@tangle-network/agent-eval 0.163.2 → 0.171.0
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 +127 -0
- package/README.md +2 -0
- package/dist/adapters/http.d.ts +108 -0
- package/dist/adapters/http.d.ts.map +1 -0
- package/dist/adapters/http.js +208 -0
- package/dist/adapters/http.js.map +1 -0
- package/dist/analyst/index.d.ts +40 -70
- package/dist/analyst/index.d.ts.map +1 -1
- package/dist/analyst/index.js +18 -311
- package/dist/analyst/index.js.map +1 -1
- package/dist/{backend-integrity-DxuQCu_A.d.ts → backend-integrity-e79K3UPD.d.ts} +3 -3
- package/dist/{backend-integrity-DxuQCu_A.d.ts.map → backend-integrity-e79K3UPD.d.ts.map} +1 -1
- package/dist/{benchmark-BhT16ep9.js → benchmark-C4wk_Sjr.js} +10 -3
- package/dist/benchmark-C4wk_Sjr.js.map +1 -0
- package/dist/{benchmark-command-CF-4GEWZ.js → benchmark-command-D8k3Gf0J.js} +236 -251
- package/dist/benchmark-command-D8k3Gf0J.js.map +1 -0
- package/dist/{benchmark-CGPp-kDC.d.ts → benchmark-h-h4bfqj.d.ts} +3 -3
- package/dist/{benchmark-CGPp-kDC.d.ts.map → benchmark-h-h4bfqj.d.ts.map} +1 -1
- package/dist/benchmarks/index.d.ts +5 -5
- 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 +8 -8
- package/dist/campaign/index.js +7 -7
- package/dist/{campaign-DQZmc2Dq.js → campaign-B72njjHj.js} +15 -14
- package/dist/campaign-B72njjHj.js.map +1 -0
- package/dist/{canonical-IL-Bu-14.js → canonical-DPyQ_rpt.js} +22 -2
- package/dist/{canonical-IL-Bu-14.js.map → canonical-DPyQ_rpt.js.map} +1 -1
- package/dist/{chat-client-DlMlAeYI.js → chat-client-DEtybj5i.js} +5 -5
- package/dist/{chat-client-DlMlAeYI.js.map → chat-client-DEtybj5i.js.map} +1 -1
- package/dist/cli.js +2 -2
- package/dist/{client-CX7KqIdB.js → client-BvwNkIRN.js} +2 -2
- package/dist/{client-CX7KqIdB.js.map → client-BvwNkIRN.js.map} +1 -1
- package/dist/{client-L9VVPkim.d.ts → client-CDtcZ3p9.d.ts} +4 -4
- package/dist/{client-L9VVPkim.d.ts.map → client-CDtcZ3p9.d.ts.map} +1 -1
- package/dist/contract/index.d.ts +12 -703
- package/dist/contract/index.js +11 -11
- package/dist/{counterfactual-BaFUWK3H.d.ts → counterfactual-Bee5_BIn.d.ts} +4 -4
- package/dist/{counterfactual-BaFUWK3H.d.ts.map → counterfactual-Bee5_BIn.d.ts.map} +1 -1
- package/dist/{default-registry-G9CKMNkc.d.ts → default-registry-B0s2zU-s.d.ts} +6 -6
- package/dist/{default-registry-G9CKMNkc.d.ts.map → default-registry-B0s2zU-s.d.ts.map} +1 -1
- package/dist/{define-agent-eval-Dx1JnPEa.d.ts → define-agent-eval-Cjy2yhqP.d.ts} +7 -7
- package/dist/{define-agent-eval-Dx1JnPEa.d.ts.map → define-agent-eval-Cjy2yhqP.d.ts.map} +1 -1
- package/dist/{define-agent-eval-D08pWIJb.js → define-agent-eval-Dy8QgxAI.js} +20 -8
- package/dist/{define-agent-eval-D08pWIJb.js.map → define-agent-eval-Dy8QgxAI.js.map} +1 -1
- package/dist/{dspy-rlm-engine-DhA9qKIm.js → dspy-rlm-engine-CS3qcCEk.js} +3 -10
- package/dist/dspy-rlm-engine-CS3qcCEk.js.map +1 -0
- package/dist/{emitter-D_jYSGRd.d.ts → emitter-Bvnu0VzL.d.ts} +3 -3
- package/dist/{emitter-D_jYSGRd.d.ts.map → emitter-Bvnu0VzL.d.ts.map} +1 -1
- package/dist/{engine-Cu5qD5Fc.d.ts → engine-CAmTUk52.d.ts} +7 -7
- package/dist/{engine-Cu5qD5Fc.d.ts.map → engine-CAmTUk52.d.ts.map} +1 -1
- package/dist/{eval-campaign-BfohKmzx.js → eval-campaign-JDTeE6Pl.js} +4 -4
- package/dist/{eval-campaign-BfohKmzx.js.map → eval-campaign-JDTeE6Pl.js.map} +1 -1
- package/dist/{exact-types-qnexxJ1Z.d.ts → exact-types-BEecmnWm.d.ts} +2 -2
- package/dist/{exact-types-qnexxJ1Z.d.ts.map → exact-types-BEecmnWm.d.ts.map} +1 -1
- package/dist/experiment/index.d.ts +5 -5
- package/dist/experiment/index.js +4 -4
- package/dist/{experiment-tracker-DCO6Cz4s.d.ts → experiment-tracker-Dm8yQMqb.d.ts} +2 -2
- package/dist/{experiment-tracker-DCO6Cz4s.d.ts.map → experiment-tracker-Dm8yQMqb.d.ts.map} +1 -1
- package/dist/{external-optimizer-process-BFmh36vW.js → external-optimizer-process-CQxylYeG.js} +4 -11
- package/dist/external-optimizer-process-CQxylYeG.js.map +1 -0
- package/dist/{external-optimizer-subprocess-CqLMW3nh.js → external-optimizer-subprocess-Cex8Da2i.js} +25 -11
- package/dist/external-optimizer-subprocess-Cex8Da2i.js.map +1 -0
- package/dist/{failure-cluster-CXL8NbEw.d.ts → failure-cluster-6YSvsKlp.d.ts} +3 -3
- package/dist/{failure-cluster-CXL8NbEw.d.ts.map → failure-cluster-6YSvsKlp.d.ts.map} +1 -1
- package/dist/{feedback-trajectory-B3ZHaHV_.d.ts → feedback-trajectory-DIqpCyF0.d.ts} +6 -6
- package/dist/{feedback-trajectory-B3ZHaHV_.d.ts.map → feedback-trajectory-DIqpCyF0.d.ts.map} +1 -1
- package/dist/fuzz.d.ts +1 -1
- package/dist/{skillopt-optimization-method-x7TTF23P.d.ts → heldout-gate-Dh2b62w8.d.ts} +111 -111
- package/dist/heldout-gate-Dh2b62w8.d.ts.map +1 -0
- package/dist/hosted/index.d.ts +2 -2
- package/dist/hosted/index.js +1 -1
- package/dist/{index-D-V8gCs_.d.ts → index-8VIogTyS.d.ts} +30 -22
- package/dist/index-8VIogTyS.d.ts.map +1 -0
- package/dist/{index-D-IiQIBB.d.ts → index-DMoxLG8P.d.ts} +3 -3
- package/dist/{index-D-IiQIBB.d.ts.map → index-DMoxLG8P.d.ts.map} +1 -1
- package/dist/{index-D_P7Ye43.d.ts → index-DNgf5gyG.d.ts} +2 -2
- package/dist/{index-D_P7Ye43.d.ts.map → index-DNgf5gyG.d.ts.map} +1 -1
- package/dist/{index-CGtH1piv.d.ts → index-DT73JraI.d.ts} +7 -38
- package/dist/index-DT73JraI.d.ts.map +1 -0
- package/dist/index-fNXZMCzX.d.ts +704 -0
- package/dist/index-fNXZMCzX.d.ts.map +1 -0
- package/dist/index.d.ts +197 -38
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +736 -28
- package/dist/index.js.map +1 -1
- package/dist/{insight-report-DRe8LB6d.d.ts → insight-report-08F022xN.d.ts} +4 -4
- package/dist/{insight-report-DRe8LB6d.d.ts.map → insight-report-08F022xN.d.ts.map} +1 -1
- package/dist/{integrity-DUNX9Fao.d.ts → integrity-B_EDELom.d.ts} +2 -2
- package/dist/{integrity-DUNX9Fao.d.ts.map → integrity-B_EDELom.d.ts.map} +1 -1
- package/dist/internal-BMFSR8Ns.js.map +1 -1
- package/dist/{kind-factory-DY8FdoXf.js → kind-factory-DMeEoMQZ.js} +3 -10
- package/dist/kind-factory-DMeEoMQZ.js.map +1 -0
- package/dist/ledger-core/index.js +2 -2
- package/dist/{ledger-core-BOzlRygb.js → ledger-core-PIfjCbKn.js} +2 -2
- package/dist/{ledger-core-BOzlRygb.js.map → ledger-core-PIfjCbKn.js.map} +1 -1
- package/dist/{llm-judge-Du7WQPh7.js → llm-judge-BtJ2Sfk_.js} +101 -20
- package/dist/llm-judge-BtJ2Sfk_.js.map +1 -0
- package/dist/matrix/index.d.ts +2 -2
- package/dist/{matrix-eXKRMHnL.d.ts → matrix-DiHmUobV.d.ts} +3 -3
- package/dist/{matrix-eXKRMHnL.d.ts.map → matrix-DiHmUobV.d.ts.map} +1 -1
- package/dist/meta-eval/index.d.ts +3 -3
- package/dist/meta-eval/index.js +1 -1
- package/dist/{mint-DfODW1KW.js → mint-DjfDUMHr.js} +2 -2
- package/dist/{mint-DfODW1KW.js.map → mint-DjfDUMHr.js.map} +1 -1
- package/dist/multishot/golden/index.d.ts +1 -1
- package/dist/multishot/index.d.ts +2 -2
- package/dist/openapi.json +4 -4
- package/dist/pipelines/index.d.ts +5 -5
- package/dist/pipelines/index.js +3 -3
- package/dist/{pareto-BqNW3LJR.d.ts → power-preflight-Ptse_Kq7.d.ts} +43 -43
- package/dist/power-preflight-Ptse_Kq7.d.ts.map +1 -0
- package/dist/{pre-registration-KN9jkh58.js → pre-registration-D94b7Of5.js} +2 -2
- package/dist/{pre-registration-KN9jkh58.js.map → pre-registration-D94b7Of5.js.map} +1 -1
- package/dist/{pre-registration-CzFCcwYk.d.ts → pre-registration-DHz6P_6f.d.ts} +2 -2
- package/dist/{pre-registration-CzFCcwYk.d.ts.map → pre-registration-DHz6P_6f.d.ts.map} +1 -1
- package/dist/{produced-state-Be0BK3RN.js → produced-state-Cm6DU_Ao.js} +4 -4
- package/dist/{produced-state-Be0BK3RN.js.map → produced-state-Cm6DU_Ao.js.map} +1 -1
- package/dist/profile-cell.js +1 -1
- package/dist/{promotion-policy-DtnOIZvk.d.ts → promotion-policy-WSXtBgBb.d.ts} +3 -3
- package/dist/{promotion-policy-DtnOIZvk.d.ts.map → promotion-policy-WSXtBgBb.d.ts.map} +1 -1
- package/dist/{run-score-lDzV0X8j.js → proposal-findings-bko3GGy-.js} +2 -31
- package/dist/proposal-findings-bko3GGy-.js.map +1 -0
- package/dist/{transient-failure-DKF5Mofa.d.ts → provenance-CafMdZKM.d.ts} +895 -891
- package/dist/provenance-CafMdZKM.d.ts.map +1 -0
- package/dist/{query-_5g6re3_.js → query-BPGMVlbM.js} +3 -3
- package/dist/{query-_5g6re3_.js.map → query-BPGMVlbM.js.map} +1 -1
- package/dist/{query-CwnHlu5p.d.ts → query-Na5gEIGd.d.ts} +3 -3
- package/dist/{query-CwnHlu5p.d.ts.map → query-Na5gEIGd.d.ts.map} +1 -1
- package/dist/{registry-8You7OK1.d.ts → registry-xEb_xfns.d.ts} +3 -3
- package/dist/{registry-8You7OK1.d.ts.map → registry-xEb_xfns.d.ts.map} +1 -1
- package/dist/{release-confidence-nGDJiiwc.js → release-confidence-CzUHc4z4.js} +3 -3
- package/dist/{release-confidence-nGDJiiwc.js.map → release-confidence-CzUHc4z4.js.map} +1 -1
- package/dist/{release-confidence-Dqt0NFep.d.ts → release-confidence-D6lQw_o7.d.ts} +4 -4
- package/dist/{release-confidence-Dqt0NFep.d.ts.map → release-confidence-D6lQw_o7.d.ts.map} +1 -1
- package/dist/reporting.d.ts +3 -3
- package/dist/reporting.js +2 -2
- package/dist/{researcher-Cz565b7D.d.ts → researcher-CMUTQXD7.d.ts} +6 -6
- package/dist/{researcher-Cz565b7D.d.ts.map → researcher-CMUTQXD7.d.ts.map} +1 -1
- package/dist/{reward-hacking-MBf7qpSB.d.ts → reward-hacking-CgPRUesA.d.ts} +2 -2
- package/dist/{reward-hacking-MBf7qpSB.d.ts.map → reward-hacking-CgPRUesA.d.ts.map} +1 -1
- package/dist/{reward-hacking-O5zKDANP.js → reward-hacking-SkxYgT0x.js} +2 -2
- package/dist/{reward-hacking-O5zKDANP.js.map → reward-hacking-SkxYgT0x.js.map} +1 -1
- package/dist/rl.d.ts +8 -8
- package/dist/rl.d.ts.map +1 -1
- package/dist/rl.js +6 -5
- package/dist/rl.js.map +1 -1
- package/dist/rollout/index.d.ts +1 -1
- package/dist/rollout/index.js +2 -2
- package/dist/{rollout-Dm2tSdiQ.js → rollout-Crypdx8s.js} +2 -2
- package/dist/{rollout-Dm2tSdiQ.js.map → rollout-Crypdx8s.js.map} +1 -1
- package/dist/{rubric-predictive-validity-CxycqzX5.d.ts → rubric-predictive-validity-DluJLCKQ.d.ts} +2 -2
- package/dist/{rubric-predictive-validity-CxycqzX5.d.ts.map → rubric-predictive-validity-DluJLCKQ.d.ts.map} +1 -1
- package/dist/{run-record-BC0ebuRP.js → run-record-DLORoL7t.js} +2 -2
- package/dist/{run-record-BC0ebuRP.js.map → run-record-DLORoL7t.js.map} +1 -1
- package/dist/{run-record-VVy4T9OW.d.ts → run-record-DQjRcYwA.d.ts} +3 -3
- package/dist/{run-record-VVy4T9OW.d.ts.map → run-record-DQjRcYwA.d.ts.map} +1 -1
- package/dist/{schema-k6ZBftVv.js → schema-CdIX2aHu.js} +5 -1
- package/dist/{schema-k6ZBftVv.js.map → schema-CdIX2aHu.js.map} +1 -1
- package/dist/{schema-Bjgdsn73.d.ts → schema-DID1Cqct.d.ts} +7 -3
- package/dist/{schema-Bjgdsn73.d.ts.map → schema-DID1Cqct.d.ts.map} +1 -1
- package/dist/{semantic-concept-judge-BsDMOwJr.js → semantic-concept-judge-I36eejJx.js} +2 -2
- package/dist/{semantic-concept-judge-BsDMOwJr.js.map → semantic-concept-judge-I36eejJx.js.map} +1 -1
- package/dist/{sequential-BLMbdrD7.js → sequential-B51qAYE4.js} +2 -2
- package/dist/{sequential-BLMbdrD7.js.map → sequential-B51qAYE4.js.map} +1 -1
- package/dist/{server-BjYiJHoJ.js → server-CCEnywOR.js} +26 -18
- package/dist/server-CCEnywOR.js.map +1 -0
- package/dist/{skillopt-optimization-method-UArRo-nr.js → skillopt-optimization-method-DzlF2RM7.js} +9 -9
- package/dist/{skillopt-optimization-method-UArRo-nr.js.map → skillopt-optimization-method-DzlF2RM7.js.map} +1 -1
- package/dist/{statistical-heldout-Cy3EhjlC.d.ts → statistical-heldout-UhiexnjU.d.ts} +3 -3
- package/dist/{statistical-heldout-Cy3EhjlC.d.ts.map → statistical-heldout-UhiexnjU.d.ts.map} +1 -1
- package/dist/{store-B06JdC56.d.ts → store-Cq9oOrI1.d.ts} +2 -2
- package/dist/{store-B06JdC56.d.ts.map → store-Cq9oOrI1.d.ts.map} +1 -1
- package/dist/{store-otlp-C_Rq5I4D.js → store-otlp-CHjBvWQY.js} +2 -2
- package/dist/{store-otlp-C_Rq5I4D.js.map → store-otlp-CHjBvWQY.js.map} +1 -1
- package/dist/{store-tool-spans-BVga3c37.js → store-tool-spans-B9o6tU8f.js} +3 -3
- package/dist/{store-tool-spans-BVga3c37.js.map → store-tool-spans-B9o6tU8f.js.map} +1 -1
- package/dist/{store-tool-spans-DPUG7UUY.d.ts → store-tool-spans-D_qMl2__.d.ts} +102 -36
- package/dist/store-tool-spans-D_qMl2__.d.ts.map +1 -0
- package/dist/storyboard/index.d.ts +1 -1
- package/dist/{summary-report-BXeQ5Ues.js → summary-report-Bgh8CpNK.js} +2 -2
- package/dist/{summary-report-BXeQ5Ues.js.map → summary-report-Bgh8CpNK.js.map} +1 -1
- package/dist/{summary-report-CC07PhEL.d.ts → summary-report-DRstQNBX.d.ts} +3 -3
- package/dist/{summary-report-CC07PhEL.d.ts.map → summary-report-DRstQNBX.d.ts.map} +1 -1
- package/dist/{task-failure-attributes-DTl-7-Kw.js → task-failure-attributes-CBGtLS_H.js} +3 -3
- package/dist/{task-failure-attributes-DTl-7-Kw.js.map → task-failure-attributes-CBGtLS_H.js.map} +1 -1
- package/dist/{tool-groups-Ci8i9ErB.d.ts → tool-groups-RGYfVWpc.d.ts} +3 -3
- package/dist/tool-groups-RGYfVWpc.d.ts.map +1 -0
- package/dist/{tool-waste-CKc7bYIg.d.ts → tool-waste-BrmLKxMw.d.ts} +4 -4
- package/dist/{tool-waste-CKc7bYIg.d.ts.map → tool-waste-BrmLKxMw.d.ts.map} +1 -1
- package/dist/{tool-waste-8BQiUc8K.js → tool-waste-CwGHzBzX.js} +2 -2
- package/dist/{tool-waste-8BQiUc8K.js.map → tool-waste-CwGHzBzX.js.map} +1 -1
- package/dist/trace-repair/index.d.ts +3 -3
- package/dist/trace-repair/index.d.ts.map +1 -1
- package/dist/trace-repair/index.js +4 -3
- package/dist/trace-repair/index.js.map +1 -1
- package/dist/traces.d.ts +59 -61
- package/dist/traces.d.ts.map +1 -1
- package/dist/traces.js +7 -7
- package/dist/{trajectory-Bi157Gun.d.ts → trajectory-r1bQqvBQ.d.ts} +3 -3
- package/dist/{trajectory-Bi157Gun.d.ts.map → trajectory-r1bQqvBQ.d.ts.map} +1 -1
- package/dist/trajectory-replay/index.d.ts +3 -3
- package/dist/trajectory-replay/index.js +1 -1
- package/dist/{types-BPb2Kf_C.d.ts → types-CMyW4GnH.d.ts} +3 -3
- package/dist/{types-BPb2Kf_C.d.ts.map → types-CMyW4GnH.d.ts.map} +1 -1
- package/dist/{types-BI4fT3HN.js → types-CiWITkGo.js} +11 -2
- package/dist/types-CiWITkGo.js.map +1 -0
- package/dist/{types-D9ssmxKL.d.ts → types-DMoNFDWi.d.ts} +6 -3
- package/dist/{types-D9ssmxKL.d.ts.map → types-DMoNFDWi.d.ts.map} +1 -1
- package/dist/{types-D4s7Z6nq.d.ts → types-JHMOqZI4.d.ts} +13 -3
- package/dist/types-JHMOqZI4.d.ts.map +1 -0
- package/dist/{verdict-B0xltqu6.js → verdict-BQ3pCFf8.js} +2 -2
- package/dist/{verdict-B0xltqu6.js.map → verdict-BQ3pCFf8.js.map} +1 -1
- package/dist/{verdict-cache-CdVVTVmn.js → verdict-cache-B3eCVQtY.js} +2 -2
- package/dist/{verdict-cache-CdVVTVmn.js.map → verdict-cache-B3eCVQtY.js.map} +1 -1
- package/dist/wire/index.d.ts +26 -11
- package/dist/wire/index.d.ts.map +1 -1
- package/dist/wire/index.js +2 -2
- package/docs/campaign-proposers.md +4 -0
- package/docs/code-agent-intake.md +64 -0
- package/docs/concepts.md +1 -1
- package/docs/design/statistics-decisions.md +1 -1
- package/docs/distributed-driver.md +3 -6
- package/docs/public-api.md +122 -106
- package/docs/wire-protocol.md +5 -3
- package/package.json +27 -19
- package/dist/benchmark-BhT16ep9.js.map +0 -1
- package/dist/benchmark-command-CF-4GEWZ.js.map +0 -1
- package/dist/campaign-DQZmc2Dq.js.map +0 -1
- package/dist/capture-fetch-CqwsJkkG.d.ts +0 -68
- package/dist/capture-fetch-CqwsJkkG.d.ts.map +0 -1
- package/dist/contract/index.d.ts.map +0 -1
- package/dist/dspy-rlm-engine-DhA9qKIm.js.map +0 -1
- package/dist/external-optimizer-process-BFmh36vW.js.map +0 -1
- package/dist/external-optimizer-subprocess-CqLMW3nh.js.map +0 -1
- package/dist/index-CGtH1piv.d.ts.map +0 -1
- package/dist/index-D-V8gCs_.d.ts.map +0 -1
- package/dist/index-vrJugRal.d.ts +0 -1
- package/dist/kind-factory-DY8FdoXf.js.map +0 -1
- package/dist/llm-judge-Du7WQPh7.js.map +0 -1
- package/dist/pareto-BqNW3LJR.d.ts.map +0 -1
- package/dist/run-score-lDzV0X8j.js.map +0 -1
- package/dist/server-BjYiJHoJ.js.map +0 -1
- package/dist/skillopt-optimization-method-x7TTF23P.d.ts.map +0 -1
- package/dist/store-tool-spans-DPUG7UUY.d.ts.map +0 -1
- package/dist/tool-groups-Ci8i9ErB.d.ts.map +0 -1
- package/dist/transient-failure-DKF5Mofa.d.ts.map +0 -1
- package/dist/types-BI4fT3HN.js.map +0 -1
- package/dist/types-D4s7Z6nq.d.ts.map +0 -1
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 { s as TraceStore } from "../store-B06JdC56.js";
|
|
3
2
|
import { p as ChatClient } from "../types-Bfk0uxRj.js";
|
|
4
|
-
import {
|
|
3
|
+
import { s as TraceStore } from "../store-Cq9oOrI1.js";
|
|
4
|
+
import { m as FeedbackTrajectoryStore } from "../feedback-trajectory-DIqpCyF0.js";
|
|
5
5
|
import { z } from "zod";
|
|
6
6
|
import { ServerType } from "@hono/node-server";
|
|
7
7
|
import { Hono } from "hono";
|
|
@@ -308,13 +308,28 @@ type ListRubricsResponse = z.infer<typeof ListRubricsResponseSchema>;
|
|
|
308
308
|
type VersionResponse = z.infer<typeof VersionResponseSchema>;
|
|
309
309
|
type ErrorResponse = z.infer<typeof ErrorResponseSchema>;
|
|
310
310
|
/**
|
|
311
|
-
* Bump on any breaking change to a request/response schema
|
|
312
|
-
*
|
|
311
|
+
* Bump the major on any breaking change to a request/response schema — the
|
|
312
|
+
* Python client compares majors and refuses a mismatch. Bump the minor when a
|
|
313
|
+
* response value changes meaning without changing its shape, which a client
|
|
314
|
+
* can detect but does not have to act on.
|
|
313
315
|
*/
|
|
314
|
-
declare const WIRE_VERSION = "1.
|
|
316
|
+
declare const WIRE_VERSION = "1.1.0";
|
|
315
317
|
/**
|
|
316
|
-
*
|
|
317
|
-
*
|
|
318
|
+
* The digest scheme `hashRubric` mints, named inside the value it produces.
|
|
319
|
+
* A consumer holding a tag from an earlier release reads a different scheme
|
|
320
|
+
* here and can tell "this package changed how it hashes" from "the rubric
|
|
321
|
+
* changed" — the two are indistinguishable when the value is an untagged hex
|
|
322
|
+
* string.
|
|
323
|
+
*/
|
|
324
|
+
declare const RUBRIC_VERSION_SCHEME = "sha256-rfc8785";
|
|
325
|
+
/**
|
|
326
|
+
* Stable identity of a rubric, as `<name>@sha256-rfc8785:<hex>`. Two scores
|
|
327
|
+
* are comparable when this matches exactly.
|
|
328
|
+
*
|
|
329
|
+
* The digest comes from `hashCanonical`, this package's only canonical-JSON
|
|
330
|
+
* encoder (RFC 8785). RFC 8785 orders keys by UTF-16 code unit, which is a
|
|
331
|
+
* property of the value alone — so the same rubric produces the same tag on
|
|
332
|
+
* every machine, which a locale-collated key order does not guarantee.
|
|
318
333
|
*/
|
|
319
334
|
declare function hashRubric(rubric: Rubric): string;
|
|
320
335
|
//#endregion
|
|
@@ -373,14 +388,14 @@ declare function handleTracesIngest(req: TracesIngestRequest, stores: IngestionS
|
|
|
373
388
|
*/
|
|
374
389
|
declare function handleFeedbackIngest(req: FeedbackTrajectory, stores: IngestionStores): Promise<FeedbackIngestResponse>;
|
|
375
390
|
//#endregion
|
|
376
|
-
//#region node_modules/.pnpm/openapi3-ts@4.6.
|
|
391
|
+
//#region node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/specification-extension.d.ts
|
|
377
392
|
type IExtensionName = `x-${string}`;
|
|
378
393
|
type IExtensionType = any;
|
|
379
394
|
type ISpecificationExtension = {
|
|
380
395
|
[extensionName: IExtensionName]: IExtensionType;
|
|
381
396
|
};
|
|
382
397
|
//#endregion
|
|
383
|
-
//#region node_modules/.pnpm/openapi3-ts@4.6.
|
|
398
|
+
//#region node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/oas-common.d.ts
|
|
384
399
|
interface ServerObject extends ISpecificationExtension {
|
|
385
400
|
url: string;
|
|
386
401
|
description?: string;
|
|
@@ -394,7 +409,7 @@ interface ServerVariableObject extends ISpecificationExtension {
|
|
|
394
409
|
description?: string;
|
|
395
410
|
}
|
|
396
411
|
//#endregion
|
|
397
|
-
//#region node_modules/.pnpm/openapi3-ts@4.6.
|
|
412
|
+
//#region node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/openapi31.d.ts
|
|
398
413
|
interface OpenAPIObject extends ISpecificationExtension {
|
|
399
414
|
openapi: string;
|
|
400
415
|
info: InfoObject;
|
|
@@ -783,5 +798,5 @@ interface StartedServer {
|
|
|
783
798
|
*/
|
|
784
799
|
declare function startServerAsync(opts?: ServeOptions): Promise<StartedServer>;
|
|
785
800
|
//#endregion
|
|
786
|
-
export { ErrorResponse, ErrorResponseSchema, FailureMode, FeedbackIngestResponse, FeedbackIngestResponseSchema, FeedbackTrajectory, FeedbackTrajectorySchema, HandleJudgeOptions, HealthResponseSchema, IngestionStores, JudgeRequest, JudgeRequestSchema, JudgeResult, JudgeResultSchema, ListRubricsResponse, ListRubricsResponseSchema, type RpcOptions, Rubric, RubricDimension, RubricDimensionSchema, RubricInfo, RubricSchema, type ServeOptions, type StartedServer, TraceEvent, TraceEventSchema, TracesIngestRequest, TracesIngestRequestSchema, TracesIngestResponse, TracesIngestResponseSchema, VersionResponse, VersionResponseSchema, WIRE_VERSION, WireError, buildOpenApi, createApp, dispatchRpc, getBuiltinRubric, handleFeedbackIngest, handleJudge, handleListRubrics, handleTracesIngest, handleVersion, hashRubric, listBuiltinRubrics, runRpcBatch, runRpcOnce, startServer, startServerAsync };
|
|
801
|
+
export { ErrorResponse, ErrorResponseSchema, FailureMode, FeedbackIngestResponse, FeedbackIngestResponseSchema, FeedbackTrajectory, FeedbackTrajectorySchema, HandleJudgeOptions, HealthResponseSchema, IngestionStores, JudgeRequest, JudgeRequestSchema, JudgeResult, JudgeResultSchema, ListRubricsResponse, ListRubricsResponseSchema, RUBRIC_VERSION_SCHEME, type RpcOptions, Rubric, RubricDimension, RubricDimensionSchema, RubricInfo, RubricSchema, type ServeOptions, type StartedServer, TraceEvent, TraceEventSchema, TracesIngestRequest, TracesIngestRequestSchema, TracesIngestResponse, TracesIngestResponseSchema, VersionResponse, VersionResponseSchema, WIRE_VERSION, WireError, buildOpenApi, createApp, dispatchRpc, getBuiltinRubric, handleFeedbackIngest, handleJudge, handleListRubrics, handleTracesIngest, handleVersion, hashRubric, listBuiltinRubrics, runRpcBatch, runRpcOnce, startServer, startServerAsync };
|
|
787
802
|
//# sourceMappingURL=index.d.ts.map
|
package/dist/wire/index.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/wire/schemas.ts","../../src/wire/handlers.ts","../../node_modules/.pnpm/openapi3-ts@4.6.
|
|
1
|
+
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/wire/schemas.ts","../../src/wire/handlers.ts","../../node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/specification-extension.d.ts","../../node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/oas-common.d.ts","../../node_modules/.pnpm/openapi3-ts@4.6.1/node_modules/openapi3-ts/dist/model/openapi31.d.ts","../../src/wire/openapi.ts","../../src/wire/rpc.ts","../../src/wire/rubrics.ts","../../src/wire/server.ts"],"x_google_ignoreList":[2,3,4],"mappings":";;;;;;;;cAoBa,uBAAqB,EAAA;;;;;;GAkBL,EAAA,KAAA;cAEvB,mBAAiB,EAAA;;;GAKE,EAAA,KAAA;cAIZ,cAAY,EAAA;;;;;;;;;;;;;;;;;;;GA6BL,EAAA,KAAA;cAIP,oBAAkB,EAAA;;;;;;;;;;;;;;;;;;;;;;;;;GAwBL,EAAA,KAAA;cAEb,mBAAiB,EAAA;;;;;;;;;GA6BL,EAAA,KAAA;cAInB,kBAAgB,EAAA;;;;;;;;;;GAcE,EAAA,KAAA;cAEX,2BAAyB,EAAA;;;;;;;;;;;;GAIL,EAAA,KAAA;cAIpB,uBAAqB,EAAA;;;;;GAWL,EAAA,KAAA;cAEhB,sBAAoB,EAAA;;;GAKL,EAAA,KAAA;;;;;;;cAUf,kBAAgB,EAAA;;;;;;;;;;;;;;;;GA0BL,EAAA,KAAA;cAEX,2BAAyB,EAAA;;;;;;;;;;;;;;;;;;GAQL,EAAA,KAAA;cAEpB,4BAA0B,EAAA;;;;;;;GAiBL,EAAA,KAAA;cA2DrB,0BAAwB,EAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA4BL,EAAA,KAAA;cAEnB,8BAA4B,EAAA;;;GAKL,EAAA,KAAA;KAExB,aAAa,EAAE,aAAa;KAC5B,sBAAsB,EAAE,aAAa;KACrC,uBAAuB,EAAE,aAAa;KACtC,qBAAqB,EAAE,aAAa;KACpC,yBAAyB,EAAE,aAAa;cAIvC,qBAAmB,EAAA;;;;;;GAcL,EAAA,KAAA;KAIf,kBAAkB,EAAE,aAAa;KACjC,cAAc,EAAE,aAAa;KAC7B,SAAS,EAAE,aAAa;KACxB,eAAe,EAAE,aAAa;KAC9B,cAAc,EAAE,aAAa;KAC7B,aAAa,EAAE,aAAa;KAC5B,sBAAsB,EAAE,aAAa;KACrC,kBAAkB,EAAE,aAAa;KACjC,gBAAgB,EAAE,aAAa;;;;;;;cAU9B;;;;;;;;cASA;;;;;;;;;;iBAWG,WAAW,QAAQ;;;;cCpXtB,kBAAkB;WAEX;WAEA;WACA;EAJlB,YACkB,cAChB,iBACgB,iBACA;;UAgJH;EACf,aAAa;EACb;;;;;;EAMA,OAAO;EACP;;EAEA,UAAU;EACV,SAAS;;iBAGW,YACpB,KAAK,cACL,UAAS,qBACR,QAAQ;iBAuEK,qBAAqB;iBAMrB,iBAAiB;;;;;;;;;UAmBhB;EACf,aAAa;EACb,gBAAgB;;;;;;;;;;;;iBAaI,mBACpB,KAAK,qBACL,QAAQ,kBACP,QAAQ;;;;;;iBA8BW,qBACpB,KAAK,oBACL,QAAQ,kBACP,QAAQ;;;KC9VC;KACA;KACA;GACP,eAAe,iBAAiB;;;;UCFpB,qBAAqB;EAClC;EACA;EACA;KACK,YAAY;;;UAGJ,6BAA6B;EAC1C;EACA;EACA;;;;UCPa,sBAAsB;EACnC;EACA,MAAM;EACN,UAAU;EACV,QAAQ;EACR,aAAa;EACb,WAAW;EACX,OAAO;EACP,eAAe;EACf,WAAW;;UAEE,mBAAmB;EAChC;EACA;EACA;EACA,UAAU;EACV,UAAU;EACV;;UAEa,sBAAsB;EACnC;EACA;EACA;;UAEa,sBAAsB;EACnC;EACA;EACA;;UAEa,yBAAyB;EACtC;KACK,iBAAiB,eAAe;;EAErC;KACK,mBAAmB,iBAAiB;;EAEzC;KACK,oBAAoB,kBAAkB;;EAE3C;KACK,kBAAkB,gBAAgB;;EAEvC;KACK,kBAAkB,oBAAoB;;EAE3C;KACK,iBAAiB,eAAe;;EAErC;KACK,yBAAyB,uBAAuB;;EAErD;KACK,eAAe,aAAa;;EAEjC;KACK,mBAAmB,iBAAiB;;EAEzC;KACK,mBAAmB,iBAAiB;;;UAG5B,oBAAoB;GAChC,eAAe;;UAIH,uBAAuB;EACpC;EACA;EACA;EACA,MAAM;EACN,MAAM;EACN,OAAO;EACP,SAAS;EACT,UAAU;EACV,OAAO;EACP,QAAQ;EACR,QAAQ;EACR,UAAU;EACV,cAAc,kBAAkB;;UAEnB,wBAAwB;EACrC;EACA;EACA;EACA,eAAe;EACf;EACA,cAAc,kBAAkB;EAChC,cAAc,oBAAoB;EAClC,YAAY;EACZ,YAAY;EACZ;EACA,WAAW;EACX,UAAU;;UAEG,oCAAoC;EACjD;EACA;;KAEQ;KACA;UACK,4BAA4B;EACzC;EACA;EACA;EACA;EACA,QAAQ;EACR;EACA;EACA,SAAS,eAAe;EACxB;KACK,gBAAgB,gBAAgB;;EAErC;EACA,UAAU;;UAEG,wBAAwB;EACrC;EACA,IAAI;;UAES,0BAA0B;EACvC;EACA,SAAS;EACT;;UAEa;GACZ,oBAAoB;;UAER,wBAAwB;EACrC,SAAS,eAAe;EACxB,WAAW;EACX;EACA,WAAW;;UAEE,uBAAuB;GACnC,mBAAmB;;UAEP;EACb;EACA;KACK,cAAc,eAAe;;EAElC;EACA;EACA;GACC;;UAEY,wBAAwB;EACrC,UAAU,iBAAiB;GAC1B,qBAAqB,iBAAiB;;UAE1B,uBAAuB;EACpC;EACA,UAAU;EACV,UAAU;EACV,QAAQ;;UAEK,wBAAwB;GACpC,eAAe,iBAAiB;;UAEpB,uBAAuB;GACnC,eAAe;;UAEH;GACZ,eAAe,eAAe;;UAElB;EACb;EACA;EACA;EACA;GACC;;UAEY;GACZ,eAAe,aAAa;;UAEhB,mBAAmB;EAChC;EACA;EACA,aAAa;EACb;EACA;EACA,SAAS;GACR;;UAEY;GACZ;;UAEY,qBAAqB;EAClC;;UAEa,kBAAkB;EAC/B;EACA;EACA,eAAe;GACd;;UAEY;GACZ,eAAe,gBAAgB;;UAEnB;EACb;EACA;EACA;;KAGQ;UACK,qBAAqB;EAClC;EACA,gBAAgB;EAChB;EACA;EACA,MAAM;EACN,eAAe;EACf;EACA;EACA;EACA,OAAO,mBAAmB;EAC1B;EACA,SAAS,eAAe;EACxB,SAAS,eAAe;EACxB,SAAS,eAAe;EACxB,MAAM,eAAe;EACrB,QAAQ,eAAe;EACvB;KACK,uBAAuB,eAAe;;EAE3C,uBAAuB,eAAe;EACtC,gBAAgB,eAAe;EAC/B;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA,eAAe,eAAe;EAC9B;EACA;;UAMa;EACb;EACA;KACK;;;UAGQ,kBAAkB;EAC/B;EACA;EACA;EACA;EACA;;KAEQ;UACK,6BAA6B;EAC1C,MAAM;EACN;EACA;EACA;EACA;EACA;EACA,QAAQ;EACR;;UAEa,yBAAyB;EACtC,WAAW;EACX,WAAW;EACX,oBAAoB;EACpB,oBAAoB;;UAEP,wBAAwB;EACrC;EACA;EACA;EACA,QAAQ;;UAEK,qBAAqB;GACjC;;UAEY;GACZ;;;;iBChRW,aAAa,yBAAyB;;;UCX5C;EACR;EACA;;UAGQ;EACR;;UAGQ;EACR;IAAS;IAAc;IAAiB;;;UAGzB;;EAEf,OAAO;EACP;;iBAGoB,YACpB,KAAK,YACL,UAAS,aACR,QAAQ,aAAa;;iBAqDF,WAAW,iBAAiB,UAAS,aAAkB;;iBAuBvD,YAAY,iBAAiB,UAAS,aAAkB;;;;iBCjB9D,iBAAiB,eAAe;;iBAKhC;;;;;;;;;;;;;UCtEC;;EAEf,SAAS;;;;;;EAMT,OAAO;;EAEP;;;;;;;;;;EAUA;IACE,mBAAmB,4BAA4B;;;iBAMnC,UAAU,OAAM,mBAAqB,0BAAA,+BAAA;UA0IpC,qBAAqB;;EAEpC;;EAEA;;iBAGc,YAAY,OAAM,eAAoB;UAUrC;EACf,QAAQ;;;;EAIR;;EAEA;;EAEA,SAAS;;;;;;;iBAQK,iBAAiB,OAAM,eAAoB,QAAQ"}
|
package/dist/wire/index.js
CHANGED
|
@@ -1,2 +1,2 @@
|
|
|
1
|
-
import { A as
|
|
2
|
-
export { ErrorResponseSchema, FeedbackIngestResponseSchema, FeedbackTrajectorySchema, HealthResponseSchema, JudgeRequestSchema, JudgeResultSchema, ListRubricsResponseSchema, RubricDimensionSchema, RubricSchema, TraceEventSchema, TracesIngestRequestSchema, TracesIngestResponseSchema, VersionResponseSchema, WIRE_VERSION, WireError, buildOpenApi, createApp, dispatchRpc, getBuiltinRubric, handleFeedbackIngest, handleJudge, handleListRubrics, handleTracesIngest, handleVersion, hashRubric, listBuiltinRubrics, runRpcBatch, runRpcOnce, startServer, startServerAsync };
|
|
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-CCEnywOR.js";
|
|
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 };
|
|
@@ -531,6 +531,10 @@ const proposer: SurfaceProposer = {
|
|
|
531
531
|
Return a label and rationale when they will help later analysis.
|
|
532
532
|
Candidate creation must not read final test results.
|
|
533
533
|
|
|
534
|
+
A proposer may also attach `attribution`: one opaque, JSON-safe record the loop carries byte-for-byte onto `GenerationCandidate.attribution` and the loop provenance record.
|
|
535
|
+
The loop never interprets it.
|
|
536
|
+
Tag it with your own schema field and validate it on readback — `PolicyEdit` candidate records (`makePolicyEditCandidateRecord` from the analyst surface) are the first producer, which is what lets a forecast in an edit be scored against the measured delta later.
|
|
537
|
+
|
|
534
538
|
`runOptimization()` rejects a candidate whose `surfaceHash` was already admitted.
|
|
535
539
|
This includes the baseline, an earlier generation, and another candidate in the same proposal.
|
|
536
540
|
The complete proposal is checked before candidate dispatch, so duplicates cannot consume candidate cells.
|
|
@@ -0,0 +1,64 @@
|
|
|
1
|
+
# Coding-agent session intake
|
|
2
|
+
|
|
3
|
+
A coding agent writes its session to a JSONL transcript.
|
|
4
|
+
This package turns that transcript into `RunRecord`s, a provider-neutral action stream, and per-session metrics, so a session that already happened can be scored, compared, and analyzed like any other run.
|
|
5
|
+
|
|
6
|
+
Import from `@tangle-network/agent-eval/contract`.
|
|
7
|
+
|
|
8
|
+
## The three steps
|
|
9
|
+
|
|
10
|
+
```ts
|
|
11
|
+
import {
|
|
12
|
+
fromCodexSession,
|
|
13
|
+
parseCodeAgentJsonlFile,
|
|
14
|
+
} from '@tangle-network/agent-eval/contract'
|
|
15
|
+
|
|
16
|
+
const { entries, malformedLines } = await parseCodeAgentJsonlFile('/path/to/rollout.jsonl')
|
|
17
|
+
const { runs, diagnostics, metrics, observations } = fromCodexSession({
|
|
18
|
+
entries,
|
|
19
|
+
malformedLines,
|
|
20
|
+
sourcePath: '/path/to/rollout.jsonl',
|
|
21
|
+
scenarioId: 'refactor-billing',
|
|
22
|
+
})
|
|
23
|
+
```
|
|
24
|
+
|
|
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
|
+
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.
|
|
28
|
+
|
|
29
|
+
## One function per harness
|
|
30
|
+
|
|
31
|
+
| Harness | Function |
|
|
32
|
+
|---|---|
|
|
33
|
+
| Codex | `fromCodexSession` |
|
|
34
|
+
| Claude Code | `fromClaudeCodeSession` |
|
|
35
|
+
| OpenCode | `fromOpenCodeSession` |
|
|
36
|
+
| Kimi Code | `fromKimiCodeSession` |
|
|
37
|
+
| Pi | `fromPiSession` |
|
|
38
|
+
|
|
39
|
+
They share one implementation and one options shape (`CodeAgentSessionIntakeOptions`); the harness decides how a session id, a terminal state, and a cost receipt are read out of the entries.
|
|
40
|
+
Supply `experimentId`, `candidateId`, `seed`, `splitTag`, `scenarioId`, `model`, `promptHash`, `configHash`, `commitSha`, and `score` when the surrounding experiment knows them — an unsupplied field stays unknown rather than becoming a zero.
|
|
41
|
+
`costProvenance` sets the cost receipt explicitly; without it a source-reported cost wins, then a token-priced estimate, then uncaptured.
|
|
42
|
+
|
|
43
|
+
`metrics` carries the per-session counts a comparison needs without re-reading the transcript: `userMessages`, `assistantMessages`, `reasoningItems`, `toolCalls`, `toolOutputs`, `toolErrors`, `unclassifiedErrors`, `patchAttempts`, `patchSuccesses`, `patchFailures`, `turnsStarted`.
|
|
44
|
+
|
|
45
|
+
## Transcripts larger than memory
|
|
46
|
+
|
|
47
|
+
`parseCodeAgentJsonlFile` holds every entry, so it grows with the transcript.
|
|
48
|
+
`streamCodeAgentJsonlFile(path)` is an async generator that keeps live memory bounded by the longest single line — use it when a rollout is large enough that the entry array itself is the problem.
|
|
49
|
+
Reading a session above V8's ~512MB string ceiling through the string path throws `ERR_STRING_TOO_LONG`; the streaming path does not.
|
|
50
|
+
|
|
51
|
+
## The action stream on its own
|
|
52
|
+
|
|
53
|
+
`observeCodeAgentSession({ source, entries })` returns just the `CodeAgentSessionObservation`: the final text, the terminal state, and the provider-neutral action list.
|
|
54
|
+
Raw prompts, tool inputs, and tool outputs stay out of that projection — keep the source JSONL as the evidence for them.
|
|
55
|
+
|
|
56
|
+
## Traces that are not coding-agent sessions
|
|
57
|
+
|
|
58
|
+
`parseAgentTrace(records)` indexes a generic agent trace, and `partitionRunsByAuthoringModel(...)` splits its runs by the model that authored them, which is what a contamination check or a per-model comparison needs.
|
|
59
|
+
`fromOtelSpans(...)` is the entry point for OTLP spans.
|
|
60
|
+
|
|
61
|
+
## Related
|
|
62
|
+
|
|
63
|
+
- [`docs/concepts.md`](./concepts.md) — what a `RunRecord` is.
|
|
64
|
+
- [`docs/trace-analysis.md`](./trace-analysis.md) — analyzing the runs once they exist.
|
package/docs/concepts.md
CHANGED
|
@@ -214,7 +214,7 @@ A rubric describes:
|
|
|
214
214
|
Built-in rubrics ship in `src/wire/rubrics.ts`, including `anti-slop` for technical-buyer voice.
|
|
215
215
|
You can also pass the same rubric shape inline at the call site.
|
|
216
216
|
|
|
217
|
-
A rubric is plain data. The
|
|
217
|
+
A rubric is plain data. The digest of that data, tagged with the scheme that produced it, is the `rubricVersion`. Two scores are only comparable if they used the same `rubricVersion`: change the rubric and you start a new comparison series.
|
|
218
218
|
|
|
219
219
|
## How verifiers work
|
|
220
220
|
|
|
@@ -354,6 +354,6 @@ Result:
|
|
|
354
354
|
- **3 fences were not TypeScript** and are now marked `text`: an object fragment in `campaign-proposers.md`, a call elided as `{ ... }` in `feedback-trajectories.md`, and a method sketch in `examples/benchmarks/README.md`.
|
|
355
355
|
- **2 fences imported a path that cannot resolve** — `@tangle-network/agent-eval/../src/trace-repair` in `trace-repair-admission.md` and `trace-repair-continuation.md` — now `@tangle-network/agent-eval/trace-repair`.
|
|
356
356
|
- **0 of the remaining 63 name an export that does not exist.** Every symbol a fence imports is on the current surface.
|
|
357
|
-
- The remaining compiler complaints are inherent to an excerpt: a name declared in the prose around it, a shorthand property, top-level `await` outside a module. Three fences import a deliberate placeholder (`your-text-optimizer`, `./my-engine`)
|
|
357
|
+
- The remaining compiler complaints are inherent to an excerpt: a name declared in the prose around it, a shorthand property, top-level `await` outside a module. Three fences import a deliberate placeholder (`your-text-optimizer`, `./my-engine`).
|
|
358
358
|
|
|
359
359
|
There is no fence checker in `scripts/` and no CI job. Compiling illustrative snippets fails builds for prose edits, and the failure mode it catches — a doc naming a symbol that no longer exists — is what the export census and review already cover. Repeat this pass by hand after a large surface change.
|
|
@@ -1,8 +1,5 @@
|
|
|
1
1
|
# Distributed campaign execution: coordinator-on-A, workers-on-B
|
|
2
2
|
|
|
3
|
-
> **Status: source-only, not published as of 0.140.x.** `@tangle-network/agent-eval/adapters/http` (`httpDispatch`, `runDispatchServer`) exists in source at `src/adapters/http.ts` but has no matching entry in `package.json` `exports`.
|
|
4
|
-
> This doc describes the design and the wire shape.
|
|
5
|
-
> Every `import` line below, run against `@tangle-network/agent-eval` installed from npm, throws `ERR_PACKAGE_PATH_NOT_EXPORTED`.
|
|
6
3
|
|
|
7
4
|
The coordinator process runs `runCampaign()`, `runImprovementLoop()`, or `compareOptimizationMethods()`.
|
|
8
5
|
The worker runs your actual agent.
|
|
@@ -62,7 +59,7 @@ Coordinator-on-A talks to one worker-on-B over HTTP.
|
|
|
62
59
|
**Coordinator side (machine A):**
|
|
63
60
|
|
|
64
61
|
```ts
|
|
65
|
-
import { httpDispatch } from '@tangle-network/agent-eval/adapters/http'
|
|
62
|
+
import { httpDispatch } from '@tangle-network/agent-eval/adapters/http'
|
|
66
63
|
|
|
67
64
|
const dispatch = httpDispatch<MyScenario, MyArtifact>({
|
|
68
65
|
url: 'https://worker.your-infra.com/dispatch',
|
|
@@ -79,7 +76,7 @@ await runImprovementLoop({ scenarios, baselineSurface, dispatchWithSurface: (sur
|
|
|
79
76
|
**Worker side (machine B):**
|
|
80
77
|
|
|
81
78
|
```ts
|
|
82
|
-
import { runDispatchServer } from '@tangle-network/agent-eval/adapters/http'
|
|
79
|
+
import { runDispatchServer } from '@tangle-network/agent-eval/adapters/http'
|
|
83
80
|
|
|
84
81
|
const handle = await runDispatchServer<MyScenario, MyArtifact>({
|
|
85
82
|
dispatch: async (scenario, ctx) => {
|
|
@@ -102,7 +99,7 @@ The coordinator picks a region per cell; the same `httpDispatch` routes to
|
|
|
102
99
|
different worker URLs based on placement.
|
|
103
100
|
|
|
104
101
|
```ts
|
|
105
|
-
import { httpDispatch } from '@tangle-network/agent-eval/adapters/http'
|
|
102
|
+
import { httpDispatch } from '@tangle-network/agent-eval/adapters/http'
|
|
106
103
|
|
|
107
104
|
const REGION_URLS: Record<string, string> = {
|
|
108
105
|
'us-east': 'https://worker-use1.your-infra.com/dispatch',
|