Files
PaperClipAI/tests/runner-e2e/operating-overhead.test.ts
T
DottaandPaperclip cff331f6ca fix(evals): retain local controller cost origin (#17990)
## Thinking Path

> - Paperclip manages AI agents and their work.
> - Its eval journals retain model costs and operating reservations
under one cap.
> - The operating policy records the controller's execution origin.
> - The schema currently accepts only GitHub-hosted and AWS Fleet
origins.
> - A local controller must therefore use an inaccurate label or fail
admission.
> - This PR adds the local origin and keeps the hosted-only schedule
gate.

## Linked Issues or Issue Description

Refs #17986. No duplicate open PR or issue was found in the
operating-overhead/local search.

**What existing behavior does this improve?**

The Product E2E operating allowance and its retained journal policy.

**Current behavior**

At master `40c1ab36cd`, an unscheduled local controller cannot set
`runner: "local"`. The generic allowance validator rejects it. Local
qualification can use the same accounting controls but cannot record its
origin correctly.

**Proposed behavior**

Accept the local origin in the generic allowance. Keep all component,
price, cap, expiry, and receipt rules. The scheduled validator still
requires `github-hosted` and rejects a local allowance.

**Reason and benefit**

Local qualification can retain an accurate execution origin without
claiming that its reservations are billed GitHub charges.

## What Changed

- Add `local` to the operating allowance's runner type and validator.
- Exercise local journal creation, model admission, retained overhead,
restart, and reconciliation in the existing boundary test.
- Check that a local allowance fails Product scheduled admission.
- Pin new hosted Runner controller admission to `github-hosted`.
Preserve original labels when reading historical manifests; those reads
cannot authorize a new worker.
- Exercise the hosted-origin rejection and historical/local reads in
both source and copied portable runtimes.
- Document the local label and its accounting limits.

## Verification

- Before the fix: the new local persistence case and schedule-boundary
assertion failed; eight tests passed.
- Focused ledger, overhead, operations, and scheduled finalization
tests: 71 passed across five files.
- Controlled Runner source and portable tests: 86 passed. The new
hosted-origin regression failed before the fix.
- Full Product support suite at the first commit: 4,524 passed, three
skipped, and one retained embedded PostgreSQL bootstrap failure. The
same database failure was previously reproduced on a clean baseline.
- Separate Node test phase on the host: 137 passed.
- Actual Linux controller: database lifecycle tests pass. After
installing its missing `zip` utility, the full Product suite has 4,527
passes and one Copilot parent-version coverage failure. That same
failure reproduces on clean master `40c1ab36cd` with the same Linux
filesystem and dependencies (14 passes, one failure). Copilot files are
unchanged. The initial seven ZIP-related failures and the watcher
failure remain retained.
- The final fresh Linux semantic check passes: `GOMEMLIMIT=2GiB GOGC=50
GOMAXPROCS=1 node node_modules/typescript/bin/tsc --singleThreaded -p
tests/runner-e2e/tsconfig.json`. The parallel compiler OOM is retained.
No semantic checks were omitted.
- Both fresh reviews are 5/5, with the addressed thread resolved. Hosted
Build first failed during unchanged Pi setup prefetch with an operation
timeout. One job rerun recovered; the original failure remains retained.
All 58 latest-head checks are now acceptable.
- No paid model requests, sandbox creation, schedule changes, or
historical journal edits.

## Risks

- The caller supplies this origin. It must describe the actual
controller.
- `local` does not prove zero compute cost, provide a default allowance,
or release a hold.
- Both maintained hosted admission paths require the GitHub-hosted
controller origin. A previously accepted hosted policy with an AWS/local
label now fails new admission. Historical reads preserve the original
evidence.

## Model Used

- OpenAI GPT-6 through Codex, with tool use and code execution. The
exact serving model ID and context window are not exposed in this
session.

## Checklist

- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-10-28 02:05:36 -05:00

171 lines
13 KiB
TypeScript

import { mkdtemp, rm } from "node:fs/promises";
import { tmpdir } from "node:os";
import path from "node:path";
import { expect, it } from "vitest";
import { EvalSpendState, openEvalCostLedger, readEvalCostLedger, type EvalSpendLimits } from "./cost-ledger.js";
import { EVAL_OVERHEAD_COMPONENTS, operatingReservationBound, scheduledOperatingOverhead, validateOperatingOverhead, type EvalOperatingOverhead } from "./operating-overhead.js";
import { evalAccountingStatus, evalReadinessStatus, readEvalOperationsReport, renderEvalOperationsReport } from "./operations-report.js";
import { costBoundConfigurationSha256 } from "./cost-bound-evidence.js";
const now = Date.parse("2026-10-21T00:00:00Z");
const allowance = (): EvalOperatingOverhead => ({ version: 1, reference: "fixture-only", runner: "aws-fleet",
verifiedAt: "2026-10-20T00:00:00Z", expiresAt: "2026-11-19T00:00:00Z",
components: EVAL_OVERHEAD_COMPONENTS.map(component => ({ component, maxUsd: "1", reference: `fixture-${component}` })) });
const currentAllowance = (time = Date.now()): EvalOperatingOverhead => ({ ...allowance(),
verifiedAt: new Date(time - 86_400_000).toISOString(), expiresAt: new Date(time + 30 * 86_400_000).toISOString() });
const runner = "ubuntu-latest";
const modelIdentity = { executionId: "claude.local.fixture", attemptId: "attempt-1", phase: "agent" as const };
it.each(["aws-fleet", "local"] as const)("keeps %s overhead inside the same cap through model completion and a journal restart", async runnerKind => {
const directory = await mkdtemp(path.join(tmpdir(), "eval-overhead-"));
const file = path.join(directory, "ledger.jsonl");
const limits: EvalSpendLimits = { campaignId: "fixture", maxUsd: "20", deadlineAt: "2099-01-01T00:00:00Z",
maxRequests: 10, maxConcurrent: 1, maxAttemptsPerExecution: 1, overhead: { ...currentAllowance(), runner: runnerKind } };
try {
const ledger = await openEvalCostLedger(file, limits);
try {
expect((await readEvalCostLedger(file)).state.snapshot()).toMatchObject({ knownUsd: "0.000000000", heldUsd: "4.000000000", inflight: 0, unknownRequests: 4 });
await ledger.admitAttempt(modelIdentity.executionId, modelIdentity.attemptId);
await expect(ledger.reserve("too-large", modelIdentity, "17")).rejects.toThrow("spending cap");
await ledger.reserve("model", modelIdentity, "16");
await ledger.observe({ observationId: "model-receipt", requestId: "model", source: "provider", provider: "anthropic",
currency: "USD", route: "direct", classification: "reported", costUsd: "1", final: true });
await ledger.finish("model", "succeeded");
expect(ledger.snapshot()).toMatchObject({ knownUsd: "1.000000000", heldUsd: "4.000000000", committedUsd: "5.000000000" });
} finally { await ledger.close(); }
const changed = structuredClone(limits); changed.overhead!.components[0]!.maxUsd = "0";
await expect(openEvalCostLedger(file, changed)).rejects.toThrow("limits changed");
const resumed = await openEvalCostLedger(file, limits);
try {
await resumed.finish("overhead:ci_compute", "cancelled");
expect(resumed.snapshot().heldUsd).toBe("4.000000000");
await resumed.observe({ observationId: "ci-partial", requestId: "overhead:ci_compute", source: "infrastructure", provider: "aws",
currency: "USD", route: "infrastructure", classification: "estimated", pricingVersion: "fixture", costUsd: "0.2", final: false });
expect(resumed.snapshot()).toMatchObject({ knownUsd: "1.200000000", heldUsd: "3.800000000", committedUsd: "5.000000000", unknownRequests: 4 });
await resumed.observe({ observationId: "ci-invoice", requestId: "overhead:ci_compute", source: "invoice", provider: "aws",
currency: "USD", route: "infrastructure", classification: "reconciled", costUsd: "0.5", final: true });
expect(resumed.snapshot()).toMatchObject({ knownUsd: "1.500000000", heldUsd: "3.000000000", committedUsd: "4.500000000", unknownRequests: 3 });
} finally { await resumed.close(); }
const report = await readEvalOperationsReport(file);
expect(report).toMatchObject({ requests: 5, unresolvedRequests: 3, knownUsd: "1.500000000", heldUsd: "3.000000000" });
expect(renderEvalOperationsReport(report)).toContain("campaign-overhead");
expect(renderEvalOperationsReport(report)).toContain("Held reservations are not charges");
} finally { await rm(directory, { recursive: true, force: true }); }
});
it("stops further model admission when a later infrastructure receipt exceeds the shared cap", async () => {
const directory = await mkdtemp(path.join(tmpdir(), "eval-overhead-overrun-"));
try {
const ledger = await openEvalCostLedger(path.join(directory, "ledger.jsonl"), { campaignId: "overrun", maxUsd: "20",
deadlineAt: "2099-01-01T00:00:00Z", maxRequests: 10, maxConcurrent: 1, maxAttemptsPerExecution: 1, overhead: currentAllowance() });
try {
await ledger.admitAttempt(modelIdentity.executionId, modelIdentity.attemptId);
await ledger.observe({ observationId: "late-invoice", requestId: "overhead:ci_compute", source: "invoice", provider: "aws",
currency: "USD", route: "infrastructure", classification: "reconciled", costUsd: "21", final: true });
expect(ledger.snapshot()).toMatchObject({ overBudget: true, committedUsd: "24.000000000" });
await expect(ledger.reserve("next", modelIdentity, "0.1")).rejects.toThrow("spending cap");
} finally { await ledger.close(); }
} finally { await rm(directory, { recursive: true, force: true }); }
});
it("requires each cost category once and rejects unknown policy fields or missing provenance", () => {
expect(validateOperatingOverhead(allowance(), "20")).toBe(4_000_000_000n);
expect(() => validateOperatingOverhead({ ...allowance(), components: allowance().components.slice(1) }, "20")).toThrow("Incomplete");
const duplicate = allowance(); duplicate.components[0] = duplicate.components[1]!;
expect(() => validateOperatingOverhead(duplicate, "20")).toThrow("component");
expect(() => validateOperatingOverhead({ ...allowance(), apiKey: "never-allowed" } as never, "20")).toThrow("fields");
const missing = allowance(); missing.components[0]!.reference = "";
expect(() => validateOperatingOverhead(missing, "20")).toThrow("component");
expect(() => validateOperatingOverhead(null as never, "20")).toThrow("fields");
expect(() => validateOperatingOverhead({ ...allowance(), expiresAt: "2027-10-20T00:00:00Z" }, "20")).toThrow("Incomplete");
expect(() => validateOperatingOverhead(allowance(), "4")).toThrow("no model allowance");
});
it("requires fresh reviewed allowances tied to the actual CI runner before scheduled work", () => {
const hosted = { ...allowance(), runner: "github-hosted" as const };
const raw = JSON.stringify(hosted);
expect(scheduledOperatingOverhead(raw, runner, "20", now)).toEqual(hosted);
expect(() => scheduledOperatingOverhead(undefined, runner, "20", now)).toThrow("Missing reviewed");
expect(() => scheduledOperatingOverhead("not json", runner, "20", now)).toThrow("JSON");
expect(() => scheduledOperatingOverhead(JSON.stringify(allowance()), runner, "20", now)).toThrow("runner");
expect(() => scheduledOperatingOverhead(JSON.stringify({ ...hosted, runner: "local" }), runner, "20", now)).toThrow("runner");
expect(() => scheduledOperatingOverhead(raw, "runs-on/fleet=paperclip-public-pr-x64/env=public-ci", "20", now)).toThrow("runner");
expect(() => scheduledOperatingOverhead(raw, "unrecognized", "20", now)).toThrow("runner");
expect(() => scheduledOperatingOverhead(raw, runner, "20", Date.parse("2026-10-19T23:59:59Z"))).toThrow("needs review");
expect(() => scheduledOperatingOverhead(raw, runner, "20", Date.parse("2026-11-18T20:00:00Z"))).toThrow("needs review");
});
it("does not turn a zero allowance into an observed zero charge", async () => {
const directory = await mkdtemp(path.join(tmpdir(), "eval-overhead-zero-"));
const overhead = allowance(); overhead.components[0]!.maxUsd = "0";
try {
const ledger = await openEvalCostLedger(path.join(directory, "ledger.jsonl"), { campaignId: "zero", maxUsd: "20",
deadlineAt: "2099-01-01T00:00:00Z", maxRequests: 10, maxConcurrent: 1, maxAttemptsPerExecution: 1, overhead });
try {
expect(ledger.snapshot().requests.find(row => row.requestId === "overhead:ci_compute"))
.toMatchObject({ costUsd: null, classification: "unknown", receiptComplete: false, reservedUsd: "0.000000000" });
} finally { await ledger.close(); }
} finally { await rm(directory, { recursive: true, force: true }); }
});
it("stops new admission after allowance expiry while preserving late invoice reconciliation", () => {
const state = new EvalSpendState({ campaignId: "expired", maxUsd: "20", deadlineAt: "2099-01-01T00:00:00Z",
maxRequests: 10, maxConcurrent: 1, maxAttemptsPerExecution: 1, overhead: allowance() });
const admitted = { type: "attempt" as const, executionId: modelIdentity.executionId, attemptId: modelIdentity.attemptId };
state.validate(admitted, now); state.apply(admitted);
const expiredAt = Date.parse(allowance().expiresAt);
expect(() => state.validate({ ...admitted, attemptId: "next" }, expiredAt)).toThrow("needs review");
expect(() => state.validate({ type: "reserve", requestId: "model", identity: modelIdentity, maxUsd: "1" }, expiredAt)).toThrow("needs review");
expect(() => state.validate({ type: "observe", observation: { observationId: "late-invoice", requestId: "overhead:ci_compute",
source: "invoice", provider: "aws", currency: "USD", route: "infrastructure", classification: "reconciled", costUsd: "0.5", final: true } }, expiredAt)).not.toThrow();
});
it("separates model readiness from unresolved overhead without releasing holds or masking model gaps", async () => {
const directory = await mkdtemp(path.join(tmpdir(), "eval-model-readiness-"));
const file = path.join(directory, "ledger.jsonl");
const overhead = { ...currentAllowance(), accountingReadiness: "model_requests" as const };
const ledger = await openEvalCostLedger(file, { campaignId: "model-readiness", maxUsd: "20",
deadlineAt: "2099-01-01T00:00:00Z", maxRequests: 10, maxConcurrent: 1, maxAttemptsPerExecution: 1, overhead });
try {
expect(evalReadinessStatus(await readEvalOperationsReport(file))).toBe("blocked");
await ledger.admitAttempt(modelIdentity.executionId, modelIdentity.attemptId);
await ledger.reserve("model", modelIdentity, "2");
expect(evalReadinessStatus(await readEvalOperationsReport(file))).toBe("blocked");
await ledger.finish("model", "cancelled");
expect(evalReadinessStatus(await readEvalOperationsReport(file))).toBe("blocked");
await ledger.observe({ observationId: "model-final", requestId: "model", source: "provider", provider: "anthropic",
currency: "USD", route: "direct", classification: "reported", costUsd: "0.5", final: true });
const report = await readEvalOperationsReport(file);
expect(evalAccountingStatus(report)).toBe("blocked");
expect(evalReadinessStatus(report)).toBe("receipts_complete");
expect(evalReadinessStatus({ ...report, accountingReadiness: undefined } as never)).toBe("blocked");
expect(report).toMatchObject({ accountingReadiness: "model_requests", heldUsd: "4.000000000",
committedUsd: "4.500000000", modelAccounting: { requests: 1, knownUsd: "0.500000000", unresolvedRequests: 0 } });
expect(renderEvalOperationsReport(report)).toContain("Operating costs remain separately unreconciled");
await ledger.observe({ observationId: "overrun", requestId: "overhead:ci_compute", source: "invoice", provider: "aws",
currency: "USD", route: "infrastructure", classification: "reconciled", costUsd: "21", final: true });
expect(evalReadinessStatus(await readEvalOperationsReport(file))).toBe("blocked");
} finally { await ledger.close(); await rm(directory, { recursive: true, force: true }); }
});
it("rejects unknown readiness policies instead of silently loosening accounting", () => {
expect(() => validateOperatingOverhead({ ...allowance(), accountingReadiness: "ignore_costs" } as never, "20"))
.toThrow("Invalid accounting readiness policy");
});
it("binds optional readiness policy to reviewed operating evidence without changing legacy hash recipes", () => {
for (const accountingReadiness of [undefined, "model_requests"] as const) {
const value = { ...allowance(), ...(accountingReadiness ? { accountingReadiness } : {}) };
const component = value.components[0]!;
const calculation = { reference: value.reference, runner: value.runner, verifiedAt: value.verifiedAt,
expiresAt: value.expiresAt, component: { ...component }, ...(accountingReadiness ? { accountingReadiness } : {}) };
component.provisionalBound = { version: 1, scope: "full_lifecycle", evidenceSha256: "a".repeat(64),
configurationSha256: costBoundConfigurationSha256(calculation) };
expect(operatingReservationBound(value, component)?.calculationSha256).toBe(component.provisionalBound.configurationSha256);
const changed = { ...value, accountingReadiness: accountingReadiness ? undefined : "model_requests" as const };
expect(() => operatingReservationBound(changed, component)).toThrow("Reviewed cost bound configuration changed");
}
});