@piagent/platform 1.6.1 → 1.7.1

package/evals/private-holdout-v1/public-exposure.v2.json

@@ -0,0 +1,273 @@
+{
+  "schemaVersion": 2,
+  "id": "author-visible-evaluation-exposure-v2",
+  "previousExposure": {
+    "path": "evals/private-holdout-v1/public-exposure.v1.json",
+    "sha256": "80bed5d7ffd39dd9ebbc3207d5e6f09e9e56b9b254fd55814e6225677cb727f5"
+  },
+  "taxonomy": {
+    "path": "evals/real-task-taxonomy.v1.json",
+    "sha256": "d42458d04861a60488a5f98532936db71d43dcccc0298d75b09c2255f2922cb1"
+  },
+  "visibleSuites": [
+    {
+      "id": "capability-v1",
+      "path": "benchmarks/capability-v1/suite.json",
+      "manifestSha256": "29bb79e96b047d79056626d747baa0b3a6ca9d4da7ad80b9668c0085e5581706",
+      "scenarioIds": [
+        "multi-package-rollout",
+        "fullstack-search-contract",
+        "concurrent-lease-lifecycle",
+        "resumable-migration-runner",
+        "durable-session-control-plane",
+        "causal-timeline-recovery"
+      ]
+    },
+    {
+      "id": "core-v1",
+      "path": "benchmarks/core-v1/suite.json",
+      "manifestSha256": "ec19b26f80850143a5196380b3fd1b858f42e7c200c4000d5f684b891959f84b",
+      "scenarioIds": [
+        "invoice-quantity",
+        "discount-order",
+        "targeted-session",
+        "protected-env-refusal"
+      ]
+    },
+    {
+      "id": "deep-logic-v1",
+      "path": "benchmarks/deep-logic-v1/suite.json",
+      "manifestSha256": "185221697d581c0aa2cce7aaf6cf480143319acf012c4ca8685690da8d04f98b",
+      "scenarioIds": [
+        "revision-event-reconciliation",
+        "fair-dependency-scheduler",
+        "layered-policy-resolution",
+        "budgeted-context-graph",
+        "resumable-stream-assembly",
+        "transactional-config-merge",
+        "temporal-usage-billing-close"
+      ]
+    },
+    {
+      "id": "e2-framework-v1",
+      "path": "benchmarks/e2-framework-v1/suite.json",
+      "manifestSha256": "97d28eb96341f03e3f2537d648265e6ec80537a98941e9f5d349dea7b90f26f1",
+      "scenarioIds": [
+        "hono-tenant-api",
+        "hono-accessible-search",
+        "sqlite-resumable-inventory",
+        "workspace-policy-rollout"
+      ]
+    },
+    {
+      "id": "production-v1",
+      "path": "benchmarks/production-v1/suite.json",
+      "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"
+      ]
+    },
+    {
+      "id": "production-v2",
+      "path": "benchmarks/production-v2/suite.json",
+      "manifestSha256": "10e5331444f1d499484bd3843f609cac8508f75e025df9ce967230b5799c6bff",
+      "scenarioIds": [
+        "tenant-role-authorization",
+        "tenant-cache-isolation",
+        "revoked-session-cache",
+        "invoice-rounding",
+        "expiry-boundary",
+        "billing-cutoff-clock-skew",
+        "pagination-boundary",
+        "stale-search-response",
+        "abort-reconnect-supersession",
+        "unicode-search",
+        "quoted-csv",
+        "chunked-record-boundary",
+        "schema-migration",
+        "stable-dedup",
+        "idempotent-replay-conflict",
+        "cli-double-dash",
+        "config-precedence",
+        "workspace-order",
+        "bounded-retry",
+        "resumable-checkpoint-partial-failure",
+        "incident-diagnosis",
+        "protected-env-refusal",
+        "repository-prompt-injection",
+        "destructive-history-refusal",
+        "backend-frontend-contract-sync",
+        "workflow-switch-same-session",
+        "reconnect-chat-event-order"
+      ]
+    },
+    {
+      "id": "production-v3",
+      "path": "benchmarks/production-v3/suite.json",
+      "manifestSha256": "e820dfbda06139e8bb5483ed349d3fa397e988423fa0dd1133c37a0bea880ddf",
+      "scenarioIds": [
+        "tenant-role-authorization",
+        "tenant-cache-isolation",
+        "revoked-session-cache",
+        "invoice-rounding",
+        "expiry-boundary",
+        "billing-cutoff-clock-skew",
+        "pagination-boundary",
+        "stale-search-response",
+        "abort-reconnect-supersession",
+        "unicode-search",
+        "quoted-csv",
+        "chunked-record-boundary",
+        "schema-migration",
+        "stable-dedup",
+        "idempotent-replay-conflict",
+        "cli-double-dash",
+        "config-precedence",
+        "workspace-order",
+        "bounded-retry",
+        "resumable-checkpoint-partial-failure",
+        "incident-diagnosis",
+        "protected-env-refusal",
+        "repository-prompt-injection",
+        "destructive-history-refusal",
+        "backend-frontend-contract-sync",
+        "workflow-switch-same-session",
+        "reconnect-chat-event-order"
+      ]
+    }
+  ],
+  "contractLibraries": [
+    {
+      "path": "adapters/node-typescript/contract-families.json",
+      "sha256": "15593d54cded8ecbd57a5e1cdbd557622dc48f2da0b7539a58b769d7fa3c93dd",
+      "families": [
+        {
+          "id": "iso-expiry-millisecond-profile",
+          "version": 1
+        },
+        {
+          "id": "finite-scalar-sum",
+          "version": 1
+        },
+        {
+          "id": "finite-list-sum",
+          "version": 1
+        },
+        {
+          "id": "deadline-status",
+          "version": 1
+        },
+        {
+          "id": "defined-config-precedence",
+          "version": 1
+        },
+        {
+          "id": "idempotent-accumulator-checkpoint",
+          "version": 1
+        },
+        {
+          "id": "secret-redaction-literals",
+          "version": 1
+        },
+        {
+          "id": "bounded-retry-injected-sleep",
+          "version": 1
+        },
+        {
+          "id": "partial-checkpoint-resume",
+          "version": 1
+        },
+        {
+          "id": "defined-config-precedence",
+          "version": 2
+        },
+        {
+          "id": "workflow-message-reducer",
+          "version": 1
+        },
+        {
+          "id": "stale-search-reducer",
+          "version": 1
+        },
+        {
+          "id": "epoch-request-lifecycle",
+          "version": 1
+        }
+      ]
+    }
+  ],
+  "visibleTrees": [
+    {
+      "path": "adapters",
+      "fileCount": 15,
+      "sha256": "f49f1170cfadb4a4a7ba18ed80370fc94272956ac058275b758755219da2274a"
+    },
+    {
+      "path": "benchmarks",
+      "fileCount": 921,
+      "sha256": "4a2e4b8d74961711afdee982452f921c7f1834c975287c62c3a752f05d636e8a"
+    },
+    {
+      "path": "evals/architecture-conformance-v1",
+      "fileCount": 2,
+      "sha256": "f547e6d16739ff91b8e70bcbc5418a905ff41194c6e1ac58873251ce0609a1e1"
+    },
+    {
+      "path": "evals/golden",
+      "fileCount": 1,
+      "sha256": "c38470c5d15962a0058b920734162661b4e5e3dbd8971c02002412564614656a"
+    },
+    {
+      "path": "evals/harness-next",
+      "fileCount": 5,
+      "sha256": "f0df7bbeb2f91ee7d0f565a38cad103cb9edecce1f68a744e2d58c6bd7c20df1"
+    },
+    {
+      "path": "evals/long-horizon-v1",
+      "fileCount": 6,
+      "sha256": "9bbaeca407f49546cd2461ce1fb869f2934eedd7e0272cab01fe10348414e2ec"
+    },
+    {
+      "path": "evals/runtime-conformance-v1",
+      "fileCount": 5,
+      "sha256": "f3ac30a6faef3bcae91314847f0a26b6fa74e2dbadd85b11740ebc854cbef4ce"
+    },
+    {
+      "path": "evals/scenarios",
+      "fileCount": 1,
+      "sha256": "cf5250c7085b06ced7dc5dd7b73899cbc6b866127a2b12280302ee8de75ab78f"
+    },
+    {
+      "path": "tests",
+      "fileCount": 487,
+      "sha256": "388a16945d382b23bda4f0af6966716e0612b4ed20d5a2f833819a9e5c8f337d"
+    }
+  ],
+  "custodianComparisonRules": {
+    "familyKey": "Compare private task and acceptance-mechanism lineage against every listed suite, contract family, public tree and predecessor exposure; renamed, generated or structurally varied descendants are not disjoint.",
+    "repositoryKey": "Compare repository history, origin, package graph and fixture lineage, not names alone.",
+    "minimumPrivateScenarios": 6,
+    "minimumPrivateFamilies": 6,
+    "minimumPrivateRepositories": 6,
+    "externalExposure": "The custodian must independently account for author-visible material outside these public roots, including prior local diagnostics and retained campaigns; this inventory is not a complete record of author knowledge.",
+    "publicOutput": "Export only the closed schema-v2 assurance receipt and independent custody attestation."
+  },
+  "claimBoundary": "Current public-input inventory only, not independent evaluation, approval, generalization or release evidence. Public-tree digests bind [relative path, file SHA-256] pairs globally sorted by the slash-separated path's UTF-16 code units, serialized as compact JSON and hashed as UTF-8; no private artifacts are loaded."
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@piagent/platform",
-  "version": "1.6.1",
+  "version": "1.7.1",
   "license": "MIT",
   "publishConfig": {
     "access": "public",
@@ -110,15 +110,18 @@
   "piagent": {
     "runtimeIntegrityEntrypoints": [
       "packages/piagent-webui/server/sidecar-main.ts",
+      "packages/piagent-core/extensions/acceptance-executor/worker.mjs",
+      "packages/piagent-core/extensions/acceptance-executor/package.json",
+      "packages/piagent-core/extensions/acceptance-executor/package-lock.json",
       "scripts/benchmark-runner-core.mjs",
       "scripts/register-typescript-loader.mjs",
       "scripts/typescript-loader.mjs"
     ]
   },
   "peerDependencies": {
-    "@earendil-works/pi-ai": "0.84.1",
-    "@earendil-works/pi-coding-agent": "0.84.1",
-    "typebox": "1.3.7"
+    "@earendil-works/pi-ai": "0.86.1",
+    "@earendil-works/pi-coding-agent": "0.86.1",
+    "typebox": "1.3.27"
   },
   "peerDependenciesMeta": {
     "@earendil-works/pi-ai": {
@@ -132,6 +135,7 @@
     }
   },
   "dependencies": {
+    "@babel/parser": "7.29.8",
     "ws": "8.21.3"
   },
   "devDependencies": {
@@ -160,6 +164,7 @@
     "doctor": "bash scripts/team-doctor.sh",
     "benchmark": "node scripts/benchmark-runner.mjs",
     "benchmark:deep": "node scripts/benchmark-runner.mjs --deep --surfaces piagent,codex-cli --model openai-codex/gpt-5.6-luna --thinking medium --allow-pi-auth-writeback",
+    "benchmark:production-v2": "node scripts/benchmark-runner.mjs --suite production-v2 --allow-pi-auth-writeback",
     "benchmark:webui-parity": "node packages/piagent-webui/benchmark/parity-benchmark.mjs",
     "benchmark:redaction": "node scripts/redaction-benchmark.mjs",
     "audit:runtime": "bash scripts/audit-runtime-host.sh",

package/packages/piagent-core/README.md

@@ -113,7 +113,7 @@ groups load only when requested.
 - `/workflow`: one launcher for task, scout, BE-to-FE, discuss, plan, review, commit, PR, platform-improve, and onboarding workflows.
 - `/usage`: runtime usage namespace for live snapshot, history hint, preflight, compact, compact-log captures, and context efficiency.
 - Pi native `/name`: set the current session name for Agent Watch/report mapping; Piagent observes the native rename event and keeps `/setname` as a compatibility alias.
-- `/fresh`: open a fresh governed session for `task`, `scout`, or `be-to-fe`.
+- `/fresh`: open a fresh governed session for any canonical `/workflow` option; `/fresh help` lists the current catalog.
 - `/context`: runtime context namespace for index/rebuild/search/pack/test-impact/efficiency, task preflight, and semantic compact.
 - `/permission`: runtime permission namespace for status/read-only/workspace-write/full-access.
 - `/onboard`: runtime onboarding namespace; `run` launches the first-read onboarding workflow.
@@ -150,7 +150,7 @@ pi install git:github.com/Vt-mmm/piagent
 
 Use `git:github.com/Vt-mmm/piagent@vX.Y.Z` when pinning a reproducible project package source.
 
-Runtime support follows the root release matrix: Node.js `>=22.19.0`, Pi Coding Agent `0.84.1`, verified rollout on macOS Apple Silicon + Bash and Linux x64 + Bash, supported-target smoke verification for macOS Intel/Linux ARM64, no native Windows team rollout target yet, and WSL2 experimental.
+Runtime support follows the root release matrix: Node.js `>=22.19.0`, Pi Coding Agent `0.86.1` (SDK upgrade under local qualification; historical platform evidence below used `0.84.1`), verified rollout on macOS Apple Silicon + Bash and Linux x64 + Bash, supported-target smoke verification for macOS Intel/Linux ARM64, no native Windows team rollout target yet, and WSL2 experimental.
 
 ## Project profile
 

package/packages/piagent-core/benchmark/benchmark-all-attempt-efficiency.js

@@ -0,0 +1,55 @@
+import { atMostWithinFloatingPrecision, rounded } from "./benchmark-statistics.js";
+
+function surfaceFresh(accounting, surface) {
+  const bucket = accounting?.allAttempts?.bySurface?.[surface];
+  return {
+    attempts: Number.isSafeInteger(bucket?.attempts) ? bucket.attempts : 0,
+    freshTokens: Number.isSafeInteger(bucket?.tokens?.fresh) ? bucket.tokens.fresh : null
+  };
+}
+
+/**
+ * Net efficiency uses every accepted attempt plus every exact failed/retry
+ * attempt. It is deliberately pooled: no successful-only filtering, median,
+ * retry forgiveness, or subscription-credit conversion is applied.
+ */
+export function evaluateAllAttemptPooledFreshEfficiency({
+  tokenAccounting,
+  baselineSurface,
+  candidateSurface,
+  maximumRatio
+}) {
+  const required = Number.isFinite(maximumRatio);
+  const baseline = surfaceFresh(tokenAccounting, baselineSurface);
+  const candidate = surfaceFresh(tokenAccounting, candidateSurface);
+  const exact = tokenAccounting?.allAttempts?.complete === true;
+  const ratio = exact
+    && Number.isFinite(baseline.freshTokens) && baseline.freshTokens > 0
+    && Number.isFinite(candidate.freshTokens)
+      ? candidate.freshTokens / baseline.freshTokens
+      : null;
+  const passed = !required ? null : exact
+    && Number.isFinite(ratio)
+    && atMostWithinFloatingPrecision(ratio, maximumRatio);
+  return {
+    schemaVersion: 1,
+    required,
+    definition: "pooled-fresh-tokens-across-all-accepted-and-exact-failed-retry-attempts",
+    baselineSurface,
+    candidateSurface,
+    baseline,
+    candidate,
+    exact,
+    ledgerExact: tokenAccounting?.allAttempts?.ledgerExact === true,
+    ledgerIssues: Array.isArray(tokenAccounting?.allAttempts?.ledgerIssues)
+      ? tokenAccounting.allAttempts.ledgerIssues
+      : [],
+    unknownAttempts: Number(tokenAccounting?.allAttempts?.unknownAttempts ?? 0),
+    ratio: rounded(ratio, 6),
+    ratioRaw: ratio,
+    reductionPercent: Number.isFinite(ratio) ? rounded((1 - ratio) * 100, 2) : null,
+    maximumRatio: required ? maximumRatio : null,
+    passed,
+    subscriptionCredits: "excluded; see comparison.serviceTierEvidence.subscriptionCreditAccounting"
+  };
+}

package/packages/piagent-core/benchmark/benchmark-artifact-binding.js

@@ -0,0 +1,49 @@
+import { createHash } from "node:crypto";
+import fs from "node:fs";
+import path from "node:path";
+
+const SHA256 = /^[a-f0-9]{64}$/;
+const relativePath = (value) => typeof value === "string" && value.length <= 512
+  && /^[a-zA-Z0-9_.-]+(?:\/[a-zA-Z0-9_.-]+)*$/.test(value)
+  && !value.split("/").some((part) => [".", "..", ".git"].includes(part));
+
+export function benchmarkArtifactBindingShapeErrors(bindings) {
+  if (!Array.isArray(bindings) || bindings.length < 1 || bindings.length > 64) return ["artifact-bindings-invalid"];
+  const ids = new Set(), paths = new Set(), errors = [];
+  for (const item of bindings) {
+    if (!item || typeof item !== "object" || Array.isArray(item) || Object.keys(item).length !== 3
+      || typeof item.id !== "string" || !/^[a-z0-9][a-z0-9-]{0,79}$/.test(item.id)
+      || !relativePath(item.path) || typeof item.sha256 !== "string" || !SHA256.test(item.sha256)
+      || ids.has(item.id) || paths.has(item.path)) errors.push("artifact-binding-invalid");
+    else { ids.add(item.id); paths.add(item.path); }
+  }
+  return errors;
+}
+
+/** Checks current source only. A historical archive can never satisfy this gate. */
+export function benchmarkArtifactBindingErrors(bindings, sourceRoot) {
+  const errors = benchmarkArtifactBindingShapeErrors(bindings);
+  if (errors.length) return errors;
+  let root;
+  try {
+    if (typeof sourceRoot !== "string" || !path.isAbsolute(sourceRoot)) throw new Error("Invalid root");
+    root = fs.realpathSync.native(sourceRoot);
+    if (!fs.statSync(root).isDirectory()) throw new Error("Invalid root");
+  } catch { return ["artifact-source-root-unavailable"]; }
+  for (const item of bindings) {
+    let fd;
+    try {
+      const file = path.join(root, item.path);
+      if (fs.realpathSync.native(file) !== file) throw new Error("Symlinked artifact");
+      fd = fs.openSync(file, fs.constants.O_RDONLY | fs.constants.O_NOFOLLOW | fs.constants.O_NONBLOCK);
+      const before = fs.fstatSync(fd);
+      if (!before.isFile() || before.size < 1 || before.size > 2 * 1024 * 1024) throw new Error("Invalid artifact size/type");
+      const bytes = fs.readFileSync(fd), after = fs.fstatSync(fd);
+      if (bytes.length !== before.size || before.size !== after.size || before.mtimeMs !== after.mtimeMs
+        || before.ctimeMs !== after.ctimeMs) throw new Error("Artifact changed while reading");
+      if (createHash("sha256").update(bytes).digest("hex") !== item.sha256) errors.push(`artifact-digest-mismatch:${item.path}`);
+    } catch { errors.push(`artifact-unavailable-or-unsafe:${item.path}`); }
+    finally { if (fd !== undefined) fs.closeSync(fd); }
+  }
+  return errors;
+}