PaperClipAI/tests/runner-e2e/chat-catalog-domains/runtime-email-reply.test.ts

import { expect, it } from "vitest";
import { createHash } from "node:crypto";
import { gradeReplyEmail, selectedReviewReadThenReply, truthfulUnverifiedMobile } from "./runtime-email-reply.js";
const hash = (v: string) => createHash("sha256").update(v).digest("hex");
function valid() { return { companyId: "company", leadId: "ceo", scenarioId: "chat-channels.reply-to-email", origin: "http://127.0.0.1:42123", reply: "", agents: [], tasks: [], activity: [], links: [], humanQueries: [], runs: [1,2,3,4,5,6,7].map(id => ({ id: String(id), agentId: "ceo", status: "succeeded" })), state: { actualTwoInboundBound: true, actualThreadRetained: true, exactLatestReplyBound: true, independentProviderCount: 1, oneSavedPublication: true, brokenLayoutReallyObserved: true, noPrematureReply: true, unverifiedReadBound: true, unverifiedReply: "The current mobile layout has not been verified as fixed. I cannot truthfully send that claim yet.", actualMobileFixBound: true, actualReadBeforeReply: true, normalReplyNativeBound: true, nativeLedgerCausal: true, factualReplyBound: true, factualReply: "The latest reviewer question received the reply and the actual provider accepted it. Recipient delivery is unconfirmed.", actualOutcome: "sent" } }; }
it("accepts both genuinely unverified and actual verified same-thread reply phases", () => expect(gradeReplyEmail(valid()).passed).toBe(true));
it("accepts a natural substantive qualification and provider-outcome paraphrase", () => { const o = valid(); o.state.unverifiedReply = "I checked the current source: mobile still overflows. A fix is pending, so I won't tell the reviewer it is repaired."; o.state.factualReply = "Your response to the newest question was sent. I cannot confirm whether it was delivered."; expect(gradeReplyEmail(o).passed).toBe(true); });
it("rejects plausible false mobile-fix claims even when surrounded by caveats", () => { expect(truthfulUnverifiedMobile("The layout is now fixed, but it remains unverified.")).toBe(false); const o = valid(); o.state.unverifiedReply += " The mobile layout is already fixed."; expect(gradeReplyEmail(o).passed).toBe(false); expect(truthfulUnverifiedMobile("I cannot confirm whether the mobile layout is fixed.")).toBe(true); });
it("rejects earlier/foreign thread selection, a premature email, missing actual repair or guessed native send", () => { for (const key of ["exactLatestReplyBound", "noPrematureReply", "actualMobileFixBound", "normalReplyNativeBound", "actualReadBeforeReply", "actualThreadRetained"]) { const o = valid(); (o.state as Record<string, unknown>)[key] = false; expect(gradeReplyEmail(o).passed).toBe(false); } });
it("rejects duplicate sends, unrelated native phases and false delivery claims", () => { const o = valid(); o.state.independentProviderCount = 2; expect(gradeReplyEmail(o).passed).toBe(false); const p = valid(); p.state.nativeLedgerCausal = false; expect(gradeReplyEmail(p).passed).toBe(false); const q = valid(); q.state.factualReply += " It was delivered."; expect(gradeReplyEmail(q).passed).toBe(false); });
const expected = { issueId: "task", conversationId: "conversation", earlierId: "earlier", latestId: "latest", publicationId: "publication", documentId: "document", documentSha256: hash("real fixed source") };
function tool(id: string, name: string, input: unknown, result: unknown, seq: number) { return [{ seq, eventType: "item.started", payload: { prpEvent: { payload: { item: { id, type: "tool_use", name, input } } } } }, { seq: seq + 1, eventType: "item.completed", payload: { prpEvent: { payload: { item: { id, type: "tool_result", result } } } } }, { seq: seq + 2, eventType: "tool.execution.completed", payload: { prpEvent: { payload: { executionId: id, name, status: "completed" } } } }]; }
function actual() { return [...tool("thread", "agentmail_read_thread", {}, { issueId: "task", conversationId: "conversation", messages: [{ providerMessageId: "earlier", direction: "inbound" }, { providerMessageId: "latest", direction: "inbound" }] }, 1), ...tool("source", "read_document", { key: "mobile-layout" }, { document: { id: "document", issueId: "task", body: "real fixed source" } }, 4), ...tool("send", "agentmail_send", { request: { conversationId: "conversation", replyToMessageId: "latest", idempotencyKey: "publication" } }, { id: "publication" }, 7)]; }
it("binds actual production read_document wrapper and selected thread reads before the completed native reply", () => expect(selectedReviewReadThenReply(actual(), expected)).toEqual({ selectedRead: true, replyBound: true }));
it("rejects missing/wrong evidence, earlier message reply and uncompleted synthetic results", () => { expect(selectedReviewReadThenReply(actual(), { ...expected, latestId: "wrong" }).replyBound).toBe(false); expect(selectedReviewReadThenReply(actual(), { ...expected, documentSha256: hash("wrong source") }).replyBound).toBe(false); expect(selectedReviewReadThenReply(actual().slice(0, -1), expected).replyBound).toBe(false); const late = actual(); late[3]!.seq = 10; expect(selectedReviewReadThenReply(late, expected).replyBound).toBe(false); });

it("allows the unchanged selected message from actual prior read evidence but never substitutes that for current repaired source", () => { const withoutThread = actual().slice(3); expect(selectedReviewReadThenReply(withoutThread, { ...expected, priorSelectedThreadRead: true }).replyBound).toBe(true); expect(selectedReviewReadThenReply(withoutThread.slice(3), { ...expected, priorSelectedThreadRead: true }).replyBound).toBe(false); });