -1
- package/dist/{feedback-trajectory-DpTTjo0q.d.ts → feedback-trajectory-B3ZHaHV_.d.ts} +7 -7
- package/dist/{feedback-trajectory-DpTTjo0q.d.ts.map → feedback-trajectory-B3ZHaHV_.d.ts.map} +1 -1
- package/dist/fuzz.js +1 -1
- package/dist/{hf-dataset-XggBupCr.js → hf-dataset-D8_RNIis.js} +4 -4
- package/dist/{hf-dataset-XggBupCr.js.map → hf-dataset-D8_RNIis.js.map} +1 -1
- package/dist/hosted/index.d.ts +3 -14
- package/dist/hosted/index.d.ts.map +1 -1
- package/dist/hosted/index.js +2 -2
- package/dist/{index-CTKpu9ry.d.ts → index-CGtH1piv.d.ts} +48 -26
- package/dist/index-CGtH1piv.d.ts.map +1 -0
- package/dist/{index-BNPtkBPf.d.ts → index-D-IiQIBB.d.ts} +5 -10
- package/dist/index-D-IiQIBB.d.ts.map +1 -0
- package/dist/{index-IQccV3Ou.d.ts → index-D-V8gCs_.d.ts} +13 -90
- package/dist/index-D-V8gCs_.d.ts.map +1 -0
- package/dist/{index-B8Ui1mr1.d.ts → index-lfaSeKSD.d.ts} +18 -2
- package/dist/index-lfaSeKSD.d.ts.map +1 -0
- package/dist/index-vrJugRal.d.ts +1 -0
- package/dist/index.d.ts +68 -125
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +61 -112
- package/dist/index.js.map +1 -1
- package/dist/{insight-report-BeT8KCgI.d.ts → insight-report-DRe8LB6d.d.ts} +4 -4
- package/dist/{insight-report-BeT8KCgI.d.ts.map → insight-report-DRe8LB6d.d.ts.map} +1 -1
- package/dist/{integrity-DysDBWDu.js → integrity-CyWSSoQS.js} +17 -4
- package/dist/integrity-CyWSSoQS.js.map +1 -0
- package/dist/{integrity-B0dZ96EO.d.ts → integrity-DUNX9Fao.d.ts} +3 -3
- package/dist/{integrity-B0dZ96EO.d.ts.map → integrity-DUNX9Fao.d.ts.map} +1 -1
- package/dist/{judge-calibration-DZkWrm5H.js → judge-calibration-zZjLz8hr.js} +2 -2
- package/dist/{judge-calibration-DZkWrm5H.js.map → judge-calibration-zZjLz8hr.js.map} +1 -1
- package/dist/{kind-factory-DmAa0h3K.js → kind-factory-DY8FdoXf.js} +3 -71
- package/dist/kind-factory-DY8FdoXf.js.map +1 -0
- package/dist/ledger-core/index.d.ts +2 -2
- package/dist/ledger-core/index.js +3 -3
- package/dist/{ledger-core-DTae9rv_.js → ledger-core-BOzlRygb.js} +2 -2
- package/dist/{ledger-core-DTae9rv_.js.map → ledger-core-BOzlRygb.js.map} +1 -1
- package/dist/{llm-client-Bg32RW0j.js → llm-client-hgDieDNN.js} +53 -98
- package/dist/llm-client-hgDieDNN.js.map +1 -0
- package/dist/{llm-judge-CVq33oz1.js → llm-judge-BhasIPFT.js} +1136 -78
- package/dist/llm-judge-BhasIPFT.js.map +1 -0
- package/dist/{matrix-DrVnRp4G.d.ts → matrix-eXKRMHnL.d.ts} +74 -72
- package/dist/matrix-eXKRMHnL.d.ts.map +1 -0
- package/dist/meta-eval/index.d.ts +8 -6
- package/dist/meta-eval/index.d.ts.map +1 -1
- package/dist/meta-eval/index.js +9 -7
- package/dist/meta-eval/index.js.map +1 -1
- package/dist/{mint-BV6tLVWl.js → mint-DfODW1KW.js} +3 -3
- package/dist/{mint-BV6tLVWl.js.map → mint-DfODW1KW.js.map} +1 -1
- package/dist/multishot/golden/index.d.ts +2 -8
- package/dist/multishot/golden/index.d.ts.map +1 -1
- package/dist/multishot/golden/index.js +56 -86
- package/dist/multishot/golden/index.js.map +1 -1
- package/dist/multishot/index.d.ts +11 -46
- package/dist/multishot/index.d.ts.map +1 -1
- package/dist/multishot/index.js +30 -83
- package/dist/multishot/index.js.map +1 -1
- package/dist/openapi.json +1 -1
- package/dist/{opencode-sqlite-DJWAXLms.js → opencode-sqlite-eK6HW6dr.js} +2 -6
- package/dist/{opencode-sqlite-DJWAXLms.js.map → opencode-sqlite-eK6HW6dr.js.map} +1 -1
- package/dist/pipelines/index.d.ts +5 -5
- package/dist/pipelines/index.js +3 -3
- package/dist/{pre-registration-zFSLEiFU.d.ts → pre-registration-CzFCcwYk.d.ts} +55 -40
- package/dist/pre-registration-CzFCcwYk.d.ts.map +1 -0
- package/dist/pre-registration-KN9jkh58.js +110 -0
- package/dist/pre-registration-KN9jkh58.js.map +1 -0
- package/dist/{produced-state-jfk8Du3b.js → produced-state-DZ89riy5.js} +8 -8
- package/dist/produced-state-DZ89riy5.js.map +1 -0
- package/dist/profile-cell.d.ts +1 -1
- package/dist/profile-cell.js +31 -5
- package/dist/profile-cell.js.map +1 -1
- package/dist/{promotion-policy-DLOUkYhI.d.ts → promotion-policy-DtnOIZvk.d.ts} +2 -2
- package/dist/{promotion-policy-DLOUkYhI.d.ts.map → promotion-policy-DtnOIZvk.d.ts.map} +1 -1
- package/dist/{query-Di7eEQ79.js → query-CHmMP42p.js} +20 -11
- package/dist/query-CHmMP42p.js.map +1 -0
- package/dist/{query-CJ_DX8vl.d.ts → query-DxPYqpmT.d.ts} +10 -4
- package/dist/query-DxPYqpmT.d.ts.map +1 -0
- package/dist/raw-provider-sink-BU29Sh8h.d.ts +134 -0
- package/dist/raw-provider-sink-BU29Sh8h.d.ts.map +1 -0
- package/dist/{registry-BQwrSYpC.d.ts → registry-8You7OK1.d.ts} +5 -7
- package/dist/{registry-BQwrSYpC.d.ts.map → registry-8You7OK1.d.ts.map} +1 -1
- package/dist/{release-confidence-BknrpBnO.js → release-confidence-DKfD2RYU.js} +28 -14
- package/dist/release-confidence-DKfD2RYU.js.map +1 -0
- package/dist/{release-confidence-4XrqlpFD.d.ts → release-confidence-Dqt0NFep.d.ts} +7 -6
- package/dist/release-confidence-Dqt0NFep.d.ts.map +1 -0
- package/dist/reporting.d.ts +3 -3
- package/dist/reporting.js +3 -3
- package/dist/{researcher-DJnoUE8c.d.ts → researcher-Cz565b7D.d.ts} +34 -21
- package/dist/researcher-Cz565b7D.d.ts.map +1 -0
- package/dist/{reward-hacking-62tojkQd.d.ts → reward-hacking-MBf7qpSB.d.ts} +2 -2
- package/dist/{reward-hacking-62tojkQd.d.ts.map → reward-hacking-MBf7qpSB.d.ts.map} +1 -1
- package/dist/{reward-hacking-DKI9T52l.js → reward-hacking-t4lB1yt8.js} +3 -3
- package/dist/{reward-hacking-DKI9T52l.js.map → reward-hacking-t4lB1yt8.js.map} +1 -1
- package/dist/rl.d.ts +11 -42
- package/dist/rl.d.ts.map +1 -1
- package/dist/rl.js +41 -24
- package/dist/rl.js.map +1 -1
- package/dist/rollout/index.d.ts +3 -3
- package/dist/rollout/index.js +7 -7
- package/dist/{rollout-ytVQ7WT8.js → rollout-Dm2tSdiQ.js} +6 -6
- package/dist/{rollout-ytVQ7WT8.js.map → rollout-Dm2tSdiQ.js.map} +1 -1
- package/dist/{rubric-predictive-validity-Cwwyd7ah.js → rubric-predictive-validity-CK8SCOg-.js} +6 -16
- package/dist/rubric-predictive-validity-CK8SCOg-.js.map +1 -0
- package/dist/{rubric-predictive-validity-C7LnNvF2.d.ts → rubric-predictive-validity-CxycqzX5.d.ts} +4 -3
- package/dist/rubric-predictive-validity-CxycqzX5.d.ts.map +1 -0
- package/dist/{run-record-D2lDdSAz.js → run-record-BC0ebuRP.js} +2 -2
- package/dist/{run-record-D2lDdSAz.js.map → run-record-BC0ebuRP.js.map} +1 -1
- package/dist/{run-record-DVV82Gwh.d.ts → run-record-VVy4T9OW.d.ts} +3 -3
- package/dist/{run-record-DVV82Gwh.d.ts.map → run-record-VVy4T9OW.d.ts.map} +1 -1
- package/dist/{schema-BtVldJ3T.d.ts → schema-Bjgdsn73.d.ts} +2 -4
- package/dist/{schema-BtVldJ3T.d.ts.map → schema-Bjgdsn73.d.ts.map} +1 -1
- package/dist/{schema-Cef2cFmb.d.ts → schema-BzWDXhOR.d.ts} +2 -5
- package/dist/schema-BzWDXhOR.d.ts.map +1 -0
- package/dist/{schema-C6DW4ZHR.js → schema-C1aaAxTf.js} +2 -2
- package/dist/schema-C1aaAxTf.js.map +1 -0
- package/dist/{schema-CRhEY1SO.js → schema-k6ZBftVv.js} +2 -8
- package/dist/{schema-CRhEY1SO.js.map → schema-k6ZBftVv.js.map} +1 -1
- package/dist/{semantic-concept-judge-laMCnTLn.js → semantic-concept-judge-BSkKKHeq.js} +14 -38
- package/dist/semantic-concept-judge-BSkKKHeq.js.map +1 -0
- package/dist/{sequential-eprocess-CbUt2htw.js → sequential-eprocess-D1jKoihe.js} +49 -2
- package/dist/sequential-eprocess-D1jKoihe.js.map +1 -0
- package/dist/{sequential-C458DXNf.js → sequential-rYW-Ophm.js} +41 -16
- package/dist/sequential-rYW-Ophm.js.map +1 -0
- package/dist/{series-convergence-BxKEgBwA.d.ts → series-convergence-D9WgpXGi.d.ts} +2 -2
- package/dist/{series-convergence-BxKEgBwA.d.ts.map → series-convergence-D9WgpXGi.d.ts.map} +1 -1
- package/dist/{server-dIWwF3j_.js → server-BtFd4uzB.js} +19 -42
- package/dist/server-BtFd4uzB.js.map +1 -0
- package/dist/{skillopt-optimization-method-DLeUcK-K.js → skillopt-optimization-method-DbaekMcn.js} +794 -8
- package/dist/skillopt-optimization-method-DbaekMcn.js.map +1 -0
- package/dist/{skillopt-optimization-method-BDD_o1xE.d.ts → skillopt-optimization-method-x7TTF23P.d.ts} +20 -7
- package/dist/{skillopt-optimization-method-BDD_o1xE.d.ts.map → skillopt-optimization-method-x7TTF23P.d.ts.map} +1 -1
- package/dist/{statistical-heldout-_woZ9q9j.d.ts → statistical-heldout-Cy3EhjlC.d.ts} +21 -8
- package/dist/statistical-heldout-Cy3EhjlC.d.ts.map +1 -0
- package/dist/{steps-AmkT-GIM.d.ts → steps-CiNVJry_.d.ts} +2 -17
- package/dist/steps-CiNVJry_.d.ts.map +1 -0
- package/dist/{store-CT9YIIve.d.ts → store-B06JdC56.d.ts} +2 -2
- package/dist/{store-CT9YIIve.d.ts.map → store-B06JdC56.d.ts.map} +1 -1
- package/dist/{store-otlp-CDYWW_8N.js → store-otlp-C_Rq5I4D.js} +2 -2
- package/dist/{store-otlp-CDYWW_8N.js.map → store-otlp-C_Rq5I4D.js.map} +1 -1
- package/dist/{store-tool-spans-Br2_IUhm.d.ts → store-tool-spans-DPUG7UUY.d.ts} +6 -6
- package/dist/{store-tool-spans-Br2_IUhm.d.ts.map → store-tool-spans-DPUG7UUY.d.ts.map} +1 -1
- package/dist/{store-tool-spans-CykkbOlv.js → store-tool-spans-Dlh9vkFK.js} +3 -3
- package/dist/{store-tool-spans-CykkbOlv.js.map → store-tool-spans-Dlh9vkFK.js.map} +1 -1
- package/dist/storyboard/index.d.ts +1 -1
- package/dist/{summary-report-Blysd6Z2.js → summary-report-BI5hUtvK.js} +7 -17
- package/dist/summary-report-BI5hUtvK.js.map +1 -0
- package/dist/{summary-report-B__Y5ub3.d.ts → summary-report-CC07PhEL.d.ts} +6 -5
- package/dist/summary-report-CC07PhEL.d.ts.map +1 -0
- package/dist/supervisor-run/index.d.ts +27 -18
- package/dist/supervisor-run/index.d.ts.map +1 -1
- package/dist/supervisor-run/index.js +104 -26
- package/dist/supervisor-run/index.js.map +1 -1
- package/dist/{task-failure-attributes--ZTP3tYO.js → task-failure-attributes-DTl-7-Kw.js} +3 -3
- package/dist/{task-failure-attributes--ZTP3tYO.js.map → task-failure-attributes-DTl-7-Kw.js.map} +1 -1
- package/dist/{tool-groups-B4tqh8jB.d.ts → tool-groups-Ci8i9ErB.d.ts} +3 -3
- package/dist/tool-groups-Ci8i9ErB.d.ts.map +1 -0
- package/dist/{tool-waste-BDdBZG1F.js → tool-waste-BqzmVdJk.js} +4 -4
- package/dist/{tool-waste-BDdBZG1F.js.map → tool-waste-BqzmVdJk.js.map} +1 -1
- package/dist/{tool-waste-DjRDEsuI.d.ts → tool-waste-Dro0gJi3.d.ts} +4 -4
- package/dist/{tool-waste-DjRDEsuI.d.ts.map → tool-waste-Dro0gJi3.d.ts.map} +1 -1
- package/dist/trace-repair/index.d.ts +4 -77
- package/dist/trace-repair/index.d.ts.map +1 -1
- package/dist/trace-repair/index.js +5 -15
- package/dist/trace-repair/index.js.map +1 -1
- package/dist/traces.d.ts +13 -23
- package/dist/traces.d.ts.map +1 -1
- package/dist/traces.js +9 -20
- package/dist/traces.js.map +1 -1
- package/dist/{trajectory-YC15QDYQ.d.ts → trajectory-Bi157Gun.d.ts} +3 -3
- package/dist/{trajectory-YC15QDYQ.d.ts.map → trajectory-Bi157Gun.d.ts.map} +1 -1
- package/dist/trajectory-replay/index.d.ts +5 -5
- package/dist/trajectory-replay/index.js +5 -5
- package/dist/{provenance-oA4-zUqm.d.ts → transient-failure-DKF5Mofa.d.ts} +468 -13
- package/dist/transient-failure-DKF5Mofa.d.ts.map +1 -0
- package/dist/types-B3jzCp0p.js.map +1 -1
- package/dist/{types-CLAwnY-L.d.ts → types-BPb2Kf_C2.d.ts} +3 -3
- package/dist/types-BPb2Kf_C2.d.ts.map +1 -0
- package/dist/types-Bfk0uxRj.d.ts +443 -0
- package/dist/types-Bfk0uxRj.d.ts.map +1 -0
- package/dist/{types-DdFNuyxQ.d.ts → types-D4s7Z6nq.d.ts} +30 -6
- package/dist/types-D4s7Z6nq.d.ts.map +1 -0
- package/dist/{types-B2NsbrNy.d.ts → types-D9ssmxKL.d.ts} +3 -3
- package/dist/{types-B2NsbrNy.d.ts.map → types-D9ssmxKL.d.ts.map} +1 -1
- package/dist/{types-yLK8gXE9.d.ts → types-DeIUdzNd.d.ts} +160 -10
- package/dist/types-DeIUdzNd.d.ts.map +1 -0
- package/dist/{verdict-BndeTAh_.js → verdict-B0xltqu6.js} +2 -2
- package/dist/{verdict-BndeTAh_.js.map → verdict-B0xltqu6.js.map} +1 -1
- package/dist/verdict-cache-CdVVTVmn.js +88 -0
- package/dist/verdict-cache-CdVVTVmn.js.map +1 -0
- package/dist/wire/index.d.ts +21 -111
- package/dist/wire/index.d.ts.map +1 -1
- package/dist/wire/index.js +2 -2
- package/docs/adapters-observability.md +9 -23
- package/docs/building-doctrine.md +3 -3
- package/docs/campaign-proposers.md +54 -15
- package/docs/concepts.md +3 -4
- package/docs/design/statistics-decisions.md +89 -1
- package/docs/eval-surface-map.md +14 -0
- package/docs/experiment.md +19 -2
- package/docs/feedback-trajectories.md +1 -1
- package/docs/multishot-golden-records.md +4 -4
- package/docs/public-api.md +1616 -0
- package/docs/research-report-methodology.md +1 -1
- package/docs/search-history-receipts.md +39 -1
- package/docs/trace-analysis.md +1 -1
- package/docs/trace-repair-admission.md +1 -1
- package/docs/trace-repair-continuation.md +1 -1
- package/docs/verdicts.md +24 -0
- package/docs/wire-protocol.md +1 -1
- package/package.json +6 -2
- package/dist/active-curriculum-C4mk67HP.js.map +0 -1
- package/dist/agent-profile-cell-BkcRDikH.d.ts.map +0 -1
- package/dist/backend-integrity-DOCa_QrR.d.ts.map +0 -1
- package/dist/benchmark-command-CAFwbH0L.js.map +0 -1
- package/dist/campaign-CN_7xJdV.js.map +0 -1
- package/dist/canonical-D-XsTQ6_.js.map +0 -1
- package/dist/client-LIuo-KPv.js.map +0 -1
- package/dist/define-agent-eval-CEQWL9Hy.d.ts.map +0 -1
- package/dist/define-agent-eval-rqNyVhVV.js.map +0 -1
- package/dist/descriptive-B5MwKfbf.js.map +0 -1
- package/dist/dspy-rlm-engine-DHI0WrUU.js.map +0 -1
- package/dist/emitter-CPBAhxum.js.map +0 -1
- package/dist/eval-campaign-C4jmuM-b.js.map +0 -1
- package/dist/exec-y-DCLqK7.js.map +0 -1
- package/dist/exporters-q9iL-2Jf.js.map +0 -1
- package/dist/external-optimizer-subprocess-Dn90UqN2.js.map +0 -1
- package/dist/index-B8Ui1mr1.d.ts.map +0 -1
- package/dist/index-BNPtkBPf.d.ts.map +0 -1
- package/dist/index-C1ravkGA.d.ts +0 -1
- package/dist/index-CTKpu9ry.d.ts.map +0 -1
- package/dist/index-IQccV3Ou.d.ts.map +0 -1
- package/dist/integrity-DysDBWDu.js.map +0 -1
- package/dist/kind-factory-DmAa0h3K.js.map +0 -1
- package/dist/llm-client-Bg32RW0j.js.map +0 -1
- package/dist/llm-judge-CVq33oz1.js.map +0 -1
- package/dist/matrix-DrVnRp4G.d.ts.map +0 -1
- package/dist/pre-registration-DakwTRXk.js +0 -96
- package/dist/pre-registration-DakwTRXk.js.map +0 -1
- package/dist/pre-registration-zFSLEiFU.d.ts.map +0 -1
- package/dist/produced-state-jfk8Du3b.js.map +0 -1
- package/dist/provenance-oA4-zUqm.d.ts.map +0 -1
- package/dist/query-CJ_DX8vl.d.ts.map +0 -1
- package/dist/query-Di7eEQ79.js.map +0 -1
- package/dist/release-confidence-4XrqlpFD.d.ts.map +0 -1
- package/dist/release-confidence-BknrpBnO.js.map +0 -1
- package/dist/researcher-DJnoUE8c.d.ts.map +0 -1
- package/dist/rubric-predictive-validity-C7LnNvF2.d.ts.map +0 -1
- package/dist/rubric-predictive-validity-Cwwyd7ah.js.map +0 -1
- package/dist/schema-C6DW4ZHR.js.map +0 -1
- package/dist/schema-Cef2cFmb.d.ts.map +0 -1
- package/dist/semantic-concept-judge-laMCnTLn.js.map +0 -1
- package/dist/sequential-C458DXNf.js.map +0 -1
- package/dist/sequential-eprocess-CbUt2htw.js.map +0 -1
- package/dist/server-dIWwF3j_.js.map +0 -1
- package/dist/skillopt-optimization-method-DLeUcK-K.js.map +0 -1
- package/dist/statistical-heldout-_woZ9q9j.d.ts.map +0 -1
- package/dist/steps-AmkT-GIM.d.ts.map +0 -1
- package/dist/summary-report-B__Y5ub3.d.ts.map +0 -1
- package/dist/summary-report-Blysd6Z2.js.map +0 -1
- package/dist/tool-groups-B4tqh8jB.d.ts.map +0 -1
- package/dist/types-CLAwnY-L.d.ts.map +0 -1
- package/dist/types-DdFNuyxQ.d.ts.map +0 -1
- package/dist/types-jUBXJ7Iz.d.ts +0 -884
- package/dist/types-jUBXJ7Iz.d.ts.map +0 -1
- package/dist/types-yLK8gXE9.d.ts.map +0 -1
- package/dist/verdict-cache-mZf5FEiY.js +0 -107
- package/dist/verdict-cache-mZf5FEiY.js.map +0 -1
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"types-jUBXJ7Iz.d.ts","names":[],"sources":["../src/trace/raw-provider-sink.ts","../src/judge-families.ts","../src/integrity/served-model.ts","../src/llm-client.ts","../src/analyst/chat-client.ts","../src/types.ts"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;KA8BY;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;;;;;;iBAmB5C,wBAAwB,OAAO,mBAAmB;UA8CjD;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;;;;;;;;;;;;;KC5QxB;;;;;;;iBAkEI,YAAY,kBAAkB;UAc7B;;EAEf;;;;EAIA;;cAGW,yBAAyB;WAGlB,UAAU;WACV;EAHlB,YACE,iBACgB,UAAU,eACV;;;;;;;;;;;;;;;iBAoBJ,kBACd,kBACA,OAAM,2BACL;;;;KCnFS;;;;;;;;;;;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;;;UCnQc;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;;EAEA,WAAW;;EAEX;;;;;;iBAOc,2BACd,SAAS,KACP,0GAGF,UAAS,mBACR;UAgCc;EACf;EACA;EACA;;EAEA;;EAEA;;EAEA;;UAGe;;EAEf;;EAEA,YAAY;EACZ,OAAO;;;;;EAKP;;;;;;;;;EASA;;;;;;;;;;;;EAYA;;EAEA;;;;;;;;;EASA;;;;;;;EAOA;;EAEA,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;;EAEf;;EAEA;EACA;;EAEA;IAAe;IAAc;;;EAE7B;;EAEA;;;;;;;EAOA,SAAS;;;;;;;;EAQT;;EAEA;;EAEA,qBAAqB;;;;;;;EAOrB;;;;;;EAMA;;EAEA,WAAW;;EAEX,eAAe;;;;;;;;EAQf,UAAU;;;;;EAKV;;EAEA;IAAiB;IAAgB;;;EAEjC,WAAW;;;;;;;;;;EAUX,8BAA8B;;;;;;;;;;;;;iBAyEhB,oBAAoB;;;;;;iBAqMpB,gBAAgB;;;;;;iBA6EV,QACpB,KAAK,gBACL,OAAM,mBACL,QAAQ;;;;;;;iBAyVW,YAAY,aAChC,KAAK,gBACL,OAAM,mBACL;EAAU,OAAO;EAAG,QAAQ;;UA6Ed;;;;;;;EAOf;;;;;EAKA,kBAAkB,eAAe;;EAEjC,kBAAkB,eAAe;;EAEjC;;;;;EAKA;;;;;;;;;;;iBAYc,eAAe,MAAM,kBAAkB,MAAK;;;;;;;;;;;;;;;;;;iBA6EtC,SACpB,eACA,OAAM;EAAqB;IAC1B;EACD;EACA;EACA;;EAEA;;EAEA;;;;;;;cAoCW;WACF;mBACQ;EAEjB,YAAY,OAAM;EAKlB,KAAK,KAAK,gBAAgB,MAAM,mBAAmB,QAAQ;EAK3D,SAAS,aACP,KAAK,gBACL,MAAM,mBACL;IAAU,OAAO;IAAG,QAAQ;;;;;;;;UC7wChB;;WAEN,WAAW;;WAEX;;WAEA;;EAGT,KAAK,KAAK,aAAa,OAAO,eAAe,QAAQ;;KAG3C;UAQK,oBAAoB,KAAK;;EAExC;;KAGU,eAAe;UAEV;;EAEf,SAAS;;EAET;;EAEA;;EAEA;;KAKU,uBACR,sBACA,yBACA,8BACA,0BACA,sBACA;UAEM;EACR;;EAEA;;UAGe,4BAA4B;EAC3C;EACA;EACA;;UAGe,+BAA+B;EAC9C;EACA;EACA;;UAGe,oCAAoC;EACnD;EACA;EACA;;;;;;UAOe,gCAAgC;EAC/C;EACA,OAAO,KAAK,aAAa,OAAO,iBAAiB,QAAQ;;;UAI1C,4BAA4B;EAC3C;EACA,OAAO,KAAK,aAAa,OAAO,iBAAiB,QAAQ;;;;;;UAO1C,0BAA0B;EACzC;EACA,UAAU,KAAK,aAAa,OAAO,iBAAiB,QAAQ;;;;;;iBAO9C,iBAAiB,MAAM,uBAAuB;;;UCjH7C;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;;UAmBe;EACf;EACA;EACA;EACA;EACA;IAAmB;IAAc;;EACjC;EACA;;UASe;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;;UAyBe;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 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"types-yLK8gXE9.d.ts","names":[],"sources":["../src/supervisor-run/types.ts"],"mappings":";;;UAiCiB;WACN;;;KAIC,SAAS,KAAK,IAAI;iBAEd,YAAY,iBAAiB;iBAI7B,cAAc,aAAa,KAAK;;iBAKhC,aAAa,GAAG;;KAWpB;;;;;UAMK;;;;;WAKN;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA;WACA;WACA;WACA;;;;;;;;;;;;;UAcM;;WAEN;;WAEA;;WAEA;;WAEA;;WAEA;;;cAIE,kBAAkB;;;;;;;;;;UAiBd;;WAEN;WACA;;WAEA;;WAEA;;;;;;WAMA;;WAEA;;WAEA;;WAEA;;WAEA;;WAEA,kBAAkB;;WAElB;;WAEA;;;;;;;WAOA;;WAEA;;WAEA;;WAEA;;;;;WAKA;IACP;IACA;IACA;IACA;;IAEA;IACA;;WAEO;;WAEA,QAAQ;;;;;;WAMR;;;;;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;;WAElB,QAAQ;WACR,SAAS;;WAET,mBAAmB;;UAGb;WACN,iBAAiB,SAAS;WAC1B,iBAAiB,SAAS;;WAE1B,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;;UAGM;WACN;WACA;WACA;WACA;WACA;WACA;;UAGM;;WAEN,OAAO;;;;;;;;WAQP,kBAAkB;;WAElB,SAAS;WACT,UAAU;;;;;;WAMV;WACA,yBAAyB;WACzB,0BAA0B,SAAS;WACnC,WAAW,kBAAkB;;UAGvB;WACN;WACA;WACA;WACA;;UAGM;WACN,WAAW;WACX,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;WACN;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;WACf,UAAU;WACV,eAAe;WACf,kBAAkB;;;;;;;;UASZ;WACN;WACA,gBAAgB;;WAEhB,eAAe;;;KAId;UASK;WACN,MAAM;WACN;WACA;WACA"}
|
|
@@ -1,107 +0,0 @@
|
|
|
1
|
-
import { createHash } from "node:crypto";
|
|
2
|
-
import { appendFileSync, existsSync, readFileSync } from "node:fs";
|
|
3
|
-
//#region src/verdict-cache.ts
|
|
4
|
-
/**
|
|
5
|
-
* Content-addressed judge-verdict caching.
|
|
6
|
-
*
|
|
7
|
-
* LAW: cache JUDGE VERDICTS only — judging the same artifact with the same
|
|
8
|
-
* judge+rubric is pure. NEVER cache agent rollouts. (A router that cached
|
|
9
|
-
* identical fanout prompts silently destroyed best-of-N diversity; rollout
|
|
10
|
-
* caching reintroduces that failure class. Judging has no diversity to
|
|
11
|
-
* destroy — same artifact + same rubric ⇒ same verdict is the desired
|
|
12
|
-
* property, not a bug.)
|
|
13
|
-
*
|
|
14
|
-
* The cache key is a sha-256 over the canonical JSON of everything that can
|
|
15
|
-
* change a verdict: the artifact content, the scenario id, the judge name,
|
|
16
|
-
* the full dimension list (key + description — the description IS the rubric
|
|
17
|
-
* text shown to the judge), and a caller-supplied `judgeVersion`.
|
|
18
|
-
* `judgeVersion` is REQUIRED: a judge whose prompt/model/ensemble changes
|
|
19
|
-
* without a version bump would otherwise silently serve stale verdicts.
|
|
20
|
-
*
|
|
21
|
-
* Strict canonicalization (`canonicalJson`) throws on undefined / function /
|
|
22
|
-
* symbol / non-finite numbers — an artifact that cannot be unambiguously
|
|
23
|
-
* serialized cannot be content-addressed, and coercing it would let two
|
|
24
|
-
* different artifacts collide on one key.
|
|
25
|
-
*/
|
|
26
|
-
function canonicalizeAt(value, path) {
|
|
27
|
-
if (value === null) return "null";
|
|
28
|
-
switch (typeof value) {
|
|
29
|
-
case "boolean": return value ? "true" : "false";
|
|
30
|
-
case "number":
|
|
31
|
-
if (!Number.isFinite(value)) throw new Error(`canonicalJson: non-finite number (${value}) at ${path} — ambiguity is an error, not a coercion`);
|
|
32
|
-
return JSON.stringify(value);
|
|
33
|
-
case "string": return JSON.stringify(value);
|
|
34
|
-
case "undefined":
|
|
35
|
-
case "function":
|
|
36
|
-
case "symbol": throw new Error(`canonicalJson: ${typeof value} at ${path} — ambiguity is an error, not a coercion`);
|
|
37
|
-
case "bigint": throw new Error(`canonicalJson: bigint at ${path} — not representable in JSON`);
|
|
38
|
-
case "object": break;
|
|
39
|
-
}
|
|
40
|
-
const obj = value;
|
|
41
|
-
if (typeof obj.toJSON === "function") return canonicalizeAt(obj.toJSON(), path);
|
|
42
|
-
if (Array.isArray(obj)) return `[${obj.map((item, i) => canonicalizeAt(item, `${path}[${i}]`)).join(",")}]`;
|
|
43
|
-
if (obj instanceof Map || obj instanceof Set) throw new Error(`canonicalJson: ${obj instanceof Map ? "Map" : "Set"} at ${path} — would serialize as '{}'; convert to a plain object/array first`);
|
|
44
|
-
return `{${Object.keys(obj).sort().map((k) => `${JSON.stringify(k)}:${canonicalizeAt(obj[k], `${path}.${k}`)}`).join(",")}}`;
|
|
45
|
-
}
|
|
46
|
-
/**
|
|
47
|
-
* Stable JSON stringify: object keys sorted recursively, so two semantically
|
|
48
|
-
* equal values produce byte-identical output regardless of key insertion
|
|
49
|
-
* order. Throws on undefined / function / symbol / NaN / ±Infinity / bigint /
|
|
50
|
-
* Map / Set — anything JSON.stringify would coerce or drop silently.
|
|
51
|
-
*
|
|
52
|
-
* Distinct from `pre-registration.ts`'s `canonicalize`/`hashJson`, which are
|
|
53
|
-
* permissive (coercion allowed) and async (web-crypto). Use THIS pair when a
|
|
54
|
-
* hash collision or silent coercion would corrupt a cache key or attestation.
|
|
55
|
-
*/
|
|
56
|
-
function canonicalJson(value) {
|
|
57
|
-
return canonicalizeAt(value, "$");
|
|
58
|
-
}
|
|
59
|
-
/** Hex sha-256 over `canonicalJson(value)`. The content address used by the
|
|
60
|
-
* verdict cache and report attestation. */
|
|
61
|
-
function contentHash(value) {
|
|
62
|
-
return createHash("sha256").update(canonicalJson(value)).digest("hex");
|
|
63
|
-
}
|
|
64
|
-
function parseCacheLine(line, path, lineNo) {
|
|
65
|
-
let parsed;
|
|
66
|
-
try {
|
|
67
|
-
parsed = JSON.parse(line);
|
|
68
|
-
} catch (err) {
|
|
69
|
-
throw new Error(`fileVerdictCache: corrupt JSONL at ${path}:${lineNo} — ${err instanceof Error ? err.message : String(err)}`);
|
|
70
|
-
}
|
|
71
|
-
const rec = parsed;
|
|
72
|
-
if (typeof rec !== "object" || rec === null || typeof rec.key !== "string" || typeof rec.score !== "object" || rec.score === null || typeof rec.score.composite !== "number" || typeof rec.score.dimensions !== "object") throw new Error(`fileVerdictCache: invalid record shape at ${path}:${lineNo} — expected {key, score:{dimensions, composite, notes}}`);
|
|
73
|
-
return rec;
|
|
74
|
-
}
|
|
75
|
-
/**
|
|
76
|
-
* JSONL-file-backed store: the full file is loaded into an in-memory index at
|
|
77
|
-
* construction; every `set` appends one line synchronously (durable before
|
|
78
|
-
* the verdict is returned). A corrupt or malformed line throws at load with
|
|
79
|
-
* file:line — a skipped line would silently re-judge (cost) or, worse, mask
|
|
80
|
-
* a half-written file that needs operator attention.
|
|
81
|
-
*/
|
|
82
|
-
function fileVerdictCache(path) {
|
|
83
|
-
const entries = /* @__PURE__ */ new Map();
|
|
84
|
-
if (existsSync(path)) {
|
|
85
|
-
const lines = readFileSync(path, "utf8").split("\n");
|
|
86
|
-
for (let i = 0; i < lines.length; i++) {
|
|
87
|
-
const line = lines[i];
|
|
88
|
-
if (line === void 0 || line.trim() === "") continue;
|
|
89
|
-
const rec = parseCacheLine(line, path, i + 1);
|
|
90
|
-
entries.set(rec.key, rec.score);
|
|
91
|
-
}
|
|
92
|
-
}
|
|
93
|
-
return {
|
|
94
|
-
get: (key) => entries.get(key),
|
|
95
|
-
set: (key, score) => {
|
|
96
|
-
appendFileSync(path, `${JSON.stringify({
|
|
97
|
-
key,
|
|
98
|
-
score
|
|
99
|
-
})}\n`, "utf8");
|
|
100
|
-
entries.set(key, score);
|
|
101
|
-
}
|
|
102
|
-
};
|
|
103
|
-
}
|
|
104
|
-
//#endregion
|
|
105
|
-
export { contentHash as n, fileVerdictCache as r, canonicalJson as t };
|
|
106
|
-
|
|
107
|
-
//# sourceMappingURL=verdict-cache-mZf5FEiY.js.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"verdict-cache-mZf5FEiY.js","names":[],"sources":["../src/verdict-cache.ts"],"sourcesContent":["/**\n * Content-addressed judge-verdict caching.\n *\n * LAW: cache JUDGE VERDICTS only — judging the same artifact with the same\n * judge+rubric is pure. NEVER cache agent rollouts. (A router that cached\n * identical fanout prompts silently destroyed best-of-N diversity; rollout\n * caching reintroduces that failure class. Judging has no diversity to\n * destroy — same artifact + same rubric ⇒ same verdict is the desired\n * property, not a bug.)\n *\n * The cache key is a sha-256 over the canonical JSON of everything that can\n * change a verdict: the artifact content, the scenario id, the judge name,\n * the full dimension list (key + description — the description IS the rubric\n * text shown to the judge), and a caller-supplied `judgeVersion`.\n * `judgeVersion` is REQUIRED: a judge whose prompt/model/ensemble changes\n * without a version bump would otherwise silently serve stale verdicts.\n *\n * Strict canonicalization (`canonicalJson`) throws on undefined / function /\n * symbol / non-finite numbers — an artifact that cannot be unambiguously\n * serialized cannot be content-addressed, and coercing it would let two\n * different artifacts collide on one key.\n */\n\nimport { createHash } from 'node:crypto'\nimport { appendFileSync, existsSync, readFileSync } from 'node:fs'\nimport type { JudgeConfig, JudgeScore, Scenario } from './campaign/types'\n\n// ── canonical JSON + content hash ─────────────────────────────────────────\n\nfunction canonicalizeAt(value: unknown, path: string): string {\n if (value === null) return 'null'\n switch (typeof value) {\n case 'boolean':\n return value ? 'true' : 'false'\n case 'number':\n if (!Number.isFinite(value)) {\n throw new Error(\n `canonicalJson: non-finite number (${value}) at ${path} — ambiguity is an error, not a coercion`,\n )\n }\n return JSON.stringify(value)\n case 'string':\n return JSON.stringify(value)\n case 'undefined':\n case 'function':\n case 'symbol':\n throw new Error(\n `canonicalJson: ${typeof value} at ${path} — ambiguity is an error, not a coercion`,\n )\n case 'bigint':\n throw new Error(`canonicalJson: bigint at ${path} — not representable in JSON`)\n case 'object':\n break\n }\n const obj = value as Record<string, unknown>\n // Honor toJSON (Date → ISO string) before structural checks — without it a\n // Date would canonicalize to '{}' and every timestamp would collide.\n if (typeof obj.toJSON === 'function') {\n return canonicalizeAt((obj as { toJSON(): unknown }).toJSON(), path)\n }\n if (Array.isArray(obj)) {\n return `[${obj.map((item, i) => canonicalizeAt(item, `${path}[${i}]`)).join(',')}]`\n }\n if (obj instanceof Map || obj instanceof Set) {\n throw new Error(\n `canonicalJson: ${obj instanceof Map ? 'Map' : 'Set'} at ${path} — would serialize as '{}'; convert to a plain object/array first`,\n )\n }\n const keys = Object.keys(obj).sort()\n const parts = keys.map((k) => `${JSON.stringify(k)}:${canonicalizeAt(obj[k], `${path}.${k}`)}`)\n return `{${parts.join(',')}}`\n}\n\n/**\n * Stable JSON stringify: object keys sorted recursively, so two semantically\n * equal values produce byte-identical output regardless of key insertion\n * order. Throws on undefined / function / symbol / NaN / ±Infinity / bigint /\n * Map / Set — anything JSON.stringify would coerce or drop silently.\n *\n * Distinct from `pre-registration.ts`'s `canonicalize`/`hashJson`, which are\n * permissive (coercion allowed) and async (web-crypto). Use THIS pair when a\n * hash collision or silent coercion would corrupt a cache key or attestation.\n */\nexport function canonicalJson(value: unknown): string {\n return canonicalizeAt(value, '$')\n}\n\n/** Hex sha-256 over `canonicalJson(value)`. The content address used by the\n * verdict cache and report attestation. */\nexport function contentHash(value: unknown): string {\n return createHash('sha256').update(canonicalJson(value)).digest('hex')\n}\n\n// ── store contract ─────────────────────────────────────────────────────────\n\n/** Pluggable verdict store. Sync or async on both legs — `cachedJudge`\n * awaits the results either way. */\nexport interface VerdictCacheStore {\n get(key: string): Promise<JudgeScore | undefined> | JudgeScore | undefined\n set(key: string, score: JudgeScore): Promise<void> | void\n}\n\n/** Process-local Map-backed store. */\nexport function inMemoryVerdictCache(): VerdictCacheStore {\n const entries = new Map<string, JudgeScore>()\n return {\n get: (key) => entries.get(key),\n set: (key, score) => {\n entries.set(key, score)\n },\n }\n}\n\ninterface VerdictCacheLine {\n key: string\n score: JudgeScore\n}\n\nfunction parseCacheLine(line: string, path: string, lineNo: number): VerdictCacheLine {\n let parsed: unknown\n try {\n parsed = JSON.parse(line)\n } catch (err) {\n throw new Error(\n `fileVerdictCache: corrupt JSONL at ${path}:${lineNo} — ${err instanceof Error ? err.message : String(err)}`,\n )\n }\n const rec = parsed as Partial<VerdictCacheLine>\n if (\n typeof rec !== 'object' ||\n rec === null ||\n typeof rec.key !== 'string' ||\n typeof rec.score !== 'object' ||\n rec.score === null ||\n typeof rec.score.composite !== 'number' ||\n typeof rec.score.dimensions !== 'object'\n ) {\n throw new Error(\n `fileVerdictCache: invalid record shape at ${path}:${lineNo} — expected {key, score:{dimensions, composite, notes}}`,\n )\n }\n return rec as VerdictCacheLine\n}\n\n/**\n * JSONL-file-backed store: the full file is loaded into an in-memory index at\n * construction; every `set` appends one line synchronously (durable before\n * the verdict is returned). A corrupt or malformed line throws at load with\n * file:line — a skipped line would silently re-judge (cost) or, worse, mask\n * a half-written file that needs operator attention.\n */\nexport function fileVerdictCache(path: string): VerdictCacheStore {\n const entries = new Map<string, JudgeScore>()\n if (existsSync(path)) {\n const lines = readFileSync(path, 'utf8').split('\\n')\n for (let i = 0; i < lines.length; i++) {\n const line = lines[i]\n if (line === undefined || line.trim() === '') continue\n const rec = parseCacheLine(line, path, i + 1)\n entries.set(rec.key, rec.score)\n }\n }\n return {\n get: (key) => entries.get(key),\n set: (key, score) => {\n appendFileSync(path, `${JSON.stringify({ key, score })}\\n`, 'utf8')\n entries.set(key, score)\n },\n }\n}\n\n// ── cached judge wrapper ───────────────────────────────────────────────────\n\nexport interface VerdictCacheStats {\n hits: number\n misses: number\n}\n\nexport interface CachedJudgeOptions {\n /** REQUIRED — part of the cache key. Bump on any change to the judge's\n * prompt, model, ensemble, or scoring logic; silent judge upgrades must\n * never serve stale verdicts. */\n judgeVersion: string\n}\n\n/** The wrapped judge: same `JudgeConfig` seam, plus hit/miss observability. */\nexport type CachedJudge<TArtifact, TScenario extends Scenario = Scenario> = JudgeConfig<\n TArtifact,\n TScenario\n> & {\n stats(): VerdictCacheStats\n}\n\n/**\n * Wrap a `JudgeConfig` so repeat judgments of the same artifact are served\n * from the store instead of re-invoking `score()`. The wrapper is generic\n * over the judge's own type parameters and preserves `appliesTo` — it is a\n * drop-in replacement anywhere a `JudgeConfig` is accepted.\n *\n * A judge that throws is NOT cached: the error propagates and the next\n * attempt re-judges (caching a failure would pin a transient outage forever).\n */\nexport function cachedJudge<TArtifact, TScenario extends Scenario = Scenario>(\n judge: JudgeConfig<TArtifact, TScenario>,\n store: VerdictCacheStore,\n options: CachedJudgeOptions,\n): CachedJudge<TArtifact, TScenario> {\n if (typeof options.judgeVersion !== 'string' || options.judgeVersion.trim() === '') {\n throw new Error('cachedJudge: judgeVersion is required and must be a non-empty string')\n }\n const stats: VerdictCacheStats = { hits: 0, misses: 0 }\n const wrapped: CachedJudge<TArtifact, TScenario> = {\n name: judge.name,\n dimensions: judge.dimensions,\n judgeVersion: options.judgeVersion,\n async score(input) {\n const key = contentHash({\n artifact: canonicalJson(input.artifact),\n scenarioId: input.scenario.id,\n judgeName: judge.name,\n dimensions: judge.dimensions,\n judgeVersion: options.judgeVersion,\n })\n const cached = await store.get(key)\n if (cached !== undefined) {\n stats.hits += 1\n return cached\n }\n const score = await judge.score(input)\n await store.set(key, score)\n stats.misses += 1\n return score\n },\n stats: () => ({ ...stats }),\n }\n if (judge.appliesTo) wrapped.appliesTo = judge.appliesTo\n return wrapped\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;AA6BA,SAAS,eAAe,OAAgB,MAAsB;CAC5D,IAAI,UAAU,MAAM,OAAO;CAC3B,QAAQ,OAAO,OAAf;EACE,KAAK,WACH,OAAO,QAAQ,SAAS;EAC1B,KAAK;GACH,IAAI,CAAC,OAAO,SAAS,KAAK,GACxB,MAAM,IAAI,MACR,qCAAqC,MAAM,OAAO,KAAK,yCACzD;GAEF,OAAO,KAAK,UAAU,KAAK;EAC7B,KAAK,UACH,OAAO,KAAK,UAAU,KAAK;EAC7B,KAAK;EACL,KAAK;EACL,KAAK,UACH,MAAM,IAAI,MACR,kBAAkB,OAAO,MAAM,MAAM,KAAK,yCAC5C;EACF,KAAK,UACH,MAAM,IAAI,MAAM,4BAA4B,KAAK,6BAA6B;EAChF,KAAK,UACH;CACJ;CACA,MAAM,MAAM;CAGZ,IAAI,OAAO,IAAI,WAAW,YACxB,OAAO,eAAgB,IAA8B,OAAO,GAAG,IAAI;CAErE,IAAI,MAAM,QAAQ,GAAG,GACnB,OAAO,IAAI,IAAI,KAAK,MAAM,MAAM,eAAe,MAAM,GAAG,KAAK,GAAG,EAAE,EAAE,CAAC,CAAC,CAAC,KAAK,GAAG,EAAE;CAEnF,IAAI,eAAe,OAAO,eAAe,KACvC,MAAM,IAAI,MACR,kBAAkB,eAAe,MAAM,QAAQ,MAAM,MAAM,KAAK,kEAClE;CAIF,OAAO,IAFM,OAAO,KAAK,GAAG,CAAC,CAAC,KACb,CAAC,CAAC,KAAK,MAAM,GAAG,KAAK,UAAU,CAAC,EAAE,GAAG,eAAe,IAAI,IAAI,GAAG,KAAK,GAAG,GAAG,GAC5E,CAAC,CAAC,KAAK,GAAG,EAAE;AAC7B;;;;;;;;;;;AAYA,SAAgB,cAAc,OAAwB;CACpD,OAAO,eAAe,OAAO,GAAG;AAClC;;;AAIA,SAAgB,YAAY,OAAwB;CAClD,OAAO,WAAW,QAAQ,CAAC,CAAC,OAAO,cAAc,KAAK,CAAC,CAAC,CAAC,OAAO,KAAK;AACvE;AA2BA,SAAS,eAAe,MAAc,MAAc,QAAkC;CACpF,IAAI;CACJ,IAAI;EACF,SAAS,KAAK,MAAM,IAAI;CAC1B,SAAS,KAAK;EACZ,MAAM,IAAI,MACR,sCAAsC,KAAK,GAAG,OAAO,KAAK,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG,GAC3G;CACF;CACA,MAAM,MAAM;CACZ,IACE,OAAO,QAAQ,YACf,QAAQ,QACR,OAAO,IAAI,QAAQ,YACnB,OAAO,IAAI,UAAU,YACrB,IAAI,UAAU,QACd,OAAO,IAAI,MAAM,cAAc,YAC/B,OAAO,IAAI,MAAM,eAAe,UAEhC,MAAM,IAAI,MACR,6CAA6C,KAAK,GAAG,OAAO,wDAC9D;CAEF,OAAO;AACT;;;;;;;;AASA,SAAgB,iBAAiB,MAAiC;CAChE,MAAM,0BAAU,IAAI,IAAwB;CAC5C,IAAI,WAAW,IAAI,GAAG;EACpB,MAAM,QAAQ,aAAa,MAAM,MAAM,CAAC,CAAC,MAAM,IAAI;EACnD,KAAK,IAAI,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;GACrC,MAAM,OAAO,MAAM;GACnB,IAAI,SAAS,KAAA,KAAa,KAAK,KAAK,MAAM,IAAI;GAC9C,MAAM,MAAM,eAAe,MAAM,MAAM,IAAI,CAAC;GAC5C,QAAQ,IAAI,IAAI,KAAK,IAAI,KAAK;EAChC;CACF;CACA,OAAO;EACL,MAAM,QAAQ,QAAQ,IAAI,GAAG;EAC7B,MAAM,KAAK,UAAU;GACnB,eAAe,MAAM,GAAG,KAAK,UAAU;IAAE;IAAK;GAAM,CAAC,EAAE,KAAK,MAAM;GAClE,QAAQ,IAAI,KAAK,KAAK;EACxB;CACF;AACF"}
|