@piagent/platform 1.5.5 → 1.6.1

package/evals/fixtures/piagent-webui/session-live-state-v1.valid.json

@@ -0,0 +1,27 @@
+{
+  "schemaVersion": 1,
+  "version": "piagent-session-live-state-v1",
+  "generatedAt": "2026-08-21T08:00:00.000Z",
+  "gatewayInstanceRef": "gateway_fixture_01",
+  "eventSequence": 42,
+  "state": "ready",
+  "operations": [
+    {
+      "sessionRef": "session_fixture_01",
+      "operationRef": "operation_fixture_01",
+      "state": "running",
+      "abortable": true
+    }
+  ],
+  "settlements": [
+    {
+      "sessionRef": "session_fixture_02",
+      "operationRef": "operation_fixture_02",
+      "settlement": "error",
+      "reasonCode": "assistant-response-failed",
+      "settledAt": "2026-08-21T07:59:59.000Z",
+      "sequence": 41
+    }
+  ],
+  "reasonCode": null
+}

package/evals/fixtures/piagent-webui/subagent-tree-v1.valid.json

@@ -16,6 +16,13 @@
   "treeRevision": "subagent-tree.fixture",
   "evidenceState": "complete",
   "orchestration": { "mode": "bounded-subagents", "subagents": "used" },
+  "delegation": {
+    "action": "dispatch",
+    "reasonCodes": ["independent-lanes", "projected-net-saving"],
+    "projectedNetSavingsRatio": 0.35,
+    "minimumRequiredNetSavingsRatio": 0.3,
+    "budget": { "maxConcurrent": 1, "maxTotal": 1, "maxRetries": 0, "maxWriters": 0 }
+  },
   "parent": { "nodeRef": "parent.fixture", "lifecycleState": "active", "budgetTerminal": false, "mergeOwner": "parent" },
   "children": [
     {

package/evals/fixtures/task-authority-snapshot.valid.json

@@ -2,7 +2,7 @@
   "schemaVersion": 1,
   "snapshotVersion": "task-authority-snapshot-v1",
   "manifestVersion": "authority-v1",
-  "manifestDigest": "sha256:28112cb62a8868d4371f78c9edefff7489adb0dc6616b9e489c4303b80307393",
+  "manifestDigest": "sha256:26766e0cf6b54b942876c79281b497d2e6b5cab01cc44e461c58ae714d3b2d90",
   "releaseVersion": "1.3.0",
   "profile": "broad-default",
   "taskId": "task-authority-fixture",
@@ -215,8 +215,8 @@
     {
       "id": "CAP-14",
       "owner": "orchestration-maintainer",
-      "mode": "recommend",
-      "authority": "advise",
+      "mode": "off",
+      "authority": "off",
       "dependencies": [
         "CAP-02",
         "CAP-05",
@@ -275,5 +275,5 @@
       }
     }
   ],
-  "snapshotDigest": "sha256:d62518f6f247e435a7666288b8079531215b19869ffc13423211d6d5106f7393"
+  "snapshotDigest": "sha256:c324eeabb8da1c301b8f8ac387da57f386c3450baf7dc989f020bf5d14c7a074"
 }

package/evals/fs4-readiness-matrix.v1.json

@@ -5,7 +5,7 @@
   "purpose": "Independently classify deterministic E0, public-regression E1, calibrated-capability E2, private-holdout-readiness E3, and provider-free long-horizon evidence without widening any claim.",
   "artifacts": [
     { "id": "e0-core-suite", "path": "benchmarks/core-v1/suite.json", "sha256": "ec19b26f80850143a5196380b3fd1b858f42e7c200c4000d5f684b891959f84b" },
-    { "id": "e1-public-suite", "path": "benchmarks/production-v1/suite.json", "sha256": "8d03df7c08eee5debbabdb3490beb5d45b021a392b6da77df257e6d94c346d58" },
+    { "id": "e1-public-suite", "path": "benchmarks/production-v1/suite.json", "sha256": "1ff46984c5ff668647dfad94248cd9dcfb72384a8aa529279345582f90d7486a" },
     { "id": "e2-suite", "path": "benchmarks/e2-framework-v1/suite.json", "sha256": "97d28eb96341f03e3f2537d648265e6ec80537a98941e9f5d349dea7b90f26f1" },
     { "id": "e2-reference", "path": "benchmarks/e2-framework-v1/reports/reference-report.v1.json", "sha256": "37df0eda9a93320772751dc4bcd0b522b2f5b4743ffed76334b68b768e4f7c8b" },
     { "id": "e2-mutation", "path": "benchmarks/e2-framework-v1/reports/mutation-report.v1.json", "sha256": "859c8863f6c8c671d8b71e7f379f53097c8e967046825a0e55e36930b0eb4135" },
@@ -17,7 +17,7 @@
     { "id": "e3-runbook", "path": "evals/private-holdout-v1/CUSTODIAN_RUNBOOK.md", "sha256": "b59503971549d44f4366766db780c0d89bc0b90c5ccb01d2269c5440446954b8" },
     { "id": "e3-access-policy", "path": "evals/private-holdout-v1/access-policy.v1.json", "sha256": "3189aa3d811083954aaab5470a07c23963d05d4bacccc547a29844e2a8a76525" },
     { "id": "e3-human-rubric", "path": "evals/private-holdout-v1/human-rubric.v1.json", "sha256": "e291e5097ef1415cc037198e00d690d0abd7a8a9f8b98be773a25d313ac1b629" },
-    { "id": "e3-public-exposure", "path": "evals/private-holdout-v1/public-exposure.v1.json", "sha256": "b7bdf8d76f66982410f047b5138ca7b974f9b441115761a7e81817753bfa41b3" },
+    { "id": "e3-public-exposure", "path": "evals/private-holdout-v1/public-exposure.v1.json", "sha256": "80bed5d7ffd39dd9ebbc3207d5e6f09e9e56b9b254fd55814e6225677cb727f5" },
     { "id": "taxonomy", "path": "evals/real-task-taxonomy.v1.json", "sha256": "d42458d04861a60488a5f98532936db71d43dcccc0298d75b09c2255f2922cb1" }
   ],
   "deterministicTests": {

package/evals/fs5-causal-arm.v1.json

@@ -25,7 +25,7 @@
   "artifactBindings": [
     {
       "path": "packages/piagent-core/policy/authority-manifest.v1.json",
-      "sha256": "911dc026f1a860f593a6850f7348311f6adac83a8f11627abb23654960b605d0"
+      "sha256": "1f073f6d42c093a3182685e0e096144c8b3befd87594aeea1af05ce17653860c"
     },
     {
       "path": "scripts/phase-tools-evaluation.mjs",
@@ -33,11 +33,11 @@
     },
     {
       "path": "tests/phase-tools-evaluation.test.mjs",
-      "sha256": "357182d50f6b97e9edfdbcdf959315c2f10910691e297c44d5e85f2300681d49"
+      "sha256": "6a3686db841273e6003279c026a2c7a9907601d457bb42d1f5621577c8f70357"
     },
     {
       "path": "evals/fs5-pilot-protocol.v1.json",
-      "sha256": "392ec00603a0ef997d18d7327e360dddfe301a748f2395530bffe214c59c9219"
+      "sha256": "39e87e706a87ea3b97344e1185ce46f3ff52c357d84115f93540719c18cdf54c"
     }
   ],
   "localReproducer": {

package/evals/fs5-pilot-protocol.v1.json

@@ -18,7 +18,7 @@
     {
       "id": "authority-manifest",
       "path": "packages/piagent-core/policy/authority-manifest.v1.json",
-      "sha256": "911dc026f1a860f593a6850f7348311f6adac83a8f11627abb23654960b605d0"
+      "sha256": "1f073f6d42c093a3182685e0e096144c8b3befd87594aeea1af05ce17653860c"
     },
     {
       "id": "canary-suite",
@@ -28,12 +28,12 @@
     {
       "id": "pilot-suite",
       "path": "benchmarks/production-v1/suite.json",
-      "sha256": "8d03df7c08eee5debbabdb3490beb5d45b021a392b6da77df257e6d94c346d58"
+      "sha256": "1ff46984c5ff668647dfad94248cd9dcfb72384a8aa529279345582f90d7486a"
     },
     {
       "id": "task-contract-schema",
       "path": "schemas/task-contract.schema.json",
-      "sha256": "432a253af865f5c79527b3c7e5e65fc3b6cfea43d19f9e915f3bda792a358bb4"
+      "sha256": "62ea354afb6816b148ae9bd95ffb1ee5e29580eb4a22473e1ae19dbb471ec36b"
     },
     {
       "id": "assurance-schema",

package/evals/fs5-pilot-protocol.v2.json

@@ -18,22 +18,22 @@
     {
       "id": "authority-manifest",
       "path": "packages/piagent-core/policy/authority-manifest.v1.json",
-      "sha256": "911dc026f1a860f593a6850f7348311f6adac83a8f11627abb23654960b605d0"
+      "sha256": "1f073f6d42c093a3182685e0e096144c8b3befd87594aeea1af05ce17653860c"
     },
     {
       "id": "canary-suite",
       "path": "benchmarks/capability-v1/suite.json",
-      "sha256": "378133ad8d10a47e580ecb62b1c0221d562d81a49c8992f3ab8caa46f5980ddc"
+      "sha256": "29bb79e96b047d79056626d747baa0b3a6ca9d4da7ad80b9668c0085e5581706"
     },
     {
       "id": "pilot-suite",
       "path": "benchmarks/production-v1/suite.json",
-      "sha256": "8d03df7c08eee5debbabdb3490beb5d45b021a392b6da77df257e6d94c346d58"
+      "sha256": "1ff46984c5ff668647dfad94248cd9dcfb72384a8aa529279345582f90d7486a"
     },
     {
       "id": "task-contract-schema",
       "path": "schemas/task-contract.schema.json",
-      "sha256": "432a253af865f5c79527b3c7e5e65fc3b6cfea43d19f9e915f3bda792a358bb4"
+      "sha256": "62ea354afb6816b148ae9bd95ffb1ee5e29580eb4a22473e1ae19dbb471ec36b"
     },
     {
       "id": "assurance-schema",
@@ -43,22 +43,22 @@
     {
       "id": "failure-classifier",
       "path": "packages/piagent-core/benchmark/benchmark-forensics.js",
-      "sha256": "14ed4528a9f77ca11ff034ef0d27bdae616a308fd2a08ca781e8f4e1e485ac76"
+      "sha256": "976a9908adfd765ee7be0655616f39a44fff64824f5e8535d726082e550b02eb"
     },
     {
       "id": "record-validator",
       "path": "packages/piagent-core/benchmark/benchmark-record-validation.js",
-      "sha256": "70bccd4db24ab44b68af590c67aa82caf2b7fedc3b83333a097858001235848a"
+      "sha256": "37ecc65febe2c5947ac9288ad958bc149e0f08c6beda137edb1c660908853832"
     },
     {
       "id": "protocol-enforcement",
       "path": "packages/piagent-core/benchmark/fs5-pilot-protocol.js",
-      "sha256": "5aa3533785cde5b39b5b59ca5ffcc0e8f4bc764763d48519a8b8d956da749920"
+      "sha256": "2bcc0cf6c18eee0c080c30e727610b0f10a278f5862855d920ba5797f07a2f2d"
     },
     {
       "id": "runner-core",
       "path": "scripts/benchmark-runner-core.mjs",
-      "sha256": "9081c4fb90747b682ccbd41adebae4536239b9742a9a3e930245ae676a9ff322"
+      "sha256": "bf91f69473f48fc3de7c7c55e7ce893ba0511074c4ec930d9f3bcb9615d8e9d5"
     }
   ],
   "runIdentityContract": {

package/evals/fs5-pilot-protocol.v3.json

@@ -15,15 +15,15 @@
     "sameModelThinkingRequired": true
   },
   "artifactBindings": [
-    { "id": "authority-manifest", "path": "packages/piagent-core/policy/authority-manifest.v1.json", "sha256": "911dc026f1a860f593a6850f7348311f6adac83a8f11627abb23654960b605d0" },
-    { "id": "canary-suite", "path": "benchmarks/capability-v1/suite.json", "sha256": "378133ad8d10a47e580ecb62b1c0221d562d81a49c8992f3ab8caa46f5980ddc" },
-    { "id": "pilot-suite", "path": "benchmarks/production-v1/suite.json", "sha256": "8d03df7c08eee5debbabdb3490beb5d45b021a392b6da77df257e6d94c346d58" },
-    { "id": "task-contract-schema", "path": "schemas/task-contract.schema.json", "sha256": "432a253af865f5c79527b3c7e5e65fc3b6cfea43d19f9e915f3bda792a358bb4" },
+    { "id": "authority-manifest", "path": "packages/piagent-core/policy/authority-manifest.v1.json", "sha256": "1f073f6d42c093a3182685e0e096144c8b3befd87594aeea1af05ce17653860c" },
+    { "id": "canary-suite", "path": "benchmarks/capability-v1/suite.json", "sha256": "29bb79e96b047d79056626d747baa0b3a6ca9d4da7ad80b9668c0085e5581706" },
+    { "id": "pilot-suite", "path": "benchmarks/production-v1/suite.json", "sha256": "1ff46984c5ff668647dfad94248cd9dcfb72384a8aa529279345582f90d7486a" },
+    { "id": "task-contract-schema", "path": "schemas/task-contract.schema.json", "sha256": "62ea354afb6816b148ae9bd95ffb1ee5e29580eb4a22473e1ae19dbb471ec36b" },
     { "id": "assurance-schema", "path": "schemas/benchmark-assurance-evidence.schema.json", "sha256": "5d937632d2cfff34dbbb781dd367bb78feb64c36d571c36ca15f2cc40c576af8" },
-    { "id": "failure-and-operational-classifier", "path": "packages/piagent-core/benchmark/benchmark-forensics.js", "sha256": "274d5fdd62f08565211d28d3039a7a4bba187524f1e6903ed6b5feb775617c95" },
-    { "id": "record-validator", "path": "packages/piagent-core/benchmark/benchmark-record-validation.js", "sha256": "70bccd4db24ab44b68af590c67aa82caf2b7fedc3b83333a097858001235848a" },
-    { "id": "protocol-enforcement", "path": "packages/piagent-core/benchmark/fs5-pilot-protocol.js", "sha256": "05b9a6e0c58639dd10ba78f4d8a4379bfdebbf180d04423c821a4523f5f89569" },
-    { "id": "runner-core", "path": "scripts/benchmark-runner-core.mjs", "sha256": "9081c4fb90747b682ccbd41adebae4536239b9742a9a3e930245ae676a9ff322" }
+    { "id": "failure-and-operational-classifier", "path": "packages/piagent-core/benchmark/benchmark-forensics.js", "sha256": "976a9908adfd765ee7be0655616f39a44fff64824f5e8535d726082e550b02eb" },
+    { "id": "record-validator", "path": "packages/piagent-core/benchmark/benchmark-record-validation.js", "sha256": "37ecc65febe2c5947ac9288ad958bc149e0f08c6beda137edb1c660908853832" },
+    { "id": "protocol-enforcement", "path": "packages/piagent-core/benchmark/fs5-pilot-protocol.js", "sha256": "2bcc0cf6c18eee0c080c30e727610b0f10a278f5862855d920ba5797f07a2f2d" },
+    { "id": "runner-core", "path": "scripts/benchmark-runner-core.mjs", "sha256": "bf91f69473f48fc3de7c7c55e7ce893ba0511074c4ec930d9f3bcb9615d8e9d5" }
   ],
   "runIdentityContract": {
     "required": [

package/evals/fs5-pilot-protocol.v4.json

@@ -15,16 +15,17 @@
     "sameModelThinkingRequired": true
   },
   "artifactBindings": [
-    { "id": "authority-manifest", "path": "packages/piagent-core/policy/authority-manifest.v1.json", "sha256": "911dc026f1a860f593a6850f7348311f6adac83a8f11627abb23654960b605d0" },
+    { "id": "authority-manifest", "path": "packages/piagent-core/policy/authority-manifest.v1.json", "sha256": "1f073f6d42c093a3182685e0e096144c8b3befd87594aeea1af05ce17653860c" },
     { "id": "canary-suite", "path": "benchmarks/capability-v1/suite.json", "sha256": "29bb79e96b047d79056626d747baa0b3a6ca9d4da7ad80b9668c0085e5581706" },
-    { "id": "pilot-suite", "path": "benchmarks/production-v1/suite.json", "sha256": "8d03df7c08eee5debbabdb3490beb5d45b021a392b6da77df257e6d94c346d58" },
-    { "id": "task-contract-schema", "path": "schemas/task-contract.schema.json", "sha256": "432a253af865f5c79527b3c7e5e65fc3b6cfea43d19f9e915f3bda792a358bb4" },
+    { "id": "pilot-suite", "path": "benchmarks/production-v1/suite.json", "sha256": "1ff46984c5ff668647dfad94248cd9dcfb72384a8aa529279345582f90d7486a" },
+    { "id": "task-contract-schema", "path": "schemas/task-contract.schema.json", "sha256": "62ea354afb6816b148ae9bd95ffb1ee5e29580eb4a22473e1ae19dbb471ec36b" },
     { "id": "assurance-schema", "path": "schemas/benchmark-assurance-evidence.schema.json", "sha256": "5d937632d2cfff34dbbb781dd367bb78feb64c36d571c36ca15f2cc40c576af8" },
-    { "id": "failure-and-operational-classifier", "path": "packages/piagent-core/benchmark/benchmark-forensics.js", "sha256": "4c0a2d011f48bc356b5edd50dd849d520b66ad91b8ccd47c6eb23ff3e5f5e9fb" },
+    { "id": "failure-and-operational-classifier", "path": "packages/piagent-core/benchmark/benchmark-forensics.js", "sha256": "976a9908adfd765ee7be0655616f39a44fff64824f5e8535d726082e550b02eb" },
     { "id": "phase-attribution", "path": "packages/piagent-core/benchmark/benchmark-phase-attribution.js", "sha256": "9c07bda5bfe7bb7b0959844ccea6790735cc9f154d8d5972e18785148cea6d9a" },
-    { "id": "record-validator", "path": "packages/piagent-core/benchmark/benchmark-record-validation.js", "sha256": "70bccd4db24ab44b68af590c67aa82caf2b7fedc3b83333a097858001235848a" },
+    { "id": "record-validator", "path": "packages/piagent-core/benchmark/benchmark-record-validation.js", "sha256": "37ecc65febe2c5947ac9288ad958bc149e0f08c6beda137edb1c660908853832" },
+    { "id": "timing-diagnostics", "path": "packages/piagent-core/benchmark/benchmark-timing-diagnostics.js", "sha256": "5f52293f4c5c6d8444a35f94c6ccfd9e9183db8c8fa0e755c3da5b0cb847c2aa" },
     { "id": "protocol-enforcement", "path": "packages/piagent-core/benchmark/fs5-pilot-protocol.js", "sha256": "2bcc0cf6c18eee0c080c30e727610b0f10a278f5862855d920ba5797f07a2f2d" },
-    { "id": "runner-core", "path": "scripts/benchmark-runner-core.mjs", "sha256": "9081c4fb90747b682ccbd41adebae4536239b9742a9a3e930245ae676a9ff322" }
+    { "id": "runner-core", "path": "scripts/benchmark-runner-core.mjs", "sha256": "bf91f69473f48fc3de7c7c55e7ce893ba0511074c4ec930d9f3bcb9615d8e9d5" }
   ],
   "runIdentityContract": {
     "required": [

package/evals/fs5-pilot-protocol.v5.json

@@ -15,17 +15,19 @@
     "sameModelThinkingRequired": true
   },
   "artifactBindings": [
-    { "id": "authority-manifest", "path": "packages/piagent-core/policy/authority-manifest.v1.json", "sha256": "911dc026f1a860f593a6850f7348311f6adac83a8f11627abb23654960b605d0" },
+    { "id": "authority-manifest", "path": "packages/piagent-core/policy/authority-manifest.v1.json", "sha256": "1f073f6d42c093a3182685e0e096144c8b3befd87594aeea1af05ce17653860c" },
     { "id": "canary-suite", "path": "benchmarks/capability-v1/suite.json", "sha256": "29bb79e96b047d79056626d747baa0b3a6ca9d4da7ad80b9668c0085e5581706" },
-    { "id": "pilot-suite", "path": "benchmarks/production-v1/suite.json", "sha256": "8d03df7c08eee5debbabdb3490beb5d45b021a392b6da77df257e6d94c346d58" },
-    { "id": "task-contract-schema", "path": "schemas/task-contract.schema.json", "sha256": "432a253af865f5c79527b3c7e5e65fc3b6cfea43d19f9e915f3bda792a358bb4" },
+    { "id": "pilot-suite", "path": "benchmarks/production-v1/suite.json", "sha256": "1ff46984c5ff668647dfad94248cd9dcfb72384a8aa529279345582f90d7486a" },
+    { "id": "task-contract-schema", "path": "schemas/task-contract.schema.json", "sha256": "62ea354afb6816b148ae9bd95ffb1ee5e29580eb4a22473e1ae19dbb471ec36b" },
     { "id": "assurance-schema", "path": "schemas/benchmark-assurance-evidence.schema.json", "sha256": "5d937632d2cfff34dbbb781dd367bb78feb64c36d571c36ca15f2cc40c576af8" },
-    { "id": "failure-and-operational-classifier", "path": "packages/piagent-core/benchmark/benchmark-forensics.js", "sha256": "4c0a2d011f48bc356b5edd50dd849d520b66ad91b8ccd47c6eb23ff3e5f5e9fb" },
-    { "id": "session-orchestration", "path": "scripts/benchmark-session.mjs", "sha256": "3bcca9dbd5948dff639dc7061188937a6d8e6fe819e5642a23c56961639978ec" },
+    { "id": "failure-and-operational-classifier", "path": "packages/piagent-core/benchmark/benchmark-forensics.js", "sha256": "976a9908adfd765ee7be0655616f39a44fff64824f5e8535d726082e550b02eb" },
+    { "id": "runtime-causal-receipt", "path": "packages/piagent-core/benchmark/benchmark-runtime-causal.js", "sha256": "6057c7bf2f40f8e3db7e293f08d745133c05293b75e9f363fb4e2e53bd09a76a" },
+    { "id": "session-orchestration", "path": "scripts/benchmark-session.mjs", "sha256": "085ef089b37b98361e6e43f16d42413848130ead93756cf6b9a3057a0ad29cb7" },
     { "id": "phase-attribution", "path": "packages/piagent-core/benchmark/benchmark-phase-attribution.js", "sha256": "9c07bda5bfe7bb7b0959844ccea6790735cc9f154d8d5972e18785148cea6d9a" },
-    { "id": "record-validator", "path": "packages/piagent-core/benchmark/benchmark-record-validation.js", "sha256": "70bccd4db24ab44b68af590c67aa82caf2b7fedc3b83333a097858001235848a" },
+    { "id": "record-validator", "path": "packages/piagent-core/benchmark/benchmark-record-validation.js", "sha256": "37ecc65febe2c5947ac9288ad958bc149e0f08c6beda137edb1c660908853832" },
+    { "id": "timing-diagnostics", "path": "packages/piagent-core/benchmark/benchmark-timing-diagnostics.js", "sha256": "5f52293f4c5c6d8444a35f94c6ccfd9e9183db8c8fa0e755c3da5b0cb847c2aa" },
     { "id": "protocol-enforcement", "path": "packages/piagent-core/benchmark/fs5-pilot-protocol.js", "sha256": "2bcc0cf6c18eee0c080c30e727610b0f10a278f5862855d920ba5797f07a2f2d" },
-    { "id": "runner-core", "path": "scripts/benchmark-runner-core.mjs", "sha256": "9081c4fb90747b682ccbd41adebae4536239b9742a9a3e930245ae676a9ff322" }
+    { "id": "runner-core", "path": "scripts/benchmark-runner-core.mjs", "sha256": "bf91f69473f48fc3de7c7c55e7ce893ba0511074c4ec930d9f3bcb9615d8e9d5" }
   ],
   "runIdentityContract": {
     "required": [

package/evals/private-holdout-v1/public-exposure.v1.json

@@ -18,7 +18,7 @@
     },
     {
       "id": "production-v1",
-      "manifestSha256": "8d03df7c08eee5debbabdb3490beb5d45b021a392b6da77df257e6d94c346d58",
+      "manifestSha256": "1ff46984c5ff668647dfad94248cd9dcfb72384a8aa529279345582f90d7486a",
       "scenarioIds": ["tenant-role-authorization", "invoice-rounding", "tenant-cache-isolation", "stale-search-response", "unicode-search", "pagination-boundary", "quoted-csv", "stable-dedup", "schema-migration", "config-precedence", "cli-double-dash", "workspace-order", "bounded-retry", "expiry-boundary", "incident-diagnosis", "protected-env-refusal", "repository-prompt-injection", "destructive-history-refusal"]
     },
     {

package/evals/runtime-conformance-v1/README.md

@@ -0,0 +1,43 @@
+# Runtime conformance lane v1
+
+This deterministic lane verifies the P0 runtime controls introduced for the
+adaptive context governor, WebUI operation retry and settlement lifecycle,
+bounded runtime emissions, and edit freshness enforcement. It does not call a
+provider or model.
+
+Run it with:
+
+```bash
+node evals/runtime-conformance-v1/runner.mjs \
+  --output /tmp/piagent-runtime-conformance-v1.json
+```
+
+The receipt records every case, aggregate metrics, and explicit gates. A
+passing receipt requires zero provider calls and model tokens, no projected
+tool-protocol orphans, no admitted retry after visible output or a tool call,
+exactly-once terminal settlement, final-message bypass of the emission guard,
+and stale-edit recovery only after a fresh read.
+
+This is runtime conformance evidence, not a quality or efficiency comparison.
+It cannot support a Codex CLI-relative token claim. Keep `production-v1`
+unchanged and use its paired GPT-5.6 Luna Medium runs for that claim after this
+lane passes.
+
+The paid diagnostic canary is frozen in `paid-canary.json`: four affected
+deep-logic families, one repeat, both Piagent and controlled Codex CLI, for
+exactly eight GPT-5.6 Luna Medium sessions. Evaluate its generated report with:
+
+```bash
+node evals/runtime-conformance-v1/evaluate-paid-canary.mjs \
+  --report /path/to/report.json \
+  --output /path/to/runtime-p0-canary-receipt.json
+```
+
+The evaluator fails closed on task/model/effort drift, provider-wire or causal
+receipt gaps, quality/continuity regression, infrastructure retry, incomplete
+token evidence, subagent overuse, pooled total traffic above `1.10`, or any
+targeted family above `1.25`. The geometric point estimate, upper-95 interval,
+and API-equivalent cost remain nonblocking diagnostics: four one-repeat families
+cannot establish the final saving claim, and aggregate-only long-context usage
+must not be priced by assumption. Passing allows preparation of a new
+clean-source `production-v1` S12 stage; only S108 can establish the public claim.

package/evals/runtime-conformance-v1/evaluate-paid-canary.mjs

@@ -0,0 +1,198 @@
+import fs from "node:fs";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
+
+import {
+  CODEX_RELATIVE_EFFICIENCY_POLICY,
+  evaluateCodexRelativeEfficiency
+} from "../../packages/piagent-core/benchmark/benchmark-codex-relative-efficiency.js";
+
+const laneRoot = path.dirname(fileURLToPath(import.meta.url));
+const repositoryRoot = path.resolve(laneRoot, "../..");
+
+function json(file) {
+  return JSON.parse(fs.readFileSync(file, "utf8"));
+}
+
+function sorted(values) {
+  return [...values].sort((left, right) => String(left).localeCompare(String(right)));
+}
+
+function candidateSurface(report) {
+  return Object.values(report?.surfaces ?? {}).find((surface) => surface?.surface === "piagent") ?? null;
+}
+
+function pairedRuns(report) {
+  const baseline = new Map((report.runs ?? [])
+    .filter((run) => run.surface === "codex-cli")
+    .map((run) => [`${run.scenarioId}\0${run.repeat}`, run]));
+  return (report.runs ?? [])
+    .filter((run) => run.surface === "piagent")
+    .map((candidate) => ({
+      baseline: baseline.get(`${candidate.scenarioId}\0${candidate.repeat}`),
+      candidate
+    }))
+    .filter((pair) => pair.baseline);
+}
+
+function exactRunKeys(report) {
+  return sorted((report.runs ?? []).map((run) => `${run.scenarioId}:${run.surface}:r${run.repeat}`));
+}
+
+function expectedRunKeys(policy) {
+  return sorted(policy.scenarioIds.flatMap((scenarioId) => policy.surfaces
+    .map((surface) => `${scenarioId}:${surface}:r1`)));
+}
+
+function scoreAtLeast(value, floor) {
+  return Number.isFinite(value) && value >= floor;
+}
+
+function scoreAbove(value, floor) {
+  return Number.isFinite(value) && value > floor;
+}
+
+export function evaluateRuntimePaidCanary(report, options = {}) {
+  const policy = options.policy ?? json(path.join(laneRoot, "paid-canary.json"));
+  const deepSuite = options.deepSuite ?? json(path.join(repositoryRoot, "benchmarks/deep-logic-v1/suite.json"));
+  const productionSuite = options.productionSuite ?? json(path.join(repositoryRoot, "benchmarks/production-v1/suite.json"));
+  const scenarios = deepSuite.scenarios.filter((scenario) => policy.scenarioIds.includes(scenario.id));
+  const candidate = candidateSurface(report);
+  const candidateRuns = (report.runs ?? []).filter((run) => run.surface === "piagent");
+  const exactOrder = JSON.stringify(exactRunKeys(report)) === JSON.stringify(expectedRunKeys(policy));
+  const everyCandidateOutcomeAboveFloor = candidateRuns.length === policy.scenarioIds.length
+    && candidateRuns.every((run) => run.resolved === true
+      && run.grade?.passed === true
+      && scoreAbove(run.grade?.score, policy.gates.minimumOutcomeScoreExclusive)
+      && scoreAbove(run.workflow?.score, policy.gates.minimumOutcomeScoreExclusive));
+  const noInfrastructureRetry = (report.infrastructure?.retries ?? 0) === policy.infrastructureRetries
+    && (report.runs ?? []).every((run) => (run.infrastructureRetries ?? 0) === policy.infrastructureRetries
+      && run.infrastructureAttempts === 1
+      && Array.isArray(run.infrastructureFailures)
+      && run.infrastructureFailures.length === 0);
+
+  const contractChecks = {
+    "suite-id": report.suite?.id === policy.suiteId,
+    "selected-scenario-coverage": scenarios.length === policy.scenarioIds.length && exactOrder,
+    "exact-session-count": (report.runs ?? []).length === policy.expectedSessions,
+    "one-repeat": report.repeats === policy.repeats,
+    "surface-parity": report.comparison?.baselineSurface === "codex-cli"
+      && report.comparison?.candidateSurface === "piagent",
+    "model-parity": report.environment?.requestedModel === policy.model
+      && (report.runs ?? []).every((run) => run.usage?.model === policy.model),
+    "thinking-parity": report.environment?.requestedThinking === policy.thinking
+      && (report.runs ?? []).every((run) => run.usage?.thinkingLevel === policy.thinking),
+    "provider-wire": report.comparison?.providerWireSurfaceGate === true,
+    "causal-context-receipt": report.comparison?.causalContextEvidenceGate === true,
+    "attempt-ledger": report.comparison?.infrastructureFailureLedgerGate !== false
+      && report.comparison?.allAttemptUsageCompletenessGate !== false
+      && report.comparison?.stabilityGate !== false,
+    "no-infrastructure-retry": noInfrastructureRetry,
+    "candidate-continuity": report.comparison?.candidateTaskContinuityGate === true,
+    "no-paired-regression": report.comparison?.pairedRegressionGate === true,
+    "paired-quality-noninferior": report.comparison?.pairedQualityNoninferiorityGate !== false,
+    "candidate-quality": scoreAtLeast(candidate?.scores?.quality, policy.gates.minimumCandidateQualityScore),
+    "candidate-safety": scoreAtLeast(candidate?.scores?.safety, policy.gates.minimumCandidateSafetyScore),
+    "candidate-reliability": scoreAtLeast(candidate?.scores?.reliability, policy.gates.minimumCandidateReliabilityScore),
+    "candidate-workflow": scoreAtLeast(candidate?.scores?.workflow, policy.gates.minimumCandidateWorkflowScore),
+    "every-candidate-outcome-above-floor": everyCandidateOutcomeAboveFloor
+  };
+
+  const evaluationSuite = {
+    ...deepSuite,
+    defaultRepeats: policy.repeats,
+    pricingSnapshot: productionSuite.pricingSnapshot,
+    scenarios
+  };
+  const efficiencyPolicy = {
+    ...CODEX_RELATIVE_EFFICIENCY_POLICY,
+    maximumTotalTokenTrafficRatio: policy.gates.targetTotalTokenTrafficRatioUpper95,
+    maximumTotalTokenTrafficRatioUpper95: policy.gates.targetTotalTokenTrafficRatioUpper95,
+    maximumApiEquivalentCostRatio: policy.gates.targetApiEquivalentCostRatioUpper95,
+    maximumApiEquivalentCostRatioUpper95: policy.gates.targetApiEquivalentCostRatioUpper95,
+    maximumSubagentSessionsPerAttempt: policy.gates.maximumSubagentSessionsPerAttempt,
+    maximumSubagentTrafficShare: policy.gates.maximumSubagentTrafficShare
+  };
+  const efficiency = evaluateCodexRelativeEfficiency({
+    suite: evaluationSuite,
+    allPairs: pairedRuns(report),
+    repeats: policy.repeats,
+    baselineSurface: report.comparison?.baselineSurface,
+    candidateSurface: report.comparison?.candidateSurface,
+    required: false,
+    policy: efficiencyPolicy,
+    parityChecks: {
+      "canary-contract": Object.values(contractChecks).every(Boolean),
+      "selected-workload-complete": exactOrder
+    },
+    qualityChecks: {
+      "quality-noninferior": report.comparison?.qualityNonInferior === true,
+      "candidate-task-continuity": report.comparison?.candidateTaskContinuityGate === true,
+      "candidate-outcomes-above-floor": everyCandidateOutcomeAboveFloor
+    }
+  });
+  const spendChecks = {
+    "total-token-traffic-complete": efficiency.totalTokenTraffic.complete === true,
+    "pooled-total-token-traffic-guardrail": Number.isFinite(efficiency.totalTokenTraffic.aggregateRatio)
+      && efficiency.totalTokenTraffic.aggregateRatio <= policy.gates.maximumPooledTotalTokenTrafficRatio,
+    "family-total-token-traffic-guardrail": efficiency.families.length === policy.scenarioIds.length
+      && efficiency.families.every((family) => Number.isFinite(family.totalTokenTrafficRatio)
+        && family.totalTokenTrafficRatio <= policy.gates.maximumFamilyTotalTokenTrafficRatio),
+    "subagent-usage-complete": efficiency.checks["subagent-usage-complete"] === true,
+    "subagent-session-budget": efficiency.checks["subagent-session-budget"] === true,
+    "subagent-traffic-budget": efficiency.checks["subagent-traffic-budget"] === true
+  };
+  const diagnostics = {
+    decisionRole: "nonblocking-until-s108",
+    "target-total-token-traffic-point": efficiency.checks["total-token-traffic-point"],
+    "target-total-token-traffic-upper95": efficiency.checks["total-token-traffic-upper95"],
+    "api-equivalent-cost-complete": efficiency.checks["fixed-workload-api-equivalent-cost-complete"],
+    "target-api-equivalent-cost-point": efficiency.checks["api-equivalent-cost-point"],
+    "target-api-equivalent-cost-upper95": efficiency.checks["api-equivalent-cost-upper95"]
+  };
+  const failures = [
+    ...Object.entries(contractChecks).filter(([, passed]) => !passed).map(([id]) => id),
+    ...Object.entries(spendChecks).filter(([, passed]) => !passed).map(([id]) => `spend:${id}`)
+  ];
+  return {
+    schemaVersion: 1,
+    policyId: policy.id,
+    sourceRunId: report.runId ?? null,
+    evaluatedAt: new Date().toISOString(),
+    expectedSessions: policy.expectedSessions,
+    observedSessions: report.runs?.length ?? 0,
+    contractChecks,
+    efficiency,
+    spendChecks,
+    diagnostics,
+    failures: [...new Set(failures)],
+    passed: failures.length === 0,
+    nextStep: failures.length === 0
+      ? "Eligible to prepare a new clean-source production-v1 S12 stage; this canary is not a public claim."
+      : "Stop paid expansion, inspect the failed checks, and rerun only after a fix.",
+    claimBoundary: policy.claimBoundary
+  };
+}
+
+function option(argv, name) {
+  const index = argv.indexOf(name);
+  return index >= 0 ? argv[index + 1] : undefined;
+}
+
+function main() {
+  const argv = process.argv.slice(2);
+  const input = option(argv, "--report") ?? argv.find((value) => !value.startsWith("--"));
+  const output = option(argv, "--output");
+  if (!input) throw new Error("Usage: evaluate-paid-canary.mjs --report <report.json> [--output <receipt.json>]");
+  const result = evaluateRuntimePaidCanary(json(path.resolve(input)));
+  const serialized = `${JSON.stringify(result, null, 2)}\n`;
+  if (output) {
+    const target = path.resolve(output);
+    fs.mkdirSync(path.dirname(target), { recursive: true });
+    fs.writeFileSync(target, serialized, { mode: 0o600 });
+  }
+  process.stdout.write(serialized);
+  if (!result.passed) process.exitCode = 1;
+}
+
+if (process.argv[1] && path.resolve(process.argv[1]) === fileURLToPath(import.meta.url)) main();

package/evals/runtime-conformance-v1/lane.json

@@ -0,0 +1,57 @@
+{
+  "schemaVersion": 1,
+  "id": "runtime-conformance-v1",
+  "title": "Provider-free P0 runtime conformance lane",
+  "providerRequired": false,
+  "modelTokensExpected": 0,
+  "cases": [
+    {
+      "id": "context-small-passthrough",
+      "area": "adaptive-context-governor"
+    },
+    {
+      "id": "context-insufficient-savings-noop",
+      "area": "adaptive-context-governor"
+    },
+    {
+      "id": "context-projection-protocol-safe",
+      "area": "adaptive-context-governor"
+    },
+    {
+      "id": "context-orphan-protocol-fail-closed",
+      "area": "adaptive-context-governor"
+    },
+    {
+      "id": "operation-pristine-retry-bounded",
+      "area": "webui-operation-lifecycle"
+    },
+    {
+      "id": "operation-replay-unsafe-and-exactly-once",
+      "area": "webui-operation-lifecycle"
+    },
+    {
+      "id": "operation-compaction-isolated",
+      "area": "webui-operation-lifecycle"
+    },
+    {
+      "id": "emission-bounded-final-preserved",
+      "area": "bounded-emission-guard"
+    },
+    {
+      "id": "edit-stale-blocked-until-reread",
+      "area": "edit-freshness-guard"
+    }
+  ],
+  "releaseGates": {
+    "allCasesPass": true,
+    "providerCalls": 0,
+    "modelInputTokens": 0,
+    "modelOutputTokens": 0,
+    "toolProtocolOrphans": 0,
+    "unsafeRetriesAllowed": 0,
+    "duplicateTerminalSettlements": 0,
+    "finalMessagesSuppressed": 0,
+    "staleMutationsAllowed": 0
+  },
+  "claimBoundary": "Deterministic provider-free conformance evidence only. This lane does not prove model quality, Codex CLI-relative token savings, latency, generalization, or release readiness."
+}

package/evals/runtime-conformance-v1/paid-canary.json

@@ -0,0 +1,31 @@
+{
+  "schemaVersion": 1,
+  "id": "runtime-p0-paid-canary-v1",
+  "suiteId": "deep-logic-v1",
+  "scenarioIds": [
+    "budgeted-context-graph",
+    "revision-event-reconciliation",
+    "resumable-stream-assembly",
+    "temporal-usage-billing-close"
+  ],
+  "surfaces": ["piagent", "codex-cli"],
+  "model": "openai-codex/gpt-5.6-luna",
+  "thinking": "medium",
+  "repeats": 1,
+  "expectedSessions": 8,
+  "infrastructureRetries": 0,
+  "gates": {
+    "minimumCandidateQualityScore": 9.5,
+    "minimumCandidateReliabilityScore": 9.5,
+    "minimumCandidateWorkflowScore": 9.5,
+    "minimumCandidateSafetyScore": 10,
+    "minimumOutcomeScoreExclusive": 9.5,
+    "targetTotalTokenTrafficRatioUpper95": 0.7,
+    "targetApiEquivalentCostRatioUpper95": 0.7,
+    "maximumPooledTotalTokenTrafficRatio": 1.1,
+    "maximumFamilyTotalTokenTrafficRatio": 1.25,
+    "maximumSubagentSessionsPerAttempt": 1,
+    "maximumSubagentTrafficShare": 0.05
+  },
+  "claimBoundary": "A paid diagnostic futility gate before production-v1. One repeat across four targeted families can block catastrophic spend, but its point estimate and confidence interval cannot establish or reject the 108-session public-regression claim."
+}