@tangle-network/agent-eval 0.187.2 → 0.188.0
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 +55 -0
- package/dist/adapters/http.d.ts +2 -2
- package/dist/{agent-profile-uAIQK3Nr.d.ts → agent-profile-CfbuZk8b.d.ts} +4 -4
- package/dist/{agent-profile-uAIQK3Nr.d.ts.map → agent-profile-CfbuZk8b.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 +6 -6
- package/dist/{analyze-runs-DapYZz02.d.ts → analyze-runs-CLz5O01x.d.ts} +5 -5
- package/dist/{analyze-runs-DapYZz02.d.ts.map → analyze-runs-CLz5O01x.d.ts.map} +1 -1
- package/dist/{benchmark-command-DGokT8zd.js → benchmark-command-CdUGku3W.js} +12 -12
- package/dist/benchmark-command-CdUGku3W.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 +5 -5
- package/dist/{campaign-DlxdTOvw.js → campaign-DZuysBbS.js} +6 -6
- package/dist/{campaign-DlxdTOvw.js.map → campaign-DZuysBbS.js.map} +1 -1
- 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-CFj9dfcn.d.ts} +4 -4
- package/dist/{client-DCc7a8ZC.d.ts.map → client-CFj9dfcn.d.ts.map} +1 -1
- package/dist/contract/index.d.ts +11 -11
- package/dist/contract/index.js +6 -6
- 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/{default-registry-DMxeb4GD.d.ts → default-registry-BLsGfnoO.d.ts} +6 -6
- package/dist/{default-registry-DMxeb4GD.d.ts.map → default-registry-BLsGfnoO.d.ts.map} +1 -1
- package/dist/{default-registry-Ca2IWH1X.js → default-registry-BtZ_t-9b.js} +2 -2
- package/dist/{default-registry-Ca2IWH1X.js.map → default-registry-BtZ_t-9b.js.map} +1 -1
- package/dist/{define-agent-eval-C0EqJ9Gz.js → define-agent-eval-ChtddasD.js} +3 -3
- package/dist/{define-agent-eval-C0EqJ9Gz.js.map → define-agent-eval-ChtddasD.js.map} +1 -1
- package/dist/{define-agent-eval-CGBwPPaq.d.ts → define-agent-eval-CsicmDxt.d.ts} +7 -7
- package/dist/{define-agent-eval-CGBwPPaq.d.ts.map → define-agent-eval-CsicmDxt.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 +83 -134
- package/dist/diagnosis/index.js.map +1 -1
- package/dist/{dspy-rlm-engine-BE9lCzzL.js → dspy-rlm-engine-sPX7B5ac.js} +3 -3
- package/dist/{dspy-rlm-engine-BE9lCzzL.js.map → dspy-rlm-engine-sPX7B5ac.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-Gw2n1irt.d.ts} +15 -5
- package/dist/engine-Gw2n1irt.d.ts.map +1 -0
- package/dist/{eval-campaign-XFHEpN86.js → eval-campaign-CeotQ1ZR.js} +7 -4
- package/dist/{eval-campaign-XFHEpN86.js.map → eval-campaign-CeotQ1ZR.js.map} +1 -1
- package/dist/evaluation.d.ts +2 -2
- package/dist/{exact-types-BEHN7tnq.d.ts → exact-types-CrhPoDp8.d.ts} +2 -2
- package/dist/{exact-types-BEHN7tnq.d.ts.map → exact-types-CrhPoDp8.d.ts.map} +1 -1
- package/dist/experiment/index.d.ts +3 -3
- package/dist/experiment/index.js +1 -2
- package/dist/experiment/index.js.map +1 -1
- package/dist/{external-optimizer-process-BbhvXTFx.js → external-optimizer-process-D3BLpZyR.js} +2 -2
- package/dist/{external-optimizer-process-BbhvXTFx.js.map → external-optimizer-process-D3BLpZyR.js.map} +1 -1
- package/dist/{external-optimizer-subprocess-DMgmqTB6.js → external-optimizer-subprocess-BrfsgL92.js} +2 -2
- package/dist/{external-optimizer-subprocess-DMgmqTB6.js.map → external-optimizer-subprocess-BrfsgL92.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-BJi4t3Vf.d.ts} +6 -6
- package/dist/{feedback-trajectory-CytTR7u4.d.ts.map → feedback-trajectory-BJi4t3Vf.d.ts.map} +1 -1
- package/dist/hosted/index.d.ts +2 -2
- package/dist/{index-CKZa_rnt.d.ts → index-B-s8qDRJ.d.ts} +2 -2
- package/dist/{index-CKZa_rnt.d.ts.map → index-B-s8qDRJ.d.ts.map} +1 -1
- package/dist/{index-YZrbnBSV.d.ts → index-BD1oSbHr.d.ts} +11 -11
- package/dist/{index-YZrbnBSV.d.ts.map → index-BD1oSbHr.d.ts.map} +1 -1
- package/dist/{index-vYNHCV2G.d.ts → index-Ca0_IjFu.d.ts} +7 -7
- package/dist/{index-vYNHCV2G.d.ts.map → index-Ca0_IjFu.d.ts.map} +1 -1
- package/dist/{index-BRsrfSHT.d.ts → index-eTTnPe3r.d.ts} +9 -9
- package/dist/{index-BRsrfSHT.d.ts.map → index-eTTnPe3r.d.ts.map} +1 -1
- package/dist/index.d.ts +383 -68
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +1482 -207
- package/dist/index.js.map +1 -1
- package/dist/{insight-report-DpOo-3iL.d.ts → insight-report-DeE1solA.d.ts} +4 -4
- package/dist/{insight-report-DpOo-3iL.d.ts.map → insight-report-DeE1solA.d.ts.map} +1 -1
- 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/{kind-factory-V79Hw0AU.js → kind-factory-g1ROw67A.js} +15 -4
- package/dist/{kind-factory-V79Hw0AU.js.map → kind-factory-g1ROw67A.js.map} +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-DF18J3CR.js} +9 -9
- package/dist/llm-judge-DF18J3CR.js.map +1 -0
- package/dist/{matrix-CPKsRY8Y.d.ts → matrix-BvA1WUpj.d.ts} +2 -2
- package/dist/{matrix-CPKsRY8Y.d.ts.map → matrix-BvA1WUpj.d.ts.map} +1 -1
- package/dist/meta-eval/index.d.ts +4 -4
- package/dist/{mint-BI14xehJ.d.ts → mint-CJFSYBIU.d.ts} +3 -3
- package/dist/{mint-BI14xehJ.d.ts.map → mint-CJFSYBIU.d.ts.map} +1 -1
- package/dist/multishot/golden/index.d.ts +1 -1
- package/dist/multishot/index.d.ts +2 -2
- package/dist/openapi.json +1 -1
- package/dist/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-CF8Z43Fj.d.ts} +2 -2
- package/dist/{prime-protocol-CDBcL9NM.d.ts.map → prime-protocol-CF8Z43Fj.d.ts.map} +1 -1
- package/dist/{produced-state-Cn9ECr4v.js → produced-state-xsMMjYVc.js} +3 -3
- package/dist/{produced-state-Cn9ECr4v.js.map → produced-state-xsMMjYVc.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-BM2Kdk8T.d.ts} +4 -4
- package/dist/{registry-D5Ly3GRV.d.ts.map → registry-BM2Kdk8T.d.ts.map} +1 -1
- package/dist/{release-confidence-CSp3MsU4.d.ts → release-confidence-CG4Z9QLB.d.ts} +4 -4
- package/dist/{release-confidence-CSp3MsU4.d.ts.map → release-confidence-CG4Z9QLB.d.ts.map} +1 -1
- 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/{researcher-D7E92STT.d.ts → researcher-Bjcd_8T_.d.ts} +7 -7
- package/dist/{researcher-D7E92STT.d.ts.map → researcher-Bjcd_8T_.d.ts.map} +1 -1
- package/dist/{reward-hacking-BIQ2BWLG.d.ts → reward-hacking-DrKXJtgH.d.ts} +2 -2
- package/dist/{reward-hacking-BIQ2BWLG.d.ts.map → reward-hacking-DrKXJtgH.d.ts.map} +1 -1
- package/dist/rl.d.ts +8 -8
- package/dist/rl.js +1 -1
- package/dist/rollout/index.d.ts +2 -2
- package/dist/{rubric-predictive-validity-CTU8y8Cv.d.ts → rubric-predictive-validity-B14yBAmM.d.ts} +2 -2
- package/dist/{rubric-predictive-validity-CTU8y8Cv.d.ts.map → rubric-predictive-validity-B14yBAmM.d.ts.map} +1 -1
- package/dist/{run-record-D9wS35fK.d.ts → run-record-xEULpgML.d.ts} +2 -2
- package/dist/{run-record-D9wS35fK.d.ts.map → run-record-xEULpgML.d.ts.map} +1 -1
- package/dist/{schema-CR5cpjQ3.d.ts → schema-2g7trUwH.d.ts} +17 -2
- package/dist/schema-2g7trUwH.d.ts.map +1 -0
- package/dist/schema-CSf6qWgZ.js.map +1 -1
- package/dist/{semantic-concept-judge-Cnt0TWp0.js → semantic-concept-judge-70LxrCSL.js} +5 -5
- package/dist/{semantic-concept-judge-Cnt0TWp0.js.map → semantic-concept-judge-70LxrCSL.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-C60h0Got.js} +5 -5
- package/dist/{skillopt-optimization-method-Cc9SVLqi.js.map → skillopt-optimization-method-C60h0Got.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-CPv-y6sK.d.ts} +3 -3
- package/dist/{statistical-heldout-DTmhTW8P.d.ts.map → statistical-heldout-CPv-y6sK.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-DPgzSQKS.js} +2 -2
- package/dist/{store-otlp-2gYCy-v0.js.map → store-otlp-DPgzSQKS.js.map} +1 -1
- package/dist/{store-tool-spans-B3bX9yqn.d.ts → store-tool-spans-C2lLwDzg.d.ts} +9 -54
- package/dist/store-tool-spans-C2lLwDzg.d.ts.map +1 -0
- package/dist/{store-tool-spans-B1e6Auc2.js → store-tool-spans-ClPxcvh2.js} +4 -4
- package/dist/{store-tool-spans-B1e6Auc2.js.map → store-tool-spans-ClPxcvh2.js.map} +1 -1
- package/dist/storyboard/index.d.ts +1 -1
- package/dist/{summary-report-Iy6kXo8x.d.ts → summary-report-SNvJoAvD.d.ts} +3 -3
- package/dist/{summary-report-Iy6kXo8x.d.ts.map → summary-report-SNvJoAvD.d.ts.map} +1 -1
- 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-CCQqAa16.d.ts} +3 -3
- package/dist/tool-groups-CCQqAa16.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 +8 -8
- 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--18oHwVF.d.ts → types-5ak2PqnJ.d.ts} +3 -3
- package/dist/{types--18oHwVF.d.ts.map → types-5ak2PqnJ.d.ts.map} +1 -1
- package/dist/{types-BTRcIqFr.d.ts → types-B1vDZ694.d.ts} +4 -4
- package/dist/{types-BTRcIqFr.d.ts.map → types-B1vDZ694.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-BjB2wZM9.d.ts → types-DDp2Mb8Y.d.ts} +3 -3
- package/dist/{types-BjB2wZM9.d.ts.map → types-DDp2Mb8Y.d.ts.map} +1 -1
- package/dist/wire/index.d.ts +3 -3
- package/dist/wire/index.js +1 -1
- package/docs/diagnosis.md +2 -3
- package/docs/eval-surface-map.md +1 -1
- package/docs/experiment.md +0 -1
- package/docs/redaction.md +59 -0
- package/docs/trace-contracts.md +102 -0
- package/docs/verdicts.md +1 -1
- package/package.json +3 -2
- package/dist/benchmark-command-DGokT8zd.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/integrity-Cy9WHAtb.js.map +0 -1
- package/dist/integrity-rGOfSUle.d.ts.map +0 -1
- package/dist/llm-judge-BVso5x-u.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/schema-CR5cpjQ3.d.ts.map +0 -1
- package/dist/store-tool-spans-B3bX9yqn.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-CBbLtr2J.d.ts.map +0 -1
package/CHANGELOG.md
CHANGED
|
@@ -6,6 +6,61 @@ All notable changes to `@tangle-network/agent-eval` and its sibling `agent-eval-
|
|
|
6
6
|
|
|
7
7
|
## Unreleased
|
|
8
8
|
|
|
9
|
+
## [0.188.0] — 2026-09-24
|
|
10
|
+
|
|
11
|
+
### Added
|
|
12
|
+
|
|
13
|
+
- `precedes` and `neverUnless` take `{ order }`: `start-order` (default), `finish-before-start`, or `all-occurrences`.
|
|
14
|
+
- New trace-contract operators: `atMost(p, max)`, `tokensAtMost(p, max)` (an unknown token count fails), `run({ requireCompleted, allowedStatuses, maxDurationMs })` (unknown statuses are rejected), and `argument(p, { pointer, check, occurrence })`, an RFC 6901 JSON Pointer check on tool-call arguments that fails when the arguments were not captured.
|
|
15
|
+
- Predicates gain `kind` (read through agent-trace-contract's classifier), `model`, `not`, and the `{ oneOf: [...] }` matcher.
|
|
16
|
+
- A contract can be scoped to one subtree with `scope({ root })`, and can list `alternative(id, ...)` rule sets of which one must pass.
|
|
17
|
+
- `evaluateTraceContract` and the contract types are exported from the package root.
|
|
18
|
+
- `/diagnosis` `rankFirstFailure()` names the first failure of one run by a fixed precedence: the innermost failing span, then a failed `agent.outcome`, then an explicit `none`.
|
|
19
|
+
Failures that ended at the same instant, or without an end time, are `ambiguous` rather than ordered by guess.
|
|
20
|
+
The failing span's message is classified by the failure taxonomy, so its `blame` separates machine and provider failures from the agent's own.
|
|
21
|
+
`diagnoseSpans` reports one per run as `facts.firstFailures`.
|
|
22
|
+
- `/pipelines` `diffSteps()` diffs any two step lists, and `diffStepsFromSpans()` orders one run's flat OTLP spans for it, so trace consumers share one diff.
|
|
23
|
+
- `compileTraceContractSpec()` compiles a declarative JSON contract (`run`, `tools`, `llm`, `scope`, `alternatives.anyOf`, low-level `rules`) onto these operators.
|
|
24
|
+
Parsing is strict: an unknown key or status is an error that names the closest known word.
|
|
25
|
+
`lintTraceContractSpec()` reports contradictions and likely mistakes, and `explainTraceContract()` states each rule in one line.
|
|
26
|
+
See `docs/trace-contracts.md`.
|
|
27
|
+
- `/traces` owns the one redaction core ([docs/redaction.md](./docs/redaction.md)): `redact(value, { profile })` with `default`, `share` and `strict` profiles and per-profile string byte caps, `redactText`, and `classifyKey`, which normalizes camel, kebab and dotted field names and keeps token counts (`inputTokens`, `max_tokens`, `token_count`) and names such as `author`.
|
|
28
|
+
A string that holds a credential (bearer, JWT, `sk-`, `sk-ant-`, `AIza`, `gh*_`, `AKIA`, PEM, `key=value` and others) is replaced whole.
|
|
29
|
+
`knownSecrets` removes exact values in plain, base64, base64url and URL-encoded form.
|
|
30
|
+
- `assessShareSafety`, `redactForShare`, `combineVerdicts` and `shareAllowed` give a share verdict of `SAFE`, `SAFE_WITH_WARNINGS`, `UNSAFE` or `UNKNOWN`; `redactForShare` re-scans the redacted output, and a part the scanner cannot read makes the verdict `UNKNOWN`.
|
|
31
|
+
- `pnpm redaction:corpus` runs the core over `scripts/redaction-corpus.json`, the must-flag and must-not-flag cases (including agent-inspect's safety corpus, MIT).
|
|
32
|
+
|
|
33
|
+
### Changed
|
|
34
|
+
|
|
35
|
+
- `TraceEmitter` never throws a failed store write into the traced run.
|
|
36
|
+
It counts written and dropped writes, reports each failure to `onCaptureError` (default: one process warning), and writes the counts onto the Run as `capture` when it ends.
|
|
37
|
+
`assertRunCaptured` reports `dropped_writes` for a run whose producer dropped records, so an incomplete run no longer reads as a shorter one.
|
|
38
|
+
- `TraceEmitter.within` runs its callback in its own async context, so parallel `within` calls parent their own children without passing `parentSpanId`.
|
|
39
|
+
`currentSpanId()` names the span a new span would be parented to.
|
|
40
|
+
- Trace contracts no longer pass when a required step never ran.
|
|
41
|
+
`precedes(a, b)` now fails when either endpoint is missing.
|
|
42
|
+
For the conditional form ("every `b` needs an earlier `a`; a run without `b` passes"), use `neverUnless(b, a)`, which keeps that meaning.
|
|
43
|
+
- Ordering rules read timestamps only.
|
|
44
|
+
A span whose needed `startedAt` or `endedAt` is missing fails the rule; array position is no longer taken as time, and mixed timestamps no longer throw for rules that do not order.
|
|
45
|
+
- Every rule reports `pass`, `fail`, or `error` in `ContractVerdict.ruleExecutions`.
|
|
46
|
+
A rule whose predicate throws, or a scope that selects no unique subtree, is `error`; the verdict's `status` is then `error` and it is never valid.
|
|
47
|
+
Errored rules are absent from `scores` rather than scored 0.
|
|
48
|
+
- `/pipelines` `firstDivergenceView` pairs steps by id, then by position with the same name and kind, then by name and kind anywhere, instead of by index alone.
|
|
49
|
+
One inserted step no longer marks every later step as diverged.
|
|
50
|
+
The report adds `diff`: paired steps with their field differences, steps only in A or only in B, and a first divergence classified as `changed`, `replaced`, `only-in-a`, `only-in-b` or `reordered`.
|
|
51
|
+
Paired steps now also compare `status`, so a step that failed in one run and passed in the other is a divergence.
|
|
52
|
+
The unused `stepEquals` option is removed.
|
|
53
|
+
- Every trace analyst tool (`buildTraceAnalysisToolDescriptors`) now ends its description with `UNTRUSTED_TRACE_TEXT`, a warning that returned trace text is untrusted evidence and never instructions.
|
|
54
|
+
Before, only `readSpanSource` warned.
|
|
55
|
+
Each descriptor also declares `readOnly: true` and `idempotent: true`, so a transport such as an MCP server publishes them without restating them.
|
|
56
|
+
|
|
57
|
+
### Removed
|
|
58
|
+
|
|
59
|
+
- **The rule-list redactor is replaced by the redaction core.** Removed `DEFAULT_REDACTION_RULES`, `RedactionRule`, `redactString` and `redactValue` (root and `/traces`), and `/diagnosis` `DIAGNOSIS_SECRET_RULES`, `holdsSecret`, `redactSecrets`, `redactSecretsDeep`, `SECRET_ASSIGNMENT_PATTERN` and their types. `RedactionReport` now has `byDetector` and `findings` instead of `byRule`, and `REDACTION_VERSION` is `2.0.0`. **Migration:** `redactValue(v)` becomes `redact(v).value`, `redactString(s).output` becomes `redactText(s)`, and a custom rule list becomes `knownSecrets` or a case in the corpus. `defaultProviderRedactor` now runs the core, so a credential header keeps its name with a `[REDACTED:credential-key]` value, and `redactedFields` holds JSON Pointers.
|
|
60
|
+
- **`ExperimentTracker` and its git-provenance/persistence machinery are gone.** Removed `ExperimentTracker`, `fileExperimentStore`, `inMemoryExperimentStore`, `Experiment`, and `ExperimentProvenance` (root and `/experiment`). The class had no in-repo, agent-runtime, blueprint-agent, or agent-dev-container caller — a manual experiment becomes a search with `proposer.kind: 'human'` in the upcoming search-tree system. **Migration:** if you called `new ExperimentTracker({ store, provenanceReader })`, replace it with your own store (`create`/`addRep`/`list` become plain reads and writes of whatever you persist) plus `computeExperimentStats` and `improvementVerdict` directly — those two pure functions, `ExperimentRep`, `ExperimentStats`, `ImprovementThresholds`, `ImprovementVerdictResult`, and `ExperimentVerdict` are unchanged and still exported from the package root (blueprint-agent's held-out gate uses them as-is). Provenance capture (`git rev-parse HEAD`, etc.) is no longer built in; shell out yourself if you need it.
|
|
61
|
+
- `contractJudge`, `matchSpan`, and `assertTraceContract`, which the package root never exported.
|
|
62
|
+
- The unit tests in `tests/trace-contracts.test.ts`; `traces check` over recorded sessions is the proof.
|
|
63
|
+
|
|
9
64
|
## [0.187.2] — 2026-09-24
|
|
10
65
|
|
|
11
66
|
### Added
|
package/dist/adapters/http.d.ts
CHANGED
|
@@ -1,5 +1,5 @@
|
|
|
1
|
-
import { R as Scenario, d as DispatchContext, f as DispatchFn } from "../types-
|
|
2
|
-
import "../index-
|
|
1
|
+
import { R as Scenario, d as DispatchContext, f as DispatchFn } from "../types-B1vDZ694.js";
|
|
2
|
+
import "../index-BD1oSbHr.js";
|
|
3
3
|
//#region src/adapters/http.d.ts
|
|
4
4
|
interface HttpDispatchOptions<TScenario extends Scenario, _TArtifact> {
|
|
5
5
|
/** Static endpoint URL. Mutually exclusive with `resolveUrl`. */
|
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
import { b as CustomTokenPricing, g as CostReceiptInput } from "./cost-ledger-DbQdN3nO.js";
|
|
2
|
-
import { D as TraceAnalysisStore, c as AnalystRunInputs, f as AnalystUsageReceipt, i as AnalystFinding, p as EvidenceRef } from "./types-
|
|
3
|
-
import { C as ChatResponse, S as ChatRequest } from "./types-
|
|
4
|
-
import { c as RegistryRunOpts, n as AnalystRegistry } from "./registry-
|
|
2
|
+
import { D as TraceAnalysisStore, c as AnalystRunInputs, f as AnalystUsageReceipt, i as AnalystFinding, p as EvidenceRef } from "./types-DDp2Mb8Y.js";
|
|
3
|
+
import { C as ChatResponse, S as ChatRequest } from "./types-C0bWoSKC.js";
|
|
4
|
+
import { c as RegistryRunOpts, n as AnalystRegistry } from "./registry-BM2Kdk8T.js";
|
|
5
5
|
import { AgentProfile, AgentProfile as AgentProfile$1, HarnessType, HarnessType as HarnessType$1 } from "@tangle-network/agent-interface";
|
|
6
6
|
//#region src/campaign/external-optimizer-contracts.d.ts
|
|
7
7
|
interface ExternalOptimizerProcessLimits {
|
|
@@ -485,4 +485,4 @@ declare function agentProfileId(profile: AgentProfile): string;
|
|
|
485
485
|
declare function agentProfileHash(profile: AgentProfile): string;
|
|
486
486
|
//#endregion
|
|
487
487
|
export { runAnalystBenchmark as A, ExternalOptimizerModelBudget as B, AnalystEvidenceResolutionError as C, AnalystLatencyDistribution as D, AnalystIssueExpectation as E, ExternalOptimizerCallbackLimits as F, ExternalOptimizerProcessLimits as G, ExternalOptimizerModelCallRequest as H, ExternalOptimizerChatRequest as I, ExternalOptimizerWireCounts as J, ExternalOptimizerResumeMode as K, ExternalOptimizerEndpointFormat as L, scoreAnalystFindings as M, DEFAULT_EXTERNAL_OPTIMIZER_CALLBACK_LIMITS as N, RunAnalystBenchmarkOptions as O, DEFAULT_EXTERNAL_OPTIMIZER_PROCESS_LIMITS as P, resolveExternalOptimizerProcessLimits as Q, ExternalOptimizerEvaluationObservation as R, AnalystEvidenceResolution as S, AnalystFindingScore as T, ExternalOptimizerModelCallResult as U, ExternalOptimizerModelCall as V, ExternalOptimizerModelExecutionObservation as W, ExternalTextEvaluationRequest as X, ExternalTextCandidate as Y, resolveExternalOptimizerCallbackLimits as Z, AnalystBenchmarkProvenance as _, ProfileAxisSpec as a, AnalystBenchmarkSummary as b, expandProfileAxes as c, AnalystBenchmarkDatasetRef as d, AnalystBenchmarkDescriptor as f, AnalystBenchmarkOutput as g, AnalystBenchmarkObservation as h, HarnessType$1 as i, traceStoreEvidenceResolver as j, registryBenchmarkRunner as k, harnessAxisOf as l, AnalystBenchmarkLabelState as m, CODING_HARNESSES as n, agentProfileHash as o, AnalystBenchmarkError as p, ExternalOptimizerRunnerCommand as q, HARNESS_NATIVE_MODEL as r, agentProfileId as s, AgentProfile$1 as t, AnalystBenchmarkCase as u, AnalystBenchmarkResult as v, AnalystEvidenceResolver as w, AnalystEvidenceExpectation as x, AnalystBenchmarkRunner as y, ExternalOptimizerEvaluationRefusalReason as z };
|
|
488
|
-
//# sourceMappingURL=agent-profile-
|
|
488
|
+
//# sourceMappingURL=agent-profile-CfbuZk8b.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"agent-profile-
|
|
1
|
+
{"version":3,"file":"agent-profile-CfbuZk8b.d.ts","names":[],"sources":["../src/campaign/external-optimizer-contracts.ts","../src/analyst/benchmark-scoring.ts","../src/analyst/benchmark.ts","../src/agent-profile.ts"],"mappings":";;;;;;UAWiB;;EAEf;;EAEA;;EAEA;;cAGW,2CAA2C,SAAS;UAOhD;;EAEf;;EAEA;;cAGW,4CAA4C,SAAS;UAMjD;EACf;EACA;EACA,MAAM,OAAO;;EAEb,SAAS,QAAQ;;KAGP;KAEA,iCAAiC;UAE5B;EACf,WAAW;EACX;;iBAUc,sCACd,OAAO,QAAQ,6CACf,iBACC;iBAmBa,uCACd,OAAO,QAAQ,8CACf,iBACC;KAkBE,aAAa,KAAK,eAAc,6BACjC,IACA,0BAA0B,gBACf,aAAa,OACtB,+BACc,WAAW,IAAI,aAAa,EAAE,SAC1C;;KAGI,+BAA+B,aACzC,KAAK;EAA0B;;KAGrB;;UAMK;;WAEN;;WAEA,SAAS;;WAET,iBAAiB;WACjB,QAAQ;;;KAIP;WAEG;;WAEA,UAAU;;;;;;;WAOV,SAPU;;WASV;;WAGA;;WAEA;;WAEA,SATkC;;WAWlC;;;;;;;;;;;;KAaH,8BACV,SAAS,sCACN,QAAQ;;KAGD;WAEG;WACA;WACA;WACA;WACA;WACA;WACA;WACA;;WAGA;WACA;WACA;WACA;WACA;WACA;WACA;WACA;;KAGH;;;;;;KAUA;WAEG;WACA;WACA,WAAW;WACX;;WAGA;WACA;WACA,WAAW;WACX;WACA;;WAEA;WACA;;WAGA;WACA;WACA,QAAQ;WACR,YAAY;WACZ;WACA;;UAGE;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;;;EAUA;;;;;;EAMA,UAAU;;EAEV;;;UAIe;;EAEf;;EAEA;;;;iBCtQc,qBACd,UAAU,KAAK,oEACf,mBAAmB,mBAClB;;;UCIc;EACf;EACA,OAAO;;UAGQ;EACf;EACA;EACA;EACA;EACA,oBAAoB;EACpB;;EAEA,4BAA4B;;KAGlB;UAEK,qBAAqB;EACpC;;EAEA;;EAEA,YAAY;EACZ,OAAO;EACP,yBAAyB;;EAEzB,2BAA2B;EAC3B;EACA,WAAW;;UAGI;EACf;EACA;EACA;EACA;EACA;EACA,mBAAmB;EACnB;EACA;EACA;EACA;;EAEA;;EAEA;;EAEA;EACA;;UAGe;EACf,UAAU;EACV;EACA;;UAGe;EACf;EACA;EACA,oBAAoB;EACpB,QAAQ;;EAER;;KAGU,wBAAwB,qBAAqB;EACvD;EACA,WAAW;EACX,UAAU;EACV,SAAS;gBACK;;;;;iBAMA,2BAA2B,QACzC,WAAW,OAAO,WAAW,qBAC5B,wBAAwB;UAoBV;EACf,mBAAmB;EACnB,QAAQ;EACR,WAAW;;;;;EAKX;;EAEA,QAAQ;;UAGO;EACf;EACA;EACA;EACA;;UAGe,uBAAuB;EACtC;EACA,QACE,OAAO,QACP;IAAW;IAAgB;IAAoB,SAAS;MACvD,yBAAyB,QAAQ;;UAGrB;EACf;EACA;EACA;EACA,YAAY;EACZ;EACA;EACA;EACA;EACA,mBAAmB;EACnB,OAAO;EACP,qBAAqB;EACrB;EACA,eAAe;EACf,QAAQ;EACR,iBAAiB;EACjB,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;EACA;EACA;EACA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;EAEA;;EAEA;;EAEA;;EAEA;EACA,WAAW;EACX;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;;UAGe;EACf;EACA,UAAU;EACV;EACA,cAAc;EACd,WAAW;;UAGI,mCAAmC;EAClD;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf,YAAY;EACZ,cAAc;EACd,WAAW;;UAGI,2BAA2B;EAC1C,gBAAgB,qBAAqB;EACrC,kBAAkB,uBAAuB;EACzC;EACA;EACA;EACA,kBAAkB,wBAAwB;EAC1C,YAAY;;EAEZ,+BAA+B;EAC/B,iBAAiB,aAAa,uCAAuC;EACrE,SAAS;;iBAGW,oBAAoB,QACxC,SAAS,2BAA2B,UACnC,QAAQ;iBAyDK,wBAAwB;EACtC;EACA,UAAU;EACV,aAAa,KAAK;;EAElB;IACE,uBAAuB;;;;;;;;;;cCpUd,2BAA2B;UAOvB;;;EAGf,MAAM;;EAEN,qBAAqB;;;EAGrB;;;;;;EAMA;;;;;;;cAQW;;;;;;;;;;;;;;;;;;;iBAoBG,kBAAkB,MAAM,kBAAkB;;;;;;;;iBAwD1C,cACd,SAAS,KAAK;EACX,SAAS;EAAa;;;;;;;;;iBAiBX,eAAe,SAAS;;;;;;;;;;;iBAmDxB,iBAAiB,SAAS"}
|
package/dist/analyst/index.d.ts
CHANGED
|
@@ -1,14 +1,14 @@
|
|
|
1
1
|
import { b as CustomTokenPricing, c as CostLedgerHandle } from "../cost-ledger-DbQdN3nO.js";
|
|
2
|
-
import { D as TraceAnalysisStore, _ as ProposalFinding, a as AnalystInputKind, b as makeFinding, c as AnalystRunInputs, d as AnalystSeverity, f as AnalystUsageReceipt, g as ExecutionProbeRequest, h as ExecutionProbeOutcome, i as AnalystFinding, l as AnalystRunResult, m as ExecutionProbe, n as AnalystContext, o as AnalystRequirements, p as EvidenceRef, r as AnalystCost, s as AnalystRunEvent, t as Analyst, u as AnalystRunSummary, v as ProposalFindingOrigin, x as makeProposalFinding, y as computeFindingId, z as TraceAnalystSpan } from "../types-
|
|
3
|
-
import { A as createChatClient, C as ChatResponse, D as MockTransportOpts, E as CustomTransportOpts, O as OpenAiCompatibleTransportOpts, S as ChatRequest, T as CreateChatClientOpts, b as ChatCallOpts, k as SandboxSdkTransportOpts, w as ChatTransport, x as ChatClient } from "../types-
|
|
4
|
-
import { A as runAnalystBenchmark, C as AnalystEvidenceResolutionError, D as AnalystLatencyDistribution, E as AnalystIssueExpectation, M as scoreAnalystFindings, O as RunAnalystBenchmarkOptions, S as AnalystEvidenceResolution, T as AnalystFindingScore, V as ExternalOptimizerModelCall, W as ExternalOptimizerModelExecutionObservation, _ as AnalystBenchmarkProvenance, b as AnalystBenchmarkSummary, d as AnalystBenchmarkDatasetRef, f as AnalystBenchmarkDescriptor, g as AnalystBenchmarkOutput, h as AnalystBenchmarkObservation, j as traceStoreEvidenceResolver, k as registryBenchmarkRunner, m as AnalystBenchmarkLabelState, p as AnalystBenchmarkError, q as ExternalOptimizerRunnerCommand, t as AgentProfile, u as AnalystBenchmarkCase, v as AnalystBenchmarkResult, w as AnalystEvidenceResolver, x as AnalystEvidenceExpectation, y as AnalystBenchmarkRunner } from "../agent-profile-
|
|
5
|
-
import { a as ExactAnalystRunPolicySnapshot, c as ExactAnalystSnapshot, d as ExactExecutionComponentSnapshot, i as ExactAnalystRunEvent, l as ExactCapableAnalyst, n as ExactAnalystExecutionPlanSnapshot, o as ExactAnalystRunResult, r as ExactAnalystRunCompletion, s as ExactAnalystRunSummary, t as ExactAnalystBudgetSnapshot, u as ExactExecutionComponentIdentity } from "../exact-types-
|
|
6
|
-
import { a as ExactAnalystBudgetPolicy, c as RegistryRunOpts, i as BudgetPolicy, n as AnalystRegistry, o as ExactAnalystRunExecutionError, r as AnalystRegistryOptions, s as ExactRegistryRunOpts, t as AnalystHooks } from "../registry-
|
|
7
|
-
import { a as createTraceAnalyst, c as BehavioralAnalystOptions, i as TraceAnalystDefinition, l as behavioralAnalyst, n as buildDefaultAnalystRegistry, o as renderPriorFindings, r as CreateTraceAnalystOptions, s as runTraceAnalyst, t as DefaultAnalystRegistryOptions, u as deriveEfficiencyFindings } from "../default-registry-
|
|
8
|
-
import { a as TraceAnalystLimits, c as RawAnalystEvidence, d as evidenceRefsFromRawFinding, f as parseRawFinding, i as TraceAnalysisEngineResult, l as RawAnalystFinding, n as TraceAnalysisEngine, o as resolveTraceAnalystLimits, r as TraceAnalysisEngineRequest, s as RAW_FINDING_SCHEMA_PROMPT, t as DEFAULT_TRACE_ANALYST_OUTPUT_TOKENS, u as RawAnalystFindingSchema } from "../engine-
|
|
9
|
-
import { n as buildTraceToolsForGroup, t as TraceToolGroupName } from "../tool-groups-
|
|
10
|
-
import { A as SemanticConceptJudgeInput, C as DspyRlmTraceEngineOptions, E as createChatTraceEngine, S as renderFindingSubject, T as ChatTraceEngineOptions, _ as FindingSubject, a as FAILURE_MODE_KIND_SPEC, b as findingSubjectGrammarPromptFor, c as emitControlIntegrityFindings, d as FindingsStore, f as PersistedFinding, g as FINDING_SUBJECT_SYNTAX, h as FINDING_SUBJECT_KINDS, i as IMPROVEMENT_KIND_SPEC, j as SemanticConceptJudgeOptions, l as DiffPolicy, m as diffFindings, n as KNOWLEDGE_POISONING_KIND_SPEC, o as CONTROL_INTEGRITY_ANALYST, p as defaultIsMaterial, r as KNOWLEDGE_GAP_KIND_SPEC, s as ControlIntegrityAnalyst, t as DEFAULT_TRACE_ANALYST_KINDS, u as FindingsDiff, v as FindingSubjectKind, w as createDspyRlmTraceEngine, x as parseFindingSubject, y as KIND_EXPECTED_SUBJECTS } from "../index-
|
|
11
|
-
import { A as ReplyRowDecoded, C as primeProtocolSha256, D as DecodedReply, E as runPrimeExchange, M as decodeReplyRows, O as ReplyContract, S as normalizePrimeUsage, T as projectPrimeTrajectory, _ as buildPrimePrompt, a as PrimeProjectionOutcome, b as extractPrimeJsonObject, c as PrimeProtocolIdentity, d as PrimeRepairPromptSpec, f as PrimeRepairState, g as analystUsageReceiptFromPrimeUsage, h as PrimeTurnRecord, i as PrimeProjectionDelivery, j as ReplyRowRejection, k as ReplyEnvelope, l as PrimeRawUsage, m as PrimeRowDecoded, n as PrimeExchangeOutcome, o as PrimeProjectionSource, p as PrimeReplyContract, r as PrimeFailure, s as PrimePromptSpec, t as PrimeExchangeOptions, u as PrimeRejectedRow, v as buildPrimeRepairPrompt, w as primeReplyDefect, x as mergePrimeRawUsage, y as emptyPrimeRawUsage } from "../prime-protocol-
|
|
2
|
+
import { D as TraceAnalysisStore, _ as ProposalFinding, a as AnalystInputKind, b as makeFinding, c as AnalystRunInputs, d as AnalystSeverity, f as AnalystUsageReceipt, g as ExecutionProbeRequest, h as ExecutionProbeOutcome, i as AnalystFinding, l as AnalystRunResult, m as ExecutionProbe, n as AnalystContext, o as AnalystRequirements, p as EvidenceRef, r as AnalystCost, s as AnalystRunEvent, t as Analyst, u as AnalystRunSummary, v as ProposalFindingOrigin, x as makeProposalFinding, y as computeFindingId, z as TraceAnalystSpan } from "../types-DDp2Mb8Y.js";
|
|
3
|
+
import { A as createChatClient, C as ChatResponse, D as MockTransportOpts, E as CustomTransportOpts, O as OpenAiCompatibleTransportOpts, S as ChatRequest, T as CreateChatClientOpts, b as ChatCallOpts, k as SandboxSdkTransportOpts, w as ChatTransport, x as ChatClient } from "../types-C0bWoSKC.js";
|
|
4
|
+
import { A as runAnalystBenchmark, C as AnalystEvidenceResolutionError, D as AnalystLatencyDistribution, E as AnalystIssueExpectation, M as scoreAnalystFindings, O as RunAnalystBenchmarkOptions, S as AnalystEvidenceResolution, T as AnalystFindingScore, V as ExternalOptimizerModelCall, W as ExternalOptimizerModelExecutionObservation, _ as AnalystBenchmarkProvenance, b as AnalystBenchmarkSummary, d as AnalystBenchmarkDatasetRef, f as AnalystBenchmarkDescriptor, g as AnalystBenchmarkOutput, h as AnalystBenchmarkObservation, j as traceStoreEvidenceResolver, k as registryBenchmarkRunner, m as AnalystBenchmarkLabelState, p as AnalystBenchmarkError, q as ExternalOptimizerRunnerCommand, t as AgentProfile, u as AnalystBenchmarkCase, v as AnalystBenchmarkResult, w as AnalystEvidenceResolver, x as AnalystEvidenceExpectation, y as AnalystBenchmarkRunner } from "../agent-profile-CfbuZk8b.js";
|
|
5
|
+
import { a as ExactAnalystRunPolicySnapshot, c as ExactAnalystSnapshot, d as ExactExecutionComponentSnapshot, i as ExactAnalystRunEvent, l as ExactCapableAnalyst, n as ExactAnalystExecutionPlanSnapshot, o as ExactAnalystRunResult, r as ExactAnalystRunCompletion, s as ExactAnalystRunSummary, t as ExactAnalystBudgetSnapshot, u as ExactExecutionComponentIdentity } from "../exact-types-CrhPoDp8.js";
|
|
6
|
+
import { a as ExactAnalystBudgetPolicy, c as RegistryRunOpts, i as BudgetPolicy, n as AnalystRegistry, o as ExactAnalystRunExecutionError, r as AnalystRegistryOptions, s as ExactRegistryRunOpts, t as AnalystHooks } from "../registry-BM2Kdk8T.js";
|
|
7
|
+
import { a as createTraceAnalyst, c as BehavioralAnalystOptions, i as TraceAnalystDefinition, l as behavioralAnalyst, n as buildDefaultAnalystRegistry, o as renderPriorFindings, r as CreateTraceAnalystOptions, s as runTraceAnalyst, t as DefaultAnalystRegistryOptions, u as deriveEfficiencyFindings } from "../default-registry-BLsGfnoO.js";
|
|
8
|
+
import { a as TraceAnalystLimits, c as RawAnalystEvidence, d as evidenceRefsFromRawFinding, f as parseRawFinding, i as TraceAnalysisEngineResult, l as RawAnalystFinding, n as TraceAnalysisEngine, o as resolveTraceAnalystLimits, r as TraceAnalysisEngineRequest, s as RAW_FINDING_SCHEMA_PROMPT, t as DEFAULT_TRACE_ANALYST_OUTPUT_TOKENS, u as RawAnalystFindingSchema } from "../engine-Gw2n1irt.js";
|
|
9
|
+
import { n as buildTraceToolsForGroup, t as TraceToolGroupName } from "../tool-groups-CCQqAa16.js";
|
|
10
|
+
import { A as SemanticConceptJudgeInput, C as DspyRlmTraceEngineOptions, E as createChatTraceEngine, S as renderFindingSubject, T as ChatTraceEngineOptions, _ as FindingSubject, a as FAILURE_MODE_KIND_SPEC, b as findingSubjectGrammarPromptFor, c as emitControlIntegrityFindings, d as FindingsStore, f as PersistedFinding, g as FINDING_SUBJECT_SYNTAX, h as FINDING_SUBJECT_KINDS, i as IMPROVEMENT_KIND_SPEC, j as SemanticConceptJudgeOptions, l as DiffPolicy, m as diffFindings, n as KNOWLEDGE_POISONING_KIND_SPEC, o as CONTROL_INTEGRITY_ANALYST, p as defaultIsMaterial, r as KNOWLEDGE_GAP_KIND_SPEC, s as ControlIntegrityAnalyst, t as DEFAULT_TRACE_ANALYST_KINDS, u as FindingsDiff, v as FindingSubjectKind, w as createDspyRlmTraceEngine, x as parseFindingSubject, y as KIND_EXPECTED_SUBJECTS } from "../index-Ca0_IjFu.js";
|
|
11
|
+
import { A as ReplyRowDecoded, C as primeProtocolSha256, D as DecodedReply, E as runPrimeExchange, M as decodeReplyRows, O as ReplyContract, S as normalizePrimeUsage, T as projectPrimeTrajectory, _ as buildPrimePrompt, a as PrimeProjectionOutcome, b as extractPrimeJsonObject, c as PrimeProtocolIdentity, d as PrimeRepairPromptSpec, f as PrimeRepairState, g as analystUsageReceiptFromPrimeUsage, h as PrimeTurnRecord, i as PrimeProjectionDelivery, j as ReplyRowRejection, k as ReplyEnvelope, l as PrimeRawUsage, m as PrimeRowDecoded, n as PrimeExchangeOutcome, o as PrimeProjectionSource, p as PrimeReplyContract, r as PrimeFailure, s as PrimePromptSpec, t as PrimeExchangeOptions, u as PrimeRejectedRow, v as buildPrimeRepairPrompt, w as primeReplyDefect, x as mergePrimeRawUsage, y as emptyPrimeRawUsage } from "../prime-protocol-CF8Z43Fj.js";
|
|
12
12
|
import { i as nodeHttpPrimeBridgeTransport, n as PrimeBridgeTransportRequest, r as PrimeBridgeTransportResult, t as PrimeBridgeTransport } from "../prime-bridge-transport-6feEglLf.js";
|
|
13
13
|
import { z } from "zod";
|
|
14
14
|
//#region src/analyst/adapters.d.ts
|
|
@@ -1181,11 +1181,11 @@ declare function runAnalystBenchmarkCommand(argv: readonly string[], env?: NodeJ
|
|
|
1181
1181
|
declare const ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM = "sha256-canonical-source-manifest";
|
|
1182
1182
|
declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM = "sha256-canonical-file-manifest";
|
|
1183
1183
|
declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES: readonly string[];
|
|
1184
|
-
declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "
|
|
1184
|
+
declare const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "c61fa0edc749ac199316c36caf4ff568681667ba5d8e6b01b22b6f1583ec6523";
|
|
1185
1185
|
declare const ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256 = "1e03f2daed356d60316aabefb407ec1e437ac94d408d61eea4ae096e9c6fbb5b";
|
|
1186
1186
|
declare const ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256 = "4dba263b6256a30d56c7fdb2d992d3a953c0035d731f359b704db806f68f75ac";
|
|
1187
1187
|
declare const ANALYST_BENCHMARK_IMPLEMENTATION_FILES: readonly string[];
|
|
1188
|
-
declare const ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 = "
|
|
1188
|
+
declare const ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 = "de168cc2cbca1623a88fc0b033bbf92f92d5fb1bd2aa317d8c2c453eb61fd9df";
|
|
1189
1189
|
declare function analystBenchmarkImplementationDigest(): string;
|
|
1190
1190
|
declare function analystBenchmarkDependencyLockDigest(): string;
|
|
1191
1191
|
//#endregion
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/analyst/adapters.ts","../../src/analyst/benchmark-agentrx-calibration.ts","../../src/analyst/benchmark-dataset-types.ts","../../src/analyst/benchmark-dataset-agentrx.ts","../../src/analyst/benchmark-dataset-codetrace.ts","../../src/analyst/benchmark-dataset-utils.ts","../../src/analyst/benchmark-verification-outcome.ts","../../src/analyst/benchmark-verification-artifacts.ts","../../src/analyst/benchmark-public-types.ts","../../src/analyst/benchmark-public-adapters.ts","../../src/analyst/benchmark-public-data.ts","../../src/analyst/definition.ts","../../src/analyst/benchmark-public-prompt.ts","../../src/analyst/benchmark-public-model.ts","../../src/analyst/benchmark-public-rlm.ts","../../src/analyst/benchmark-runner-prime.ts","../../src/analyst/benchmark-comparison.ts","../../src/analyst/benchmark-public-calibration.ts","../../src/analyst/benchmark-command-artifact.ts","../../src/analyst/benchmark-command-result.ts","../../src/analyst/benchmark-command.ts","../../src/analyst/benchmark-implementation.ts","../../src/analyst/benchmark-instructions-override.ts","../../src/analyst/benchmark-report.ts","../../src/analyst/benchmark-summary.ts","../../src/analyst/bind.ts","../../src/analyst/define.ts","../../src/analyst/kinds/skill-usage.ts","../../src/analyst/parse-tolerant.ts","../../src/analyst/proposal-findings.ts"],"mappings":";;;;;;;;;;;;;;UA8CiB;EACf;EACA;;EAEA,SAAS,KAAK;;EAEd;;iBAGc,kCACd,MAAM,kCACL,QAAQ;;;cCtDE;UAEI;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA,SAAS;;iBAGK,4BACd,QAAQ,wBACR,2BACC;iBAkBa,iCAAiC,SAAS;;;KCtD9C;UAEK;EACf,YAAY;EACZ;EACA;EACA;EACA;EACA;;UAGe;EACf,eAAe;EACf,mBAAmB;EACnB;IAAe,YAAY;IAAY;;EACvC,wBAAwB;EACxB;EACA;EACA;;UAGe;EACf,UAAU;EACV;EACA;EACA;EACA;;UAGe;EACf,UAAU;EACV,mBAAmB;EACnB;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA,iBAAiB;;KAGP,0CAEC,sCACA,iCACA;UAEI;EACf;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA,oCAAoC;;UAGrB;EACf,eAAe;EACf,WAAW,sBAAsB;;KAGvB;UAEK;;;;;EAKf,WAAW;;UAGI,kCACP,yBACN;UAEa,oCAAoC;EACnD;;EAEA;;UAGe,yCAAyC;EACxD;EACA;EACA;EACA;;UAGe,2CACP,kCACN;;;iBCtGY,qBAAqB,QACnC,KAAK,YACL,OAAO,QACP,UAAS,8BACR,qBAAqB;;iBAsGR,6BACd,mBAAmB,YACnB,iBACA,UAAS,mCACR;iBA6Ea,yBAAyB;;iBAoNzB,iBAAiB;;;iBC/YjB,mBAAmB,QACjC,KAAK,mBACL,OAAO,QACP,UAAS,4BACR,qBAAqB;;iBAwFR,gCACd,2BACA,aAAa,uBACb,UAAS,qCACR;;;iBClHa,wBAAwB;;;KCA5B;UAEK;EACf;EACA;EACA,QAAQ;;UAGO;EACf,QAAQ;EACR;EAKA;IAAe;IAAe;;EAC9B,SAAS;EACT;EACA;EACA;EACA;;UAGe;EACf;EACA;;iBA8Ec,yBACd,gBAAgB,2BACf;;;cC/FU;KAED;UAEK;EACf,MAAM;EACN;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;EACT;EACA;EACA;EACA;EACA;EACA,OAAO;EACP,cAAc;EACd,UAAU,OAAO;;UAGF;EACf,UAAU;EACV,SAAS;EACT,OAAO,MAAM;IAA6B;;;iBAatB,mCAAmC;EACvD;EACA,KAAK;EACL;IACE,QAAQ;iBAwHI,kCACd,kBACA,iBACA,WAAW,6BACX;;;KC7KU;;;;;;;;UASK;;WAEN;;WAEA;;;UAIM;EACf,MAAM;EACN;EACA,kBAAkB,aAAa;;EAE/B,UAAU;;UAGK;;EAEf,MAAM;;EAEN;;EAEA,kBAAkB,aAAa;EAC/B;EACA;EACA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA,UAAU;;EAEV;;EAEA,uBAAuB;EACvB;IACE,SAAS;IACT;IACA;IACA;IACA;IACA;IACA;IACA;IACA;;;;;;IAMA;;EAEF,aAAa;EACb;IACE;IACA;;;;;;;;;KAUQ,sCAAsC,KAChD,iEAGA,QAAQ,KAAK;UAEE;EACf,OAAO,qBAAqB;EAC5B;EACA;EACA;EACA,YAAY;IACV;IACA;IACA;;EAEF,uBAAuB;EACvB,WAAW;;UAGI;EACf;EACA;EACA,QAAQ;;UAGO;EACf,OAAO;EACP,OAAO;EACP,OAAO;EACP,YAAY;EACZ,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;EACA;EACA,QAAQ;EACR,UAAU;;;;;;;;;;UChHK;EACf;EACA;;EAEA;EACA;EACA,UAAU;EACV;EACA;EACA;EACA;EACA,WAAW;;;;;;;;;UAUI;EACf;EACA;;EAEA,kCAAkC;;EAElC;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;UASe;EACf;EACA,OAAO;;iBAGO,8BAA8B,uBAAuB;iBAiB/C,6BAA6B;EACjD,SAAS;EACT;EACA,mBAAmB;EACnB;EACA,OAAO;EACP,SAAS;IACP;EACF,UAAU;EACV,aAAa;;EAEb,aAAa;;;;;;;;;iBA4MO,6BAA6B;EACjD;EACA,iBAAiB;EACjB,OAAO;EACP;EACA;EACA,SAAS;IACP;EACF,UAAU;EACV,aAAa;EACb,YAAY;;;;iBCrQQ,wBACpB,eACC,QAAQ,MAAM;iBA2BD,0BACd,SAAS,+BACT,eAAe,2BACf;EAAW;EAAe;IACzB,MAAM;iBAwBO,6BACd,SAAS,+BACT,eAAe,4BACd;iBAyCa,+BACd,SAAS,+BACT,iBAAiB,2BACjB,mBAAmB,2BACnB,eACC;iBAcmB,8BAA8B;EAClD,SAAS;EACT;EACA;EACA;EACA;EACA;EACA;IACE,QAAQ;;;;;;;;KCnIA,yBAAyB,KAAK;;;;;;KAS9B;WAEG;;WAEA;;;;;;WAMA;;WAGA;;WAEA;;;WAIA;WACA,WAAW;;;WAIX;WACA,WAAW;;UAKT;;WAEN;;WAEA;;WAEA;;WAEA,eAAe;;UAGT;;;;;WAKN;;UAKM;EACf,UAAU;;EAEV;;UAGe,eAAe;;EAE9B;EACA,eAAe;EACf,OAAO;EACP;EACA;;EAEA;EACA,SAAS;;;UAIM,sBAAsB;WAC5B;EACT,kBAAkB;;WAET,cAAc,SAAS;;;;;EAKhC,OAAO,iBAAiB,gBAAgB;;EAExC,QAAQ,iBAAiB,gBAAgB;EACzC,WAAW,MAAM,eAAe,QAAQ,QAAQ;;;UAIjC,uBAAuB;WAC7B;EACT,kBAAkB;WACT,cAAc,SAAS;;WAEvB;;WAEA;;EAET,YAAY;EACZ,WAAW,MAAM,eAAe,QAAQ,QAAQ;;EAEhD,gBAAgB;IACd;IACA,mBAAmB;IACnB,OAAO;IACP,SAAS;MACP;;;UAIW,0BAA0B,aAAa;;EAEtD,KAAK,SAAS,uBAAuB;IACnC,iBAAiB;IACjB;;;EAGF,OAAO;IACL;IACA,iBAAiB;IACjB,OAAO;IACP;IACA;IACA,SAAS;MACP,QAAQ;;EAEZ,aAAa,sBAAsB,gBAAgB;;;UAIpC,4BAA4B,uBAAuB;WACzD;;WAEA;EACT,kBAAkB;WACT,cAAc,SAAS;;WAEvB,qBAAqB,SAAS;;WAE9B;;EAET,qBAAqB,8BAA8B;;EAEnD,MAAM;IACJ;IACA,mBAAmB;IACnB;IACA,OAAO;IACP,SAAS;MACP;IAAU,UAAU;IAAkB,aAAa;IAAe;;;EAEtE,YAAY,0BAA0B,aAAa;;EAEnD,mBACE,QAAQ,oCACP,uBAAuB;;KAGhB,uBAAuB,MAAM,uBAAuB,oBAC5D,sBAAsB,QACtB,uBAAuB,QACvB,4BAA4B,aAAa;UAI5B,kBAAkB,gBAAgB,uBAAuB;;WAE/D;WACA;WACA;;WAEA;WACA,SAAS;;WAET;;WAEA;WACA,YAAY;WACZ,eAAe,cAAc;;;;;;WAM7B,gBAAgB,SAAS;WACzB,QAAQ;WACR,QAAQ;;;;;;WAMR;WACA,SAAS,uBAAuB,MAAM,aAAa;;;;;;;;;cAUjD,mCAAmC;;;;;;;iBAUhC,gCAAgC,MAAM,aAAa,QACjE,YAAY,kBAAkB,MAAM,aAAa;;UAwClC;WACN;WACA,gBAAgB;WAChB,iBAAiB,YAAY;WAC7B;WACA;WACA;;WAEA;;WAEA;;UAGM;WACN;;WAEA;;WAEA,sBAAsB;WACtB,sBAAsB;;;;;;;;;iBAUjB,6BACd,aAAa,cAAc,gDAC1B;;;;;;;cCzUU;;;;cAKA;cAMA;;iBA8GG,4BAA4B,SAAS;;iBASrC,+BAA+B,SAAS;;;;iBAcxC,8BAA8B,SAAS;;;;;;;;;;;;;UC5EtC;;EAEf;;EAEA;;EAEA;;;iBAIc,8BACd,SAAS,+BACT,MAAM,6BACL,kBAAkB;;iBAsEL,kCACd,SAAS,+BACT,QAAQ,oCACP,uBAAuB;cAwhBpB,yBAAuB,EAAA;;;;;;;;;;;;;GASlB,EAAA,KAAA;cACL,gCAA8B,EAAA;;;;;;;;;;;;;;;;;;;GAsChC,EAAA,KAAA;cACE,uBAAqB,EAAA;;;;;;;;;;;;KA2BtB,yBAAyB,EAAE,aAAa;EAC3C,WAAW,EAAE,aAAa;;KAEvB,2BAA2B,EAAE,aAAa;KACnC,iCAAiC,yBAAyB;;;;;;;;;;;;;UCjtBrD;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA,cAAc;;;iBAIA,2BACd,SAAS,+BACT,MAAM,0BACL,kBAAkB,mBAAmB;;iBA+GxB,+BACd,SAAS,+BACT,QAAQ,oCACP,uBAAuB;;iBAwBV,gBAAgB,QAAQ,oCAAoC;;;UCtI3D;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA,UAAU;;EAEV,YAAY;;;;;;iBA8EE;UAIC;;EAEf;;EAEA;;;;;;iBAOc,gCACd,MAAM,+BACL,kBAAkB;;iBAyDL,2BACd,SAAS,8BACR,uBAAuB;;;;;;;;;;cC5OpB;WACQ;WACC;WACK;WACd;WACkB;WACJ;WACO;WACD;WACJ;WACK;WACd;WACJ;WACM;WACC;WACG;WACH;WACI;WACT;;KAGC,uCAAuC;UAgBlC;EACf,QAAQ;EACR;;EAEA;;EAEA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;;iBASK,sBACd,QAAQ,wBACR;EACE;EACA;EACA;EACA;EACA;IAED;;;UC3Fc;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;;iBAGK,8BACd,QAAQ,yBACP;iBAca,mCAAmC,SAAS;;;UCpC3C;EACf;EACA;EACA;IACE,SAAS;IACT;IACA;IACA;IACA;IACA,YAAY;MAAQ;MAAiB;MAAsB;;IAC3D,uBAAuB;IACvB,0BAA0B;IAC1B;MACE;MACA;MACA;MACA,QAAQ;;IAEV;MACE;MACA;;MAEA;MACA;;MAEA;MACA;MACA;MACA;MACA;MACA;MACA;MACA,UAAU;MACV;QACE;QACA;QACA;QACA;QACA;QACA;QACA;QACA;;MAEF;QACE;QACA;QACA;;MAEF;MACA;MACA;;MAEA;MACA;MACA;;;EAGJ,QAAQ;EACR,aAAa;EACb,uBAAuB;EACvB,qBAAqB;;UAGN;EACf;EACA;EACA;EACA;IACE;IACA;IACA;;;UAIa;EACf;IACE,SAAS;IACT;IACA;IACA;MACE;MACA;MACA;MACA;MACA;MACA;MACA;MACA;MACA,SAAS;MACT;QACE;QACA;QACA;QACA;QACA;QACA;QACA;QACA;;MAEF;QACE;QACA;QACA;;;IAGJ;IACA;IACA;IACA;;IAEA;IACA;IACA;IACA;;IAEA;IACA;IACA;IACA;;EAEF;IACE;IACA;IACA;IACA,YAAY;MAAQ;MAAiB;MAAsB;;IAC3D;IACA;;;UAIa;EACf;EACA;EACA;EACA;EACA,UAAU;;UAGK;EACf;EACA;EACA;EACA;IACE;IACA;IACA;IACA;;IAEA;;EAEF;EACA;IACE;IACA;IACA;;EAEF;IACE;IACA;IACA;IACA;IACA;IACA;;;UAIa;EACf;EACA;EACA;EACA,aAAa;EACb;;cAGW;cACA;cACA;cACA;;;iBC5KS,6BACpB,eACC,QAAQ;;;UCuEM;EACf,uBACE,SAAS,+BACT,QAAQ,wCACL,uBAAuB;EAC5B,2BACE,mBACA;IACE;IACA,aAAa,SAAS,OAAO;QAE5B,QAAQ;;;;;;;;;KAUH;;UAGK;EACf;EACA;;UAGe;EACf,SAAS;EACT,SAAS;EACT;EACA;EACA;EACA;EACA;EACA;EACA,OAAO;EACP;EACA;EACA;EACA;;EAEA;EACA;EACA;;EAEA;;EAEA,QAAQ;EACR;EACA;;iBAKoB,2BACpB,yBACA,MAAK,OAAO,YACZ,eAAc,sCACb;;;cCzJU;cAEA;cAEA;cAOA;cAYA;cAGA;cAGA;
|
|
1
|
+
{"version":3,"file":"index.d.ts","names":[],"sources":["../../src/analyst/adapters.ts","../../src/analyst/benchmark-agentrx-calibration.ts","../../src/analyst/benchmark-dataset-types.ts","../../src/analyst/benchmark-dataset-agentrx.ts","../../src/analyst/benchmark-dataset-codetrace.ts","../../src/analyst/benchmark-dataset-utils.ts","../../src/analyst/benchmark-verification-outcome.ts","../../src/analyst/benchmark-verification-artifacts.ts","../../src/analyst/benchmark-public-types.ts","../../src/analyst/benchmark-public-adapters.ts","../../src/analyst/benchmark-public-data.ts","../../src/analyst/definition.ts","../../src/analyst/benchmark-public-prompt.ts","../../src/analyst/benchmark-public-model.ts","../../src/analyst/benchmark-public-rlm.ts","../../src/analyst/benchmark-runner-prime.ts","../../src/analyst/benchmark-comparison.ts","../../src/analyst/benchmark-public-calibration.ts","../../src/analyst/benchmark-command-artifact.ts","../../src/analyst/benchmark-command-result.ts","../../src/analyst/benchmark-command.ts","../../src/analyst/benchmark-implementation.ts","../../src/analyst/benchmark-instructions-override.ts","../../src/analyst/benchmark-report.ts","../../src/analyst/benchmark-summary.ts","../../src/analyst/bind.ts","../../src/analyst/define.ts","../../src/analyst/kinds/skill-usage.ts","../../src/analyst/parse-tolerant.ts","../../src/analyst/proposal-findings.ts"],"mappings":";;;;;;;;;;;;;;UA8CiB;EACf;EACA;;EAEA,SAAS,KAAK;;EAEd;;iBAGc,kCACd,MAAM,kCACL,QAAQ;;;cCtDE;UAEI;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA,SAAS;;iBAGK,4BACd,QAAQ,wBACR,2BACC;iBAkBa,iCAAiC,SAAS;;;KCtD9C;UAEK;EACf,YAAY;EACZ;EACA;EACA;EACA;EACA;;UAGe;EACf,eAAe;EACf,mBAAmB;EACnB;IAAe,YAAY;IAAY;;EACvC,wBAAwB;EACxB;EACA;EACA;;UAGe;EACf,UAAU;EACV;EACA;EACA;EACA;;UAGe;EACf,UAAU;EACV,mBAAmB;EACnB;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA;EACA,iBAAiB;;KAGP,0CAEC,sCACA,iCACA;UAEI;EACf;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA,oCAAoC;;UAGrB;EACf,eAAe;EACf,WAAW,sBAAsB;;KAGvB;UAEK;;;;;EAKf,WAAW;;UAGI,kCACP,yBACN;UAEa,oCAAoC;EACnD;;EAEA;;UAGe,yCAAyC;EACxD;EACA;EACA;EACA;;UAGe,2CACP,kCACN;;;iBCtGY,qBAAqB,QACnC,KAAK,YACL,OAAO,QACP,UAAS,8BACR,qBAAqB;;iBAsGR,6BACd,mBAAmB,YACnB,iBACA,UAAS,mCACR;iBA6Ea,yBAAyB;;iBAoNzB,iBAAiB;;;iBC/YjB,mBAAmB,QACjC,KAAK,mBACL,OAAO,QACP,UAAS,4BACR,qBAAqB;;iBAwFR,gCACd,2BACA,aAAa,uBACb,UAAS,qCACR;;;iBClHa,wBAAwB;;;KCA5B;UAEK;EACf;EACA;EACA,QAAQ;;UAGO;EACf,QAAQ;EACR;EAKA;IAAe;IAAe;;EAC9B,SAAS;EACT;EACA;EACA;EACA;;UAGe;EACf;EACA;;iBA8Ec,yBACd,gBAAgB,2BACf;;;cC/FU;KAED;UAEK;EACf,MAAM;EACN;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;EACT;EACA;EACA;EACA;EACA;EACA,OAAO;EACP,cAAc;EACd,UAAU,OAAO;;UAGF;EACf,UAAU;EACV,SAAS;EACT,OAAO,MAAM;IAA6B;;;iBAatB,mCAAmC;EACvD;EACA,KAAK;EACL;IACE,QAAQ;iBAwHI,kCACd,kBACA,iBACA,WAAW,6BACX;;;KC7KU;;;;;;;;UASK;;WAEN;;WAEA;;;UAIM;EACf,MAAM;EACN;EACA,kBAAkB,aAAa;;EAE/B,UAAU;;UAGK;;EAEf,MAAM;;EAEN;;EAEA,kBAAkB,aAAa;EAC/B;EACA;EACA;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA,UAAU;;EAEV;;EAEA,uBAAuB;EACvB;IACE,SAAS;IACT;IACA;IACA;IACA;IACA;IACA;IACA;IACA;;;;;;IAMA;;EAEF,aAAa;EACb;IACE;IACA;;;;;;;;;KAUQ,sCAAsC,KAChD,iEAGA,QAAQ,KAAK;UAEE;EACf,OAAO,qBAAqB;EAC5B;EACA;EACA;EACA,YAAY;IACV;IACA;IACA;;EAEF,uBAAuB;EACvB,WAAW;;UAGI;EACf;EACA;EACA,QAAQ;;UAGO;EACf,OAAO;EACP,OAAO;EACP,OAAO;EACP,YAAY;EACZ,QAAQ;;UAGO;EACf;EACA;EACA;EACA;EACA;EACA;EACA,QAAQ;EACR,UAAU;;;;;;;;;;UChHK;EACf;EACA;;EAEA;EACA;EACA,UAAU;EACV;EACA;EACA;EACA;EACA,WAAW;;;;;;;;;UAUI;EACf;EACA;;EAEA,kCAAkC;;EAElC;;EAEA;;EAEA;;EAEA;;EAEA;;;;;;;;UASe;EACf;EACA,OAAO;;iBAGO,8BAA8B,uBAAuB;iBAiB/C,6BAA6B;EACjD,SAAS;EACT;EACA,mBAAmB;EACnB;EACA,OAAO;EACP,SAAS;IACP;EACF,UAAU;EACV,aAAa;;EAEb,aAAa;;;;;;;;;iBA4MO,6BAA6B;EACjD;EACA,iBAAiB;EACjB,OAAO;EACP;EACA;EACA,SAAS;IACP;EACF,UAAU;EACV,aAAa;EACb,YAAY;;;;iBCrQQ,wBACpB,eACC,QAAQ,MAAM;iBA2BD,0BACd,SAAS,+BACT,eAAe,2BACf;EAAW;EAAe;IACzB,MAAM;iBAwBO,6BACd,SAAS,+BACT,eAAe,4BACd;iBAyCa,+BACd,SAAS,+BACT,iBAAiB,2BACjB,mBAAmB,2BACnB,eACC;iBAcmB,8BAA8B;EAClD,SAAS;EACT;EACA;EACA;EACA;EACA;EACA;IACE,QAAQ;;;;;;;;KCnIA,yBAAyB,KAAK;;;;;;KAS9B;WAEG;;WAEA;;;;;;WAMA;;WAGA;;WAEA;;;WAIA;WACA,WAAW;;;WAIX;WACA,WAAW;;UAKT;;WAEN;;WAEA;;WAEA;;WAEA,eAAe;;UAGT;;;;;WAKN;;UAKM;EACf,UAAU;;EAEV;;UAGe,eAAe;;EAE9B;EACA,eAAe;EACf,OAAO;EACP;EACA;;EAEA;EACA,SAAS;;;UAIM,sBAAsB;WAC5B;EACT,kBAAkB;;WAET,cAAc,SAAS;;;;;EAKhC,OAAO,iBAAiB,gBAAgB;;EAExC,QAAQ,iBAAiB,gBAAgB;EACzC,WAAW,MAAM,eAAe,QAAQ,QAAQ;;;UAIjC,uBAAuB;WAC7B;EACT,kBAAkB;WACT,cAAc,SAAS;;WAEvB;;WAEA;;EAET,YAAY;EACZ,WAAW,MAAM,eAAe,QAAQ,QAAQ;;EAEhD,gBAAgB;IACd;IACA,mBAAmB;IACnB,OAAO;IACP,SAAS;MACP;;;UAIW,0BAA0B,aAAa;;EAEtD,KAAK,SAAS,uBAAuB;IACnC,iBAAiB;IACjB;;;EAGF,OAAO;IACL;IACA,iBAAiB;IACjB,OAAO;IACP;IACA;IACA,SAAS;MACP,QAAQ;;EAEZ,aAAa,sBAAsB,gBAAgB;;;UAIpC,4BAA4B,uBAAuB;WACzD;;WAEA;EACT,kBAAkB;WACT,cAAc,SAAS;;WAEvB,qBAAqB,SAAS;;WAE9B;;EAET,qBAAqB,8BAA8B;;EAEnD,MAAM;IACJ;IACA,mBAAmB;IACnB;IACA,OAAO;IACP,SAAS;MACP;IAAU,UAAU;IAAkB,aAAa;IAAe;;;EAEtE,YAAY,0BAA0B,aAAa;;EAEnD,mBACE,QAAQ,oCACP,uBAAuB;;KAGhB,uBAAuB,MAAM,uBAAuB,oBAC5D,sBAAsB,QACtB,uBAAuB,QACvB,4BAA4B,aAAa;UAI5B,kBAAkB,gBAAgB,uBAAuB;;WAE/D;WACA;WACA;;WAEA;WACA,SAAS;;WAET;;WAEA;WACA,YAAY;WACZ,eAAe,cAAc;;;;;;WAM7B,gBAAgB,SAAS;WACzB,QAAQ;WACR,QAAQ;;;;;;WAMR;WACA,SAAS,uBAAuB,MAAM,aAAa;;;;;;;;;cAUjD,mCAAmC;;;;;;;iBAUhC,gCAAgC,MAAM,aAAa,QACjE,YAAY,kBAAkB,MAAM,aAAa;;UAwClC;WACN;WACA,gBAAgB;WAChB,iBAAiB,YAAY;WAC7B;WACA;WACA;;WAEA;;WAEA;;UAGM;WACN;;WAEA;;WAEA,sBAAsB;WACtB,sBAAsB;;;;;;;;;iBAUjB,6BACd,aAAa,cAAc,gDAC1B;;;;;;;cCzUU;;;;cAKA;cAMA;;iBA8GG,4BAA4B,SAAS;;iBASrC,+BAA+B,SAAS;;;;iBAcxC,8BAA8B,SAAS;;;;;;;;;;;;;UC5EtC;;EAEf;;EAEA;;EAEA;;;iBAIc,8BACd,SAAS,+BACT,MAAM,6BACL,kBAAkB;;iBAsEL,kCACd,SAAS,+BACT,QAAQ,oCACP,uBAAuB;cAwhBpB,yBAAuB,EAAA;;;;;;;;;;;;;GASlB,EAAA,KAAA;cACL,gCAA8B,EAAA;;;;;;;;;;;;;;;;;;;GAsChC,EAAA,KAAA;cACE,uBAAqB,EAAA;;;;;;;;;;;;KA2BtB,yBAAyB,EAAE,aAAa;EAC3C,WAAW,EAAE,aAAa;;KAEvB,2BAA2B,EAAE,aAAa;KACnC,iCAAiC,yBAAyB;;;;;;;;;;;;;UCjtBrD;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA;;EAEA,cAAc;;;iBAIA,2BACd,SAAS,+BACT,MAAM,0BACL,kBAAkB,mBAAmB;;iBA+GxB,+BACd,SAAS,+BACT,QAAQ,oCACP,uBAAuB;;iBAwBV,gBAAgB,QAAQ,oCAAoC;;;UCtI3D;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA,UAAU;;EAEV,YAAY;;;;;;iBA8EE;UAIC;;EAEf;;EAEA;;;;;;iBAOc,gCACd,MAAM,+BACL,kBAAkB;;iBAyDL,2BACd,SAAS,8BACR,uBAAuB;;;;;;;;;;cC5OpB;WACQ;WACC;WACK;WACd;WACkB;WACJ;WACO;WACD;WACJ;WACK;WACd;WACJ;WACM;WACC;WACG;WACH;WACI;WACT;;KAGC,uCAAuC;UAgBlC;EACf,QAAQ;EACR;;EAEA;;EAEA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;;iBASK,sBACd,QAAQ,wBACR;EACE;EACA;EACA;EACA;EACA;IAED;;;UC3Fc;EACf;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;EAEA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;EACA;;UAGe;EACf;EACA;EACA,SAAS;;iBAGK,8BACd,QAAQ,yBACP;iBAca,mCAAmC,SAAS;;;UCpC3C;EACf;EACA;EACA;IACE,SAAS;IACT;IACA;IACA;IACA;IACA,YAAY;MAAQ;MAAiB;MAAsB;;IAC3D,uBAAuB;IACvB,0BAA0B;IAC1B;MACE;MACA;MACA;MACA,QAAQ;;IAEV;MACE;MACA;;MAEA;MACA;;MAEA;MACA;MACA;MACA;MACA;MACA;MACA;MACA,UAAU;MACV;QACE;QACA;QACA;QACA;QACA;QACA;QACA;QACA;;MAEF;QACE;QACA;QACA;;MAEF;MACA;MACA;;MAEA;MACA;MACA;;;EAGJ,QAAQ;EACR,aAAa;EACb,uBAAuB;EACvB,qBAAqB;;UAGN;EACf;EACA;EACA;EACA;IACE;IACA;IACA;;;UAIa;EACf;IACE,SAAS;IACT;IACA;IACA;MACE;MACA;MACA;MACA;MACA;MACA;MACA;MACA;MACA,SAAS;MACT;QACE;QACA;QACA;QACA;QACA;QACA;QACA;QACA;;MAEF;QACE;QACA;QACA;;;IAGJ;IACA;IACA;IACA;;IAEA;IACA;IACA;IACA;;IAEA;IACA;IACA;IACA;;EAEF;IACE;IACA;IACA;IACA,YAAY;MAAQ;MAAiB;MAAsB;;IAC3D;IACA;;;UAIa;EACf;EACA;EACA;EACA;EACA,UAAU;;UAGK;EACf;EACA;EACA;EACA;IACE;IACA;IACA;IACA;;IAEA;;EAEF;EACA;IACE;IACA;IACA;;EAEF;IACE;IACA;IACA;IACA;IACA;IACA;;;UAIa;EACf;EACA;EACA;EACA,aAAa;EACb;;cAGW;cACA;cACA;cACA;;;iBC5KS,6BACpB,eACC,QAAQ;;;UCuEM;EACf,uBACE,SAAS,+BACT,QAAQ,wCACL,uBAAuB;EAC5B,2BACE,mBACA;IACE;IACA,aAAa,SAAS,OAAO;QAE5B,QAAQ;;;;;;;;;KAUH;;UAGK;EACf;EACA;;UAGe;EACf,SAAS;EACT,SAAS;EACT;EACA;EACA;EACA;EACA;EACA;EACA,OAAO;EACP;EACA;EACA;EACA;;EAEA;EACA;EACA;;EAEA;;EAEA,QAAQ;EACR;EACA;;iBAKoB,2BACpB,yBACA,MAAK,OAAO,YACZ,eAAc,sCACb;;;cCzJU;cAEA;cAEA;cAOA;cAYA;cAGA;cAGA;cAgHA;iBAGG;iBAIA;;;;iBC3IA,oCAAoC,eAAe;;iBAQnD,gCAAgC,eAAe;;;;;;;;;;;iBAyB/C,+BACd,SAAS,+BACT,WAAW,KAAK;;;iBCzCF,+BACd,QAAQ,wBACR,uBAAsB;;;iBCIR,gCACd,kBACA,uBAAuB,gCACtB;;;;KCcS;;WAGG;WACA;WACA;WACA,YAAY;WACZ,UAAU;;;WAIV;WACA,QAAQ;;;;;;iBAOP,YAAY,MAAM,aAAa,QAC7C,YAAY,kBAAkB,MAAM,aAAa,SACjD,YAAY,0BACX,uBAAuB;;;;;;;;;;iBCpCV,mBAAmB,YAAY,yBAAyB;UAkBvD;EACf;EACA;EACA;EACA,MAAM;EACN,SAAS,QAAQ;;UAGF,wCAAwC;;EAEvD,iBAAiB,SAAS;;;iBAIZ,oBACd,SAAS,kCACR,oBAAoB;iBACP,oBACd,SAAS,6BACR,QAAQ;;;KCpBC;;UAGK;EACf;EACA,MAAM;;EAEN;EACA;;EAEA;;EAEA;;;EAGA;;EAEA;;EAEA;EACA;EACA;;EAEA;;EAEA;;UAGe;EACf;EACA,SAAS;;UAGM;;EAEf;;EAEA;IAAc;IAAc,MAAM;;;EAElC;;;EAGA,kBAAkB;;EAElB;;;iBA2Ec,sBAAsB,QAAQ,uBAAuB;;iBAiHrD,uBACd,QAAQ,kBACR,qBACC;cA2HG,6BAA6B,oBAAoB;WAC5C;WACA;WAEA;WACA;IAAS;IAAgC;;WACzC;WACA;aACP;aACA;aACA;;EAGI,QAAQ,OAAO,kBAAkB,KAAK,iBAAiB,QAAQ;;cAS1D,qBAAmB;;;;;;;;;;;;;;;iBC5YhB,gBAAgB;;;;;iBAgBhB,WAAW;;;;iBCIX,kBAAkB,mBAAmB,WAAW;;;;;;iBAShD,uBACd,mBACA,mBACC,cAAc"}
|
package/dist/analyst/index.js
CHANGED
|
@@ -1,14 +1,14 @@
|
|
|
1
1
|
import { i as CostLedger } from "../cost-ledger-_IbkejCF.js";
|
|
2
2
|
import { n as assertProposalFindings, r as isProposalFinding } from "../abort-signal-CtzAM_sJ.js";
|
|
3
3
|
import { i as makeProposalFinding, n as computeFindingId, r as makeFinding } from "../types-C8VBbiw6.js";
|
|
4
|
-
import { B as
|
|
4
|
+
import { B as KIND_EXPECTED_SUBJECTS, F as parseRawFinding, G as resolveTraceAnalystLimits, H as parseFindingSubject, I as coerceJson, L as stripCodeFences, M as RAW_FINDING_SCHEMA_PROMPT, N as RawAnalystFindingSchema, P as evidenceRefsFromRawFinding, R as FINDING_SUBJECT_KINDS, U as renderFindingSubject, V as findingSubjectGrammarPromptFor, W as DEFAULT_TRACE_ANALYST_OUTPUT_TOKENS, i as buildTraceToolsForGroup, n as renderPriorFindings, r as runTraceAnalyst, t as createTraceAnalyst, z as FINDING_SUBJECT_SYNTAX } from "../kind-factory-g1ROw67A.js";
|
|
5
5
|
import { i as validateUsageSettlementTimeout, n as settleUsageReceiptFromCostLedger } from "../usage-receipt-Cx3xNeyB.js";
|
|
6
|
-
import { a as KNOWLEDGE_POISONING_KIND_SPEC, c as FAILURE_MODE_KIND_SPEC, d as emitControlIntegrityFindings, f as behavioralAnalyst, i as DEFAULT_TRACE_ANALYST_KINDS, l as CONTROL_INTEGRITY_ANALYST, n as AnalystRegistry, o as KNOWLEDGE_GAP_KIND_SPEC, p as deriveEfficiencyFindings, r as ExactAnalystRunExecutionError, s as IMPROVEMENT_KIND_SPEC, t as buildDefaultAnalystRegistry, u as ControlIntegrityAnalyst } from "../default-registry-
|
|
7
|
-
import { t as createDspyRlmTraceEngine } from "../dspy-rlm-engine-
|
|
8
|
-
import { a as diffFindings, i as defaultIsMaterial, n as runSemanticConceptJudge, o as createChatTraceEngine, r as FindingsStore, t as SEMANTIC_CONCEPT_JUDGE_VERSION } from "../semantic-concept-judge-
|
|
9
|
-
import { t as createChatClient } from "../chat-client-
|
|
6
|
+
import { a as KNOWLEDGE_POISONING_KIND_SPEC, c as FAILURE_MODE_KIND_SPEC, d as emitControlIntegrityFindings, f as behavioralAnalyst, i as DEFAULT_TRACE_ANALYST_KINDS, l as CONTROL_INTEGRITY_ANALYST, n as AnalystRegistry, o as KNOWLEDGE_GAP_KIND_SPEC, p as deriveEfficiencyFindings, r as ExactAnalystRunExecutionError, s as IMPROVEMENT_KIND_SPEC, t as buildDefaultAnalystRegistry, u as ControlIntegrityAnalyst } from "../default-registry-BtZ_t-9b.js";
|
|
7
|
+
import { t as createDspyRlmTraceEngine } from "../dspy-rlm-engine-sPX7B5ac.js";
|
|
8
|
+
import { a as diffFindings, i as defaultIsMaterial, n as runSemanticConceptJudge, o as createChatTraceEngine, r as FindingsStore, t as SEMANTIC_CONCEPT_JUDGE_VERSION } from "../semantic-concept-judge-70LxrCSL.js";
|
|
9
|
+
import { t as createChatClient } from "../chat-client-C4dg7Ng_.js";
|
|
10
10
|
import { a as scoreAnalystFindings, i as summarizeAnalystBenchmarkRunner, n as runAnalystBenchmark, r as traceStoreEvidenceResolver, t as registryBenchmarkRunner } from "../benchmark-BeQJyrRC.js";
|
|
11
|
-
import { $ as ANALYST_BENCHMARK_MANIFEST_FILE, A as readAnalystInstructionsOverride, B as parseVerificationOutcome, C as emptyPublicBenchmarkRunner, D as summarizeCodeTraceCalibration, E as renderCodeTraceCalibrationMarkdown, F as publicBenchmarkRlmInstructions, G as ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, H as ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, I as publicBenchmarkSystemPrompt, J as ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, K as ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, L as DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, M as MAX_INCORRECT_BLOCKS, N as MAX_INCORRECT_BLOCK_STEPS, O as analystInstructionsOverrideFromText, P as publicBenchmarkProtocolSha256, Q as ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, R as appendVerificationArtifactsToOtlp, S as adaptPublicBenchmarkFindings, T as readAnalystBenchmarkArtifact, U as ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, V as ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, W as ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, X as analystBenchmarkImplementationDigest, Y as analystBenchmarkDependencyLockDigest, Z as ANALYST_BENCHMARK_COST_LEDGER_FILE, _ as publicDirectAnalystDefinition, a as primeCodeTraceAnalystDefinition, at as codeTraceBenchCase, b as analystDefinitionAsymmetries, c as preparePublicAnalystBenchmark, ct as agentRxPredictionsToFindings, d as selectPublicBenchmarkRows, dt as normalizeBenchmarkLabel, et as ANALYST_BENCHMARK_OBSERVATIONS_FILE, f as createPublicBenchmarkRlmRunner, g as createPublicBenchmarkDirectRunner, h as runReplVariableAnalystDefinition, i as primeAnalystProtocolSha256, it as summarizeAgentRxCalibration, j as CODE_TRACE_BENCH_ANALYST_PROMPT, k as effectiveAnalystProtocolSha256, l as publicBenchmarkDistributions, lt as normalizeAgentRxCategory, m as rlmEngineLimits, n as renderAnalystBenchmarkMarkdown, nt as AGENT_RX_UPSTREAM_REVISION, o as runInlineAnalystDefinition, ot as codeTracerPredictionsToFindings, p as publicRlmAnalystDefinition, q as ANALYST_BENCHMARK_IMPLEMENTATION_FILES, r as createPrimeBenchmarkRunner, rt as renderAgentRxCalibrationMarkdown, s as loadPublicBenchmarkRows, st as agentRxBenchmarkCase, t as runAnalystBenchmarkCommand, tt as compareAnalystRunners, u as publicBenchmarkSelectionReport, ut as roundAgentRxStep, v as runChunkedAnalystDefinition, w as expandCodeTraceFailureBlocks, x as analystDefinitionProtocolSha256, y as AnalystExpressivenessError, z as loadCodeTraceVerificationArtifacts } from "../benchmark-command-
|
|
11
|
+
import { $ as ANALYST_BENCHMARK_MANIFEST_FILE, A as readAnalystInstructionsOverride, B as parseVerificationOutcome, C as emptyPublicBenchmarkRunner, D as summarizeCodeTraceCalibration, E as renderCodeTraceCalibrationMarkdown, F as publicBenchmarkRlmInstructions, G as ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256, H as ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES, I as publicBenchmarkSystemPrompt, J as ANALYST_BENCHMARK_IMPLEMENTATION_SHA256, K as ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM, L as DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES, M as MAX_INCORRECT_BLOCKS, N as MAX_INCORRECT_BLOCK_STEPS, O as analystInstructionsOverrideFromText, P as publicBenchmarkProtocolSha256, Q as ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE, R as appendVerificationArtifactsToOtlp, S as adaptPublicBenchmarkFindings, T as readAnalystBenchmarkArtifact, U as ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256, V as ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM, W as ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256, X as analystBenchmarkImplementationDigest, Y as analystBenchmarkDependencyLockDigest, Z as ANALYST_BENCHMARK_COST_LEDGER_FILE, _ as publicDirectAnalystDefinition, a as primeCodeTraceAnalystDefinition, at as codeTraceBenchCase, b as analystDefinitionAsymmetries, c as preparePublicAnalystBenchmark, ct as agentRxPredictionsToFindings, d as selectPublicBenchmarkRows, dt as normalizeBenchmarkLabel, et as ANALYST_BENCHMARK_OBSERVATIONS_FILE, f as createPublicBenchmarkRlmRunner, g as createPublicBenchmarkDirectRunner, h as runReplVariableAnalystDefinition, i as primeAnalystProtocolSha256, it as summarizeAgentRxCalibration, j as CODE_TRACE_BENCH_ANALYST_PROMPT, k as effectiveAnalystProtocolSha256, l as publicBenchmarkDistributions, lt as normalizeAgentRxCategory, m as rlmEngineLimits, n as renderAnalystBenchmarkMarkdown, nt as AGENT_RX_UPSTREAM_REVISION, o as runInlineAnalystDefinition, ot as codeTracerPredictionsToFindings, p as publicRlmAnalystDefinition, q as ANALYST_BENCHMARK_IMPLEMENTATION_FILES, r as createPrimeBenchmarkRunner, rt as renderAgentRxCalibrationMarkdown, s as loadPublicBenchmarkRows, st as agentRxBenchmarkCase, t as runAnalystBenchmarkCommand, tt as compareAnalystRunners, u as publicBenchmarkSelectionReport, ut as roundAgentRxStep, v as runChunkedAnalystDefinition, w as expandCodeTraceFailureBlocks, x as analystDefinitionProtocolSha256, y as AnalystExpressivenessError, z as loadCodeTraceVerificationArtifacts } from "../benchmark-command-CdUGku3W.js";
|
|
12
12
|
import { a as extractPrimeJsonObject, c as primeProtocolSha256, d as runPrimeExchange, f as decodeReplyRows, i as emptyPrimeRawUsage, l as primeReplyDefect, n as buildPrimePrompt, o as mergePrimeRawUsage, r as buildPrimeRepairPrompt, s as normalizePrimeUsage, t as analystUsageReceiptFromPrimeUsage, u as projectPrimeTrajectory } from "../prime-protocol-DcAK9pC8.js";
|
|
13
13
|
import { t as nodeHttpPrimeBridgeTransport } from "../prime-bridge-transport-Z3cpKGoV.js";
|
|
14
14
|
import { existsSync, readFileSync, readdirSync, statSync } from "node:fs";
|
|
@@ -1,8 +1,8 @@
|
|
|
1
|
-
import { a as RunRecord } from "./run-record-
|
|
1
|
+
import { a as RunRecord } from "./run-record-xEULpgML.js";
|
|
2
2
|
import { r as DatasetScenario } from "./dataset-DQqhOCPt.js";
|
|
3
|
-
import "./summary-report-
|
|
4
|
-
import { o as InsightReport, r as ExecutionInsight, t as CostProvenanceSummary } from "./insight-report-
|
|
5
|
-
import { n as AnalystRegistry } from "./registry-
|
|
3
|
+
import "./summary-report-SNvJoAvD.js";
|
|
4
|
+
import { o as InsightReport, r as ExecutionInsight, t as CostProvenanceSummary } from "./insight-report-DeE1solA.js";
|
|
5
|
+
import { n as AnalystRegistry } from "./registry-BM2Kdk8T.js";
|
|
6
6
|
//#region src/contract/analyze-runs.d.ts
|
|
7
7
|
interface AnalyzeRunsOptions {
|
|
8
8
|
/** The runs to analyze. */
|
|
@@ -72,4 +72,4 @@ declare function summarizeExecution(opts: SummarizeExecutionOptions): ExecutionR
|
|
|
72
72
|
declare function analyzeRuns(opts: AnalyzeRunsOptions): Promise<InsightReport>;
|
|
73
73
|
//#endregion
|
|
74
74
|
export { summarizeExecution as a, analyzeRuns as i, ExecutionReport as n, SummarizeExecutionOptions as r, AnalyzeRunsOptions as t };
|
|
75
|
-
//# sourceMappingURL=analyze-runs-
|
|
75
|
+
//# sourceMappingURL=analyze-runs-CLz5O01x.d.ts.map
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"analyze-runs-
|
|
1
|
+
{"version":3,"file":"analyze-runs-CLz5O01x.d.ts","names":[],"sources":["../src/contract/analyze-runs.ts"],"mappings":";;;;;;UAqEiB;;EAEf,MAAM;;;EAGN;;;;;EAKA;EACA;;;EAGA,8BAA8B;;;EAG9B,kBAAkB;;;EAGlB,UAAU;;;;EAIV;IACE;IACA,cAAc;;;;;EAKhB,cAAc;IAAQ;IAAe;IAAe;;;EAEpD;;;;EAIA;;;;;;;;;EASA,eAAe;;;EAGf;;UAGe;EACf,MAAM;EACN;;UAGe;EACf,WAAW;EACX,gBAAgB;;;iBAIF,mBAAmB,MAAM,4BAA4B;iBAgB/C,YAAY,MAAM,qBAAqB,QAAQ"}
|
|
@@ -5,15 +5,16 @@ import { a as resolveModelPricing } from "./metrics-DOMpkoeb.js";
|
|
|
5
5
|
import { a as CostLedgerPersistenceError, i as CostLedger, n as CostCallConflictError, o as CostReceiptCaptureError, r as CostCeilingReachedError, s as CostReservationExceededError, t as CostAccountingIncompleteError } from "./cost-ledger-_IbkejCF.js";
|
|
6
6
|
import { c as withLedgerFileLock, l as writeLedgerFileAtomically } from "./journal-BuRLm6ho.js";
|
|
7
7
|
import "./ledger-core/index.js";
|
|
8
|
-
import { S as fsCampaignStorage, i as startExternalOptimizerModelProxy, r as runWithCleanup, x as createRunCostLedger, y as resolveExternalOptimizerProcessLimits } from "./external-optimizer-subprocess-
|
|
8
|
+
import { S as fsCampaignStorage, i as startExternalOptimizerModelProxy, r as runWithCleanup, x as createRunCostLedger, y as resolveExternalOptimizerProcessLimits } from "./external-optimizer-subprocess-BrfsgL92.js";
|
|
9
|
+
import { f as redactText } from "./redact-DxwlUAvh.js";
|
|
9
10
|
import { r as makeFinding } from "./types-C8VBbiw6.js";
|
|
10
|
-
import { M as
|
|
11
|
+
import { M as RAW_FINDING_SCHEMA_PROMPT, N as RawAnalystFindingSchema, P as evidenceRefsFromRawFinding, g as TRACE_ANALYSIS_LIMITS, r as runTraceAnalyst } from "./kind-factory-g1ROw67A.js";
|
|
11
12
|
import { r as usageReceiptFromCostLedger } from "./usage-receipt-Cx3xNeyB.js";
|
|
12
|
-
import { a as callLlmJson, r as LlmResponseError, t as LlmCallError } from "./llm-client-
|
|
13
|
-
import { t as createDspyRlmTraceEngine } from "./dspy-rlm-engine-
|
|
14
|
-
import { i as otlpTextToTraceAnalysisStore, r as createOtlpBufferTraceStore, t as DEFAULT_MAX_TRACE_FILE_BYTES } from "./store-otlp-
|
|
13
|
+
import { a as callLlmJson, r as LlmResponseError, t as LlmCallError } from "./llm-client-DF9bPNYA.js";
|
|
14
|
+
import { t as createDspyRlmTraceEngine } from "./dspy-rlm-engine-sPX7B5ac.js";
|
|
15
|
+
import { i as otlpTextToTraceAnalysisStore, r as createOtlpBufferTraceStore, t as DEFAULT_MAX_TRACE_FILE_BYTES } from "./store-otlp-DPgzSQKS.js";
|
|
15
16
|
import { i as summarizeAnalystBenchmarkRunner, n as runAnalystBenchmark, r as traceStoreEvidenceResolver } from "./benchmark-BeQJyrRC.js";
|
|
16
|
-
import { n as acquireSingleRunLock } from "./external-optimizer-process-
|
|
17
|
+
import { n as acquireSingleRunLock } from "./external-optimizer-process-D3BLpZyR.js";
|
|
17
18
|
import { t as assertEqualDeclarativeTerms } from "./equal-terms-Blx9Zk5I.js";
|
|
18
19
|
import { c as primeProtocolSha256, d as runPrimeExchange, f as decodeReplyRows, n as buildPrimePrompt, t as analystUsageReceiptFromPrimeUsage, u as projectPrimeTrajectory } from "./prime-protocol-DcAK9pC8.js";
|
|
19
20
|
import { t as nodeHttpPrimeBridgeTransport } from "./prime-bridge-transport-Z3cpKGoV.js";
|
|
@@ -1414,7 +1415,7 @@ const ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES = Object.freeze([
|
|
|
1414
1415
|
"package.json",
|
|
1415
1416
|
"pnpm-lock.yaml"
|
|
1416
1417
|
]);
|
|
1417
|
-
const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "
|
|
1418
|
+
const ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 = "c61fa0edc749ac199316c36caf4ff568681667ba5d8e6b01b22b6f1583ec6523";
|
|
1418
1419
|
const ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256 = "1e03f2daed356d60316aabefb407ec1e437ac94d408d61eea4ae096e9c6fbb5b";
|
|
1419
1420
|
const ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256 = "4dba263b6256a30d56c7fdb2d992d3a953c0035d731f359b704db806f68f75ac";
|
|
1420
1421
|
const ANALYST_BENCHMARK_IMPLEMENTATION_FILES = Object.freeze([
|
|
@@ -1525,9 +1526,10 @@ const ANALYST_BENCHMARK_IMPLEMENTATION_FILES = Object.freeze([
|
|
|
1525
1526
|
"src/trace/attribute-vocabulary.ts",
|
|
1526
1527
|
"src/trace/otlp-attributes.ts",
|
|
1527
1528
|
"src/trace/raw-provider-sink.ts",
|
|
1529
|
+
"src/trace/redact.ts",
|
|
1528
1530
|
"src/verdict-cache.ts"
|
|
1529
1531
|
]);
|
|
1530
|
-
const ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 = "
|
|
1532
|
+
const ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 = "de168cc2cbca1623a88fc0b033bbf92f92d5fb1bd2aa317d8c2c453eb61fd9df";
|
|
1531
1533
|
function analystBenchmarkImplementationDigest() {
|
|
1532
1534
|
return ANALYST_BENCHMARK_IMPLEMENTATION_SHA256;
|
|
1533
1535
|
}
|
|
@@ -3442,9 +3444,7 @@ function publicBenchmarkError(error, secrets = []) {
|
|
|
3442
3444
|
};
|
|
3443
3445
|
}
|
|
3444
3446
|
function redactSensitiveText(value, secrets) {
|
|
3445
|
-
|
|
3446
|
-
for (const secret of secrets) if (secret) redacted = redacted.replaceAll(secret, "[REDACTED]");
|
|
3447
|
-
redacted = redacted.replace(/\bBearer\s+[^\s"',;]+/gi, "Bearer [REDACTED]").replace(/\b(api[_-]?key|access[_-]?token|refresh[_-]?token|password|secret)\b\s*[:=]\s*[^\s"',;]+/gi, "$1=[REDACTED]");
|
|
3447
|
+
const redacted = redactText(value, { knownSecrets: secrets });
|
|
3448
3448
|
if (redacted.length <= 500) return redacted;
|
|
3449
3449
|
const head = redacted.slice(0, 180);
|
|
3450
3450
|
const marker = `...[${redacted.length - 460} chars omitted]...`;
|
|
@@ -6315,4 +6315,4 @@ function shellQuote(value) {
|
|
|
6315
6315
|
//#endregion
|
|
6316
6316
|
export { ANALYST_BENCHMARK_MANIFEST_FILE as $, readAnalystInstructionsOverride as A, parseVerificationOutcome as B, emptyPublicBenchmarkRunner as C, summarizeCodeTraceCalibration as D, renderCodeTraceCalibrationMarkdown as E, publicBenchmarkRlmInstructions as F, ANALYST_BENCHMARK_EVIDENCE_IMPLEMENTATION_SHA256 as G, ANALYST_BENCHMARK_DEPENDENCY_LOCK_FILES as H, publicBenchmarkSystemPrompt as I, ANALYST_BENCHMARK_IMPLEMENTATION_SHA256 as J, ANALYST_BENCHMARK_IMPLEMENTATION_DIGEST_ALGORITHM as K, DEFAULT_MAX_VERIFICATION_ARTIFACT_BYTES as L, MAX_INCORRECT_BLOCKS as M, MAX_INCORRECT_BLOCK_STEPS as N, analystInstructionsOverrideFromText as O, publicBenchmarkProtocolSha256 as P, ANALYST_BENCHMARK_LOCAL_RECEIPT_FILE as Q, appendVerificationArtifactsToOtlp as R, adaptPublicBenchmarkFindings as S, readAnalystBenchmarkArtifact as T, ANALYST_BENCHMARK_DEPENDENCY_LOCK_SHA256 as U, ANALYST_BENCHMARK_DEPENDENCY_LOCK_DIGEST_ALGORITHM as V, ANALYST_BENCHMARK_EVIDENCE_DEPENDENCY_LOCK_SHA256 as W, analystBenchmarkImplementationDigest as X, analystBenchmarkDependencyLockDigest as Y, ANALYST_BENCHMARK_COST_LEDGER_FILE as Z, publicDirectAnalystDefinition as _, primeCodeTraceAnalystDefinition as a, codeTraceBenchCase as at, analystDefinitionAsymmetries as b, preparePublicAnalystBenchmark as c, agentRxPredictionsToFindings as ct, selectPublicBenchmarkRows as d, normalizeBenchmarkLabel as dt, ANALYST_BENCHMARK_OBSERVATIONS_FILE as et, createPublicBenchmarkRlmRunner as f, createPublicBenchmarkDirectRunner as g, runReplVariableAnalystDefinition as h, primeAnalystProtocolSha256 as i, summarizeAgentRxCalibration as it, CODE_TRACE_BENCH_ANALYST_PROMPT as j, effectiveAnalystProtocolSha256 as k, publicBenchmarkDistributions as l, normalizeAgentRxCategory as lt, rlmEngineLimits as m, renderAnalystBenchmarkMarkdown as n, AGENT_RX_UPSTREAM_REVISION as nt, runInlineAnalystDefinition as o, codeTracerPredictionsToFindings as ot, publicRlmAnalystDefinition as p, ANALYST_BENCHMARK_IMPLEMENTATION_FILES as q, createPrimeBenchmarkRunner as r, renderAgentRxCalibrationMarkdown as rt, loadPublicBenchmarkRows as s, agentRxBenchmarkCase as st, runAnalystBenchmarkCommand as t, compareAnalystRunners as tt, publicBenchmarkSelectionReport as u, roundAgentRxStep as ut, runChunkedAnalystDefinition as v, expandCodeTraceFailureBlocks as w, analystDefinitionProtocolSha256 as x, AnalystExpressivenessError as y, loadCodeTraceVerificationArtifacts as z };
|
|
6317
6317
|
|
|
6318
|
-
//# sourceMappingURL=benchmark-command-
|
|
6318
|
+
//# sourceMappingURL=benchmark-command-CdUGku3W.js.map
|