mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-12 06:10:44 +01:00
## Thinking Path > - Paperclip is the open source app people use to manage AI agents for work > - Budget admission decides if new agent work can start in a company > - Five admission gates waited on the cleanup or accounting of other runs > - The restart-chaos campaign found that these gates froze companies for up to 10 minutes, caused false budget pauses, denied in-run writes, and dropped wakes, comments and Retries > - The durable ownership plan (Decision 1) removes the company-wide accounting gate. It does not narrow it again > - This pull request removes the five gates. Unfinished accounting now adds its bounded estimate to exposure > - The benefit is that one run's unsettled accounting never stops the other work in its company ## Linked Issues or Issue Description Refs #19831 (plan, section D), #19838 (inventory, finding 3), #19927 (Phase 2 reconciler data), #19497, #19724, #19826. **What happened?** Admission refused new work in these cases, with no reached hard stop and no pause: - a terminal run awaited accounting (`waitingRunCount`); - a native run recovered its accounting (`recoveringRunCount`); - a budget-owned scope hold awaited reconciliation, after the hard stop no longer applied; - a budget input was older than 15 s (`budget_evaluation_work`); - the predecessor run row of a task was missing (`hasPendingRunCleanup` failed closed). **Expected behavior** Admission never waits on the cleanup or accounting of another run. Only a reached hard stop, an explicit pause or an inactive company refuses new work (invariant 3). **Steps to reproduce** Make a company with an active budget policy. Leave one failed run with `cost_accounting_pending = true`. Request work for another agent. On `master` admission refuses with "Company cannot start work while completed runs await accounting." ## What Changed - `budgets.ts`: admission refuses only for a reached hard stop or an explicit pause. The code removes `observedBlocks`, the `recoveringRunCount` check, the stale `budgetHeld` refusal and the 15 s input-age gate. It also removes the unused `inspectInvocationEligibility`. `reconcileScope` no longer returns accounting holds. - `budget-exposure.ts`: removes the `waiting` and `recovering` aggregates and the own-task subquery. The admission query is smaller. Unfinished accounting stays exposure: the larger of the held reservation and the observed cost. - `budget-evaluation-work.ts`: removes `oldestBudgetInputAt`, which nothing reads now. - `execution-blocker.ts`: `hasPendingRunCleanup` returns false for a missing run row. A run row that is deleted removes its finalization coordinator and nulls its lease references, so the row owns nothing. A predecessor that owns a lease or a coordinator still holds the successor of its own task. - `recovery/ownership-invariant-report.ts`: the I3 check no longer mirrors the removed gate. It reports a wake request or a run that was refused with an accounting-hold reason. Any such report is a regression. - `doc/SPEC-implementation.md`: adds the product rule "Admission is independent of cleanup and accounting". `doc/execution-semantics.md` is aligned. ## Verification - New `server/src/__tests__/admission-independent-of-cleanup.test.ts`, on embedded Postgres. It has one case per gate. All six cases fail on `master` and pass here. A seventh case (an explicit pause still refuses) passes on both. One case also asserts invariant 3 together with exposure: two failed runs that each hold a 400-cent reservation fill a 1000-cent budget, and a hard stop still refuses. - Existing tests that asserted the old gates now assert admission (invariants 3 and 4): `closed-native-receipt-accounting`, `budgets-service`, `budget-evaluation-work`, `heartbeat-held-comment-wake`, `ownership-reconciliation-report`. In `heartbeat-held-comment-wake`, each saved comment, Retry, assignment and continuation is delivered exactly once while another run's accounting is pending. Settling accounting later does not deliver it again. - All six files pass locally, one file per process. The server package typecheck passes. - Speed: admission is on the turn path. This change only removes work from it: two correlated aggregates, one own-task subquery and one `budget_evaluation_work` read per policy. No wait or I/O is added before `turn.start`. - Not run here: the chaos waves (K13 storm, K6 crash loop, K4 long holds). The coordinating session schedules them on the shared harness. ## Risks - Behavior change: work starts while accounting for other runs settles. Spend stays bounded by the reservation capacity check and the hard stop, which read the ledger plus observed cost. A run that has no reservation and no observed cost adds zero exposure until its charge commits. This is the gap that Decision 1 accepts. Budget enforcement was already best effort (SPEC 11.6). - A budget-owned pause with no reached hard stop no longer refuses work in budget admission. The paused company or agent is still refused by the existing pause checks until the budget worker lifts the pause. ## Model Used - Claude Opus 5.5 (`claude-opus-5-5`), Anthropic, with tool use and code execution in Claude Code. ## 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 - [ ] All Paperclip CI gates are green - [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups - [x] I will address all Greptile and reviewer comments before requesting merge 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
222 lines
14 KiB
TypeScript
222 lines
14 KiB
TypeScript
import { randomUUID } from "node:crypto";
|
|
import { and, eq, sql } from "drizzle-orm";
|
|
import { afterAll, beforeAll, describe, expect, it } from "vitest";
|
|
import { agents, budgetEvaluationWork, budgetIncidents, budgetPolicies, companies, costEvents, createDb, projects } from "@paperclipai/db";
|
|
import { budgetService, budgetServiceInTransaction, flushBudgetEvaluationWork } from "../services/budgets.js";
|
|
import { costService, createCostEventInTransaction } from "../services/costs.js";
|
|
import { withAccountingTransaction, withOperationAccountingTransaction } from "../services/accounting-transaction.js";
|
|
import { computeObservedSpends, type SpendPolicy } from "../services/budget-exposure.js";
|
|
import { createDeliveryWorkCoordinator } from "../services/delivery-work-coordinator.js";
|
|
import { registerBudgetEnforcementWork } from "../services/budget-enforcement-worker.js";
|
|
import { getEmbeddedPostgresTestSupport, startEmbeddedPostgresTestDatabase, EMBEDDED_POSTGRES_TEST_TIMEOUT_MS } from "./helpers/embedded-postgres.js";
|
|
|
|
const support = await getEmbeddedPostgresTestSupport();
|
|
const postgres = support.supported ? describe : describe.skip;
|
|
|
|
postgres("independent accounting and deferred budget decisions", () => {
|
|
let temp: Awaited<ReturnType<typeof startEmbeddedPostgresTestDatabase>>;
|
|
let db: ReturnType<typeof createDb>;
|
|
beforeAll(async () => {
|
|
temp = await startEmbeddedPostgresTestDatabase("paperclip-budget-work-");
|
|
db = createDb(temp.connectionString);
|
|
}, EMBEDDED_POSTGRES_TEST_TIMEOUT_MS);
|
|
afterAll(async () => {
|
|
await db?.$client.end({ timeout: 1 });
|
|
await temp?.cleanup();
|
|
});
|
|
async function fixture(limit = 100) {
|
|
const companyId = randomUUID(), agentId = randomUUID();
|
|
await db.insert(companies).values({ id: companyId, name: "Independent accounting", issuePrefix: `X${companyId.slice(0, 6)}` });
|
|
await db.insert(agents).values({ id: agentId, companyId, name: "Worker", role: "engineer", status: "active", adapterType: "process" });
|
|
await budgetService(db).upsertPolicy(companyId, { scopeType: "company", scopeId: companyId, amount: limit }, null);
|
|
await flushBudgetEvaluationWork(db, { companyId });
|
|
return { companyId, agentId };
|
|
}
|
|
function charge(agentId: string, idempotencyKey = randomUUID(), costCents = 60) {
|
|
return { agentId, provider: "openai", biller: "openai", model: "gpt-4", billingType: "metered_api",
|
|
costCents, occurredAt: new Date(), idempotencyKey };
|
|
}
|
|
async function pending(companyId: string) {
|
|
return db.select().from(budgetEvaluationWork).where(eq(budgetEvaluationWork.companyId, companyId));
|
|
}
|
|
|
|
it("records every concurrent charge, then coalesces one recoverable hard-stop transition", async () => {
|
|
const f = await fixture();
|
|
await Promise.all([costService(db).createEvent(f.companyId, charge(f.agentId)), costService(db).createEvent(f.companyId, charge(f.agentId))]);
|
|
expect((await pending(f.companyId))).toHaveLength(2);
|
|
expect((await db.select().from(companies).where(eq(companies.id, f.companyId)))[0].pauseReason).toBeNull();
|
|
await Promise.all([flushBudgetEvaluationWork(db, f), flushBudgetEvaluationWork(db, f)]);
|
|
expect(await pending(f.companyId)).toHaveLength(0);
|
|
expect((await db.select().from(costEvents).where(eq(costEvents.companyId, f.companyId)))).toHaveLength(2);
|
|
expect((await db.select().from(companies).where(eq(companies.id, f.companyId)))[0].pauseReason).toBe("budget");
|
|
const incidents = await db.select().from(budgetIncidents).where(and(eq(budgetIncidents.companyId, f.companyId), eq(budgetIncidents.thresholdType, "hard")));
|
|
expect(incidents).toHaveLength(1);
|
|
expect((await budgetService(db).evaluateAdmission(f.companyId, f.agentId)).block).not.toBeNull();
|
|
});
|
|
|
|
it("keeps a lower allocated input that commits after a consumed batch", async () => {
|
|
const f = await fixture(1000);
|
|
let release!: () => void, inserted!: () => void;
|
|
const ready = new Promise<void>(resolve => { inserted = resolve; });
|
|
const gate = new Promise<void>(resolve => { release = resolve; });
|
|
const earlier = withOperationAccountingTransaction(db, f.companyId, async (tx, publications) => {
|
|
await createCostEventInTransaction(tx, f.companyId, charge(f.agentId), publications);
|
|
inserted();
|
|
await gate;
|
|
});
|
|
try {
|
|
await ready;
|
|
await costService(db).createEvent(f.companyId, charge(f.agentId));
|
|
expect(await flushBudgetEvaluationWork(db, f)).toBe(1);
|
|
release();
|
|
await earlier;
|
|
expect(await pending(f.companyId)).toHaveLength(1);
|
|
expect(await flushBudgetEvaluationWork(db, f)).toBe(1);
|
|
expect(await pending(f.companyId)).toHaveLength(0);
|
|
} finally { release(); await earlier; }
|
|
});
|
|
|
|
it("does not create more work or audit charges on identical receipt replay", async () => {
|
|
const f = await fixture();
|
|
const receipt = charge(f.agentId, randomUUID(), 1);
|
|
const events = await Promise.all([costService(db).createEvent(f.companyId, receipt), costService(db).createEvent(f.companyId, receipt)]);
|
|
expect(events[0].id).toBe(events[1].id);
|
|
expect(await pending(f.companyId)).toHaveLength(1);
|
|
await expect(costService(db).createEvent(f.companyId, { ...receipt, costCents: 2 })).rejects.toMatchObject({ status: 409 });
|
|
expect(await pending(f.companyId)).toHaveLength(1);
|
|
});
|
|
|
|
// Invariant 3: unevaluated inputs are not a gate. Admission reads the ledger
|
|
// and live exposure itself, so a reached hard stop refuses at once while the
|
|
// evaluator is still behind, and an old input under the limit never refuses.
|
|
it("admits limited work while inputs await evaluation, and refuses a reached hard stop inline", async () => {
|
|
const f = await fixture(1000);
|
|
await costService(db).createEvent(f.companyId, charge(f.agentId, randomUUID(), 1));
|
|
await db.update(budgetEvaluationWork).set({ createdAt: new Date(Date.now() - 16_000) }).where(eq(budgetEvaluationWork.companyId, f.companyId));
|
|
expect((await budgetService(db).evaluateAdmission(f.companyId, f.agentId)).block).toBeNull();
|
|
await costService(db).createEvent(f.companyId, charge(f.agentId, randomUUID(), 1000));
|
|
expect(await pending(f.companyId)).not.toHaveLength(0);
|
|
expect((await budgetService(db).evaluateAdmission(f.companyId, f.agentId)).block?.reason).toMatch(/hard-stop/);
|
|
await flushBudgetEvaluationWork(db, f);
|
|
expect((await budgetService(db).evaluateAdmission(f.companyId, f.agentId)).block?.reason).toMatch(/hard-stop/);
|
|
});
|
|
|
|
it("retains policy raise and manual pause when evaluation races them", async () => {
|
|
const f = await fixture();
|
|
await costService(db).createEvent(f.companyId, charge(f.agentId, randomUUID(), 120));
|
|
await Promise.all([flushBudgetEvaluationWork(db, f), budgetService(db).upsertPolicy(f.companyId,
|
|
{ scopeType: "company", scopeId: f.companyId, amount: 1000 }, null)]);
|
|
expect((await db.select().from(companies).where(eq(companies.id, f.companyId)))[0].pauseReason).toBeNull();
|
|
await db.update(companies).set({ status: "paused", pauseReason: "manual", pausedAt: new Date() }).where(eq(companies.id, f.companyId));
|
|
await costService(db).createEvent(f.companyId, charge(f.agentId, randomUUID(), 1));
|
|
await flushBudgetEvaluationWork(db, f);
|
|
expect((await db.select().from(companies).where(eq(companies.id, f.companyId)))[0].pauseReason).toBe("manual");
|
|
});
|
|
|
|
it("does not apply another agent's unevaluated hard stop to a separately limited agent", async () => {
|
|
const f = await fixture(0);
|
|
const otherAgentId = randomUUID();
|
|
await db.insert(agents).values({ id: otherAgentId, companyId: f.companyId, name: "Other worker", role: "engineer", status: "active", adapterType: "process" });
|
|
const service = budgetService(db);
|
|
for (const agentId of [f.agentId, otherAgentId]) await service.upsertPolicy(f.companyId,
|
|
{ scopeType: "agent", scopeId: agentId, amount: 1000 }, null);
|
|
await flushBudgetEvaluationWork(db, f);
|
|
await costService(db).createEvent(f.companyId, charge(f.agentId, randomUUID(), 1000));
|
|
await db.update(budgetEvaluationWork).set({ createdAt: new Date(Date.now() - 16_000) }).where(eq(budgetEvaluationWork.companyId, f.companyId));
|
|
// The unevaluated charge reaches this agent's hard stop and refuses it
|
|
// inline; the other agent's separate limit is untouched.
|
|
expect((await service.evaluateAdmission(f.companyId, f.agentId)).block?.reason).toMatch(/hard-stop/);
|
|
expect((await service.evaluateAdmission(f.companyId, otherAgentId)).block).toBeNull();
|
|
});
|
|
|
|
it("evaluates many distinct scopes without exceeding PostgreSQL's aggregate column limit", async () => {
|
|
const f = await fixture(1000);
|
|
await costService(db).createEvent(f.companyId, charge(f.agentId, randomUUID(), 1));
|
|
const policies: SpendPolicy[] = [{ companyId: f.companyId, scopeType: "company", scopeId: f.companyId,
|
|
windowKind: "lifetime", metric: "billed_cents" }, ...Array.from({ length: 450 }, () => ({
|
|
companyId: f.companyId, scopeType: "project", scopeId: randomUUID(), windowKind: "lifetime", metric: "billed_cents",
|
|
}))];
|
|
const observed = await withOperationAccountingTransaction(db, f.companyId, tx => computeObservedSpends(tx, policies));
|
|
expect(observed).toHaveLength(451);
|
|
expect(observed[0].ledgerCents).toBe("1.0000000");
|
|
expect(observed.slice(1).every(row => row.totalExact === "0.0000000")).toBe(true);
|
|
});
|
|
|
|
it("proves the legacy mutex waits while the independent ledger transaction commits", async () => {
|
|
const f = await fixture(1000);
|
|
let release!: () => void, locked!: () => void;
|
|
const ready = new Promise<void>(resolve => { locked = resolve; });
|
|
const gate = new Promise<void>(resolve => { release = resolve; });
|
|
const holding = db.transaction(async tx => {
|
|
await tx.select().from(companies).where(eq(companies.id, f.companyId)).for("no key update");
|
|
locked(); await gate;
|
|
});
|
|
let legacyFinished = false;
|
|
let legacy: Promise<unknown> | undefined;
|
|
try {
|
|
await ready;
|
|
legacy = withAccountingTransaction(db, f.companyId, async () => { legacyFinished = true; });
|
|
await costService(db).createEvent(f.companyId, charge(f.agentId));
|
|
expect(legacyFinished).toBe(false);
|
|
expect(await pending(f.companyId)).toHaveLength(1);
|
|
} finally { release(); await holding; await legacy; }
|
|
expect(legacyFinished).toBe(true);
|
|
});
|
|
|
|
it("reevaluates a new policy committed after a no-policy batch consumed its charge", async () => {
|
|
const f = await fixture(1000);
|
|
const projectId = randomUUID();
|
|
await db.insert(projects).values({ id: projectId, companyId: f.companyId, name: "New policy scope" });
|
|
let release!: () => void, created!: () => void;
|
|
const ready = new Promise<void>(resolve => { created = resolve; });
|
|
const gate = new Promise<void>(resolve => { release = resolve; });
|
|
const policy = withOperationAccountingTransaction(db, f.companyId, async (tx, publications) => {
|
|
await budgetServiceInTransaction(tx, publications, { evaluate: true }).upsertPolicy(f.companyId,
|
|
{ scopeType: "project", scopeId: projectId, amount: 80 }, null);
|
|
created(); await gate;
|
|
});
|
|
async function bounded<T>(work: Promise<T>, phase: string): Promise<T> {
|
|
let timer: ReturnType<typeof setTimeout>;
|
|
try { return await Promise.race([work, new Promise<never>((_, reject) => {
|
|
timer = setTimeout(() => reject(new Error(`Policy race blocked at ${phase}`)), 2000);
|
|
})]); } finally { clearTimeout(timer!); }
|
|
}
|
|
try {
|
|
await bounded(Promise.race([ready, policy]), "policy prepared");
|
|
await bounded(costService(db).createEvent(f.companyId, { ...charge(f.agentId, randomUUID(), 120), projectId }), "charge commit");
|
|
expect(await bounded(flushBudgetEvaluationWork(db, f), "first evaluation")).toBe(1);
|
|
release(); await policy;
|
|
expect(await pending(f.companyId)).toHaveLength(1);
|
|
await flushBudgetEvaluationWork(db, f);
|
|
expect((await db.select().from(projects).where(eq(projects.id, projectId)))[0].pauseReason).toBe("budget");
|
|
} finally { release(); await policy; }
|
|
});
|
|
|
|
it("delivers another company's Stop when a busy company has more than a page of old inputs", async () => {
|
|
const busy = await fixture(), healthy = await fixture();
|
|
await costService(db).createEvent(busy.companyId, charge(busy.agentId, randomUUID(), 120));
|
|
await costService(db).createEvent(healthy.companyId, charge(healthy.agentId, randomUUID(), 120));
|
|
await db.insert(budgetEvaluationWork).values(Array.from({ length: 150 }, () => ({
|
|
companyId: busy.companyId, agentId: busy.agentId, createdAt: new Date(Date.now() - 20_000),
|
|
})));
|
|
let release!: () => void, locked!: () => void;
|
|
const ready = new Promise<void>(resolve => { locked = resolve; });
|
|
const gate = new Promise<void>(resolve => { release = resolve; });
|
|
const holding = db.transaction(async tx => {
|
|
await tx.select({ id: companies.id }).from(companies).where(eq(companies.id, busy.companyId)).for("no key update");
|
|
locked(); await gate;
|
|
});
|
|
const errors: unknown[] = [], delivered: string[] = [];
|
|
const coordinator = createDeliveryWorkCoordinator({ owner: db, canRun: () => true, onError: error => errors.push(error) });
|
|
try {
|
|
await ready;
|
|
await registerBudgetEnforcementWork(db, coordinator, { cancelWorkForScope: async scope => { delivered.push(scope.companyId); } }).ready;
|
|
expect(errors).toHaveLength(0);
|
|
expect(delivered).toContain(healthy.companyId);
|
|
expect((await db.select().from(companies).where(eq(companies.id, healthy.companyId)))[0].pauseReason).toBe("budget");
|
|
expect(await pending(healthy.companyId)).toHaveLength(0);
|
|
expect((await pending(busy.companyId)).length).toBeGreaterThan(100);
|
|
} finally { release(); await holding; await coordinator.stop(); }
|
|
});
|
|
});
|