@vellumai/assistant 0.10.0 → 0.10.1-dev.202606240317.ea25efe

package/src/__tests__/guardian-grant-minting.test.ts

@@ -1,607 +0,0 @@
-/**
- * Tests for M3: scoped grant minting on guardian tool-approval decisions.
- *
- * When a guardian approves a tool-approval request (one with toolName + input),
- * the approval interception flow should mint a `tool_signature` scoped grant.
- * Non-tool-approval requests and rejections must NOT mint grants.
- */
-
-import { beforeEach, describe, expect, mock, spyOn, test } from "bun:test";
-
-// ---------------------------------------------------------------------------
-// Test isolation: in-memory SQLite via temp directory
-// ---------------------------------------------------------------------------
-
-mock.module("../util/logger.js", () => ({
-  getLogger: () =>
-    new Proxy({} as Record<string, unknown>, {
-      get: () => () => {},
-    }),
-}));
-
-const _conversationMocks = new Map<string, unknown>();
-mock.module("../daemon/conversation-registry.js", () => ({
-  findConversation: (id: string) => _conversationMocks.get(id),
-}));
-
-import { GRANT_TTL_MS } from "../approvals/guardian-decision-primitive.js";
-import type { Conversation } from "../daemon/conversation.js";
-import type { TrustContext } from "../daemon/trust-context.js";
-import { getDb } from "../memory/db-connection.js";
-import { initializeDb } from "../memory/db-init.js";
-import {
-  createApprovalRequest,
-  type GuardianApprovalRequest,
-} from "../memory/guardian-approvals.js";
-import * as approvalMessageComposer from "../runtime/approval-message-composer.js";
-import * as gatewayClient from "../runtime/gateway-client.js";
-import * as pendingInteractions from "../runtime/pending-interactions.js";
-import { handleApprovalInterception } from "../runtime/routes/guardian-approval-interception.js";
-import { computeToolApprovalDigest } from "../security/tool-approval-digest.js";
-
-import "../memory/scoped-approval-grants.js";
-
-initializeDb();
-
-// ---------------------------------------------------------------------------
-// Helpers
-// ---------------------------------------------------------------------------
-
-const ASSISTANT_ID = "self";
-const GUARDIAN_USER = "guardian-user-1";
-const GUARDIAN_CHAT = "guardian-chat-1";
-const REQUESTER_USER = "requester-user-1";
-const REQUESTER_CHAT = "requester-chat-1";
-const CONVERSATION_ID = "conv-1";
-const TOOL_NAME = "execute_shell";
-const TOOL_INPUT = { command: "rm -rf /tmp/test" };
-
-function resetTables(): void {
-  try {
-    const db = getDb();
-    db.run("DELETE FROM channel_guardian_approval_requests");
-    db.run("DELETE FROM scoped_approval_grants");
-  } catch {
-    /* tables may not exist yet */
-  }
-  pendingInteractions.clear();
-}
-
-function createTestGuardianApproval(
-  requestId: string,
-  overrides: Partial<Parameters<typeof createApprovalRequest>[0]> = {},
-): GuardianApprovalRequest {
-  return createApprovalRequest({
-    runId: `run-${requestId}`,
-    requestId,
-    conversationId: CONVERSATION_ID,
-    channel: "telegram",
-    requesterExternalUserId: REQUESTER_USER,
-    requesterChatId: REQUESTER_CHAT,
-    guardianExternalUserId: GUARDIAN_USER,
-    guardianChatId: GUARDIAN_CHAT,
-    toolName: TOOL_NAME,
-    expiresAt: Date.now() + 300_000,
-    ...overrides,
-  });
-}
-
-function registerPendingInteraction(
-  requestId: string,
-  conversationId: string,
-  toolName: string,
-  input: Record<string, unknown> = TOOL_INPUT,
-): ReturnType<typeof mock> {
-  const handleConfirmationResponse = mock(() => {});
-  const _mockSession = {
-    handleConfirmationResponse,
-    ensureActorScopedHistory: async () => {},
-  } as unknown as Conversation;
-  _conversationMocks.set(conversationId, _mockSession);
-
-  pendingInteractions.register(requestId, {
-    conversationId,
-    kind: "confirmation",
-    confirmationDetails: {
-      toolName,
-      input,
-      riskLevel: "high",
-      allowlistOptions: [
-        { label: "test", description: "test", pattern: "test" },
-      ],
-      scopeOptions: [{ label: "everywhere", scope: "everywhere" }],
-    },
-  });
-
-  return handleConfirmationResponse;
-}
-
-function makeTrustContext(): TrustContext {
-  return {
-    sourceChannel: "telegram",
-    trustClass: "guardian",
-  };
-}
-
-function countGrants(): number {
-  try {
-    const db = getDb();
-    const row = db.$client
-      .prepare("SELECT count(*) as cnt FROM scoped_approval_grants")
-      .get() as { cnt: number };
-    return row.cnt;
-  } catch {
-    return 0;
-  }
-}
-
-function getLatestGrant(): Record<string, unknown> | null {
-  try {
-    const db = getDb();
-    const row = db.$client
-      .prepare(
-        "SELECT * FROM scoped_approval_grants ORDER BY created_at DESC LIMIT 1",
-      )
-      .get();
-    return (row as Record<string, unknown>) ?? null;
-  } catch {
-    return null;
-  }
-}
-
-// ═══════════════════════════════════════════════════════════════════════════
-// Tests
-// ═══════════════════════════════════════════════════════════════════════════
-
-describe("guardian grant minting on tool-approval decisions", () => {
-  let deliverSpy: ReturnType<typeof spyOn>;
-  let composeSpy: ReturnType<typeof spyOn>;
-
-  beforeEach(() => {
-    resetTables();
-    deliverSpy = spyOn(gatewayClient, "deliverChannelReply").mockResolvedValue({
-      ok: true,
-    });
-    composeSpy = spyOn(
-      approvalMessageComposer,
-      "composeApprovalMessageGenerative",
-    ).mockResolvedValue("test message");
-  });
-
-  // ── 1. approve_once via callback mints a grant ──
-
-  test("approve_once via callback for tool-approval request mints a scoped grant", async () => {
-    const requestId = "req-grant-cb-1";
-    createTestGuardianApproval(requestId);
-    registerPendingInteraction(
-      requestId,
-      CONVERSATION_ID,
-      TOOL_NAME,
-      TOOL_INPUT,
-    );
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-1",
-      callbackData: `apr:${requestId}:approve_once`,
-      content: "",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: GUARDIAN_USER,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("guardian_decision_applied");
-
-    // Verify a grant was minted
-    expect(countGrants()).toBe(1);
-
-    const grant = getLatestGrant();
-    expect(grant).not.toBeNull();
-    expect(grant!.scope_mode).toBe("tool_signature");
-    expect(grant!.tool_name).toBe(TOOL_NAME);
-    expect(grant!.status).toBe("active");
-    expect(grant!.request_channel).toBe("telegram");
-    expect(grant!.decision_channel).toBe("telegram");
-    expect(grant!.guardian_external_user_id).toBe(GUARDIAN_USER);
-    expect(grant!.requester_external_user_id).toBe(REQUESTER_USER);
-    expect(grant!.conversation_id).toBe(CONVERSATION_ID);
-    expect(grant!.execution_channel).toBeNull();
-    expect(grant!.call_session_id).toBeNull();
-
-    // Verify the input digest matches what computeToolApprovalDigest produces
-    const expectedDigest = computeToolApprovalDigest(TOOL_NAME, TOOL_INPUT);
-    expect(grant!.input_digest).toBe(expectedDigest);
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  // Reaction-based approvals are exercised against the canonical pipeline in
-  // slack-reaction-canonical-approval.test.ts — reactions route through
-  // routeGuardianReply, not handleApprovalInterception.
-
-  // ── 2. approve_once for non-tool-approval does NOT mint a grant ──
-
-  test("approve_once for informational request (no toolName) does NOT mint a grant", async () => {
-    const requestId = "req-no-grant-1";
-    // Informational requests have no meaningful tool name — the empty string
-    // signals that this is not a tool-approval request.
-    createTestGuardianApproval(requestId, { toolName: "" });
-    registerPendingInteraction(requestId, CONVERSATION_ID, "", {});
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-2",
-      callbackData: `apr:${requestId}:approve_once`,
-      content: "",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: GUARDIAN_USER,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("guardian_decision_applied");
-
-    // No grant should have been minted
-    expect(countGrants()).toBe(0);
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  // ── 2b. approve_once for zero-argument tool call DOES mint a grant ──
-
-  test("approve_once for zero-argument tool call mints a scoped grant", async () => {
-    const requestId = "req-grant-zero-arg";
-    const zeroArgTool = "get_system_status";
-    createTestGuardianApproval(requestId, { toolName: zeroArgTool });
-    // Register with empty input object to simulate a zero-argument tool call
-    registerPendingInteraction(requestId, CONVERSATION_ID, zeroArgTool, {});
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-2b",
-      callbackData: `apr:${requestId}:approve_once`,
-      content: "",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: GUARDIAN_USER,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("guardian_decision_applied");
-
-    // A grant MUST be minted even though input is {}
-    expect(countGrants()).toBe(1);
-
-    const grant = getLatestGrant();
-    expect(grant).not.toBeNull();
-    expect(grant!.scope_mode).toBe("tool_signature");
-    expect(grant!.tool_name).toBe(zeroArgTool);
-    expect(grant!.status).toBe("active");
-
-    // Verify the input digest matches what computeToolApprovalDigest produces for empty input
-    const expectedDigest = computeToolApprovalDigest(zeroArgTool, {});
-    expect(grant!.input_digest).toBe(expectedDigest);
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  // ── 3. reject does NOT mint a grant ──
-
-  test("reject decision does NOT mint a scoped grant", async () => {
-    const requestId = "req-no-grant-rej";
-    createTestGuardianApproval(requestId);
-    registerPendingInteraction(
-      requestId,
-      CONVERSATION_ID,
-      TOOL_NAME,
-      TOOL_INPUT,
-    );
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-3",
-      callbackData: `apr:${requestId}:reject`,
-      content: "",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: GUARDIAN_USER,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("guardian_decision_applied");
-
-    // No grant should have been minted
-    expect(countGrants()).toBe(0);
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  // ── 4. Identity mismatch remains fail-closed (no grant minted) ──
-
-  test("identity mismatch does NOT mint a grant and fails closed", async () => {
-    const requestId = "req-mismatch-1";
-    createTestGuardianApproval(requestId);
-    registerPendingInteraction(
-      requestId,
-      CONVERSATION_ID,
-      TOOL_NAME,
-      TOOL_INPUT,
-    );
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-4",
-      callbackData: `apr:${requestId}:approve_once`,
-      content: "",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: "wrong-guardian-user",
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.handled).toBe(true);
-    // Identity mismatch results in guardian_decision_applied (fail-closed, no actual decision applied)
-    expect(result.type).toBe("guardian_decision_applied");
-
-    // No grant should have been minted
-    expect(countGrants()).toBe(0);
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  // ── 5. Stale/already-resolved request does NOT mint a grant ──
-
-  test("stale request (already resolved) does NOT mint a grant", async () => {
-    const requestId = "req-stale-1";
-    // Create guardian approval but do NOT register a pending interaction
-    // This simulates the pending interaction being already resolved
-    createTestGuardianApproval(requestId);
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-5",
-      callbackData: `apr:${requestId}:approve_once`,
-      content: "",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: GUARDIAN_USER,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("stale_ignored");
-
-    // No grant should have been minted
-    expect(countGrants()).toBe(0);
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  // ── 6. approve_once via conversation engine mints a grant ──
-
-  test("approve_once via conversation engine mints a scoped grant", async () => {
-    const requestId = "req-grant-eng-1";
-    createTestGuardianApproval(requestId);
-    registerPendingInteraction(
-      requestId,
-      CONVERSATION_ID,
-      TOOL_NAME,
-      TOOL_INPUT,
-    );
-
-    const mockGenerator = async () => ({
-      disposition: "approve_once" as const,
-      replyText: "Approved!",
-      targetRequestId: requestId,
-    });
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-6",
-      content: "yes, approve it",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: GUARDIAN_USER,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-      approvalConversationGenerator: mockGenerator,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("guardian_decision_applied");
-
-    // Verify a grant was minted
-    expect(countGrants()).toBe(1);
-
-    const grant = getLatestGrant();
-    expect(grant).not.toBeNull();
-    expect(grant!.scope_mode).toBe("tool_signature");
-    expect(grant!.tool_name).toBe(TOOL_NAME);
-    expect(grant!.status).toBe("active");
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  // ── 7. reject via conversation engine does NOT mint a grant ──
-
-  test("reject via conversation engine does NOT mint a grant", async () => {
-    const requestId = "req-no-grant-eng-rej";
-    createTestGuardianApproval(requestId);
-    registerPendingInteraction(
-      requestId,
-      CONVERSATION_ID,
-      TOOL_NAME,
-      TOOL_INPUT,
-    );
-
-    const mockGenerator = async () => ({
-      disposition: "reject" as const,
-      replyText: "Denied.",
-      targetRequestId: requestId,
-    });
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-7",
-      content: "no, deny it",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: GUARDIAN_USER,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-      approvalConversationGenerator: mockGenerator,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("guardian_decision_applied");
-
-    // No grant should have been minted
-    expect(countGrants()).toBe(0);
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  // ── 8. Grant TTL is approximately 5 minutes ──
-
-  test("minted grant has approximately 5-minute TTL", async () => {
-    const requestId = "req-grant-ttl-1";
-    createTestGuardianApproval(requestId);
-    registerPendingInteraction(
-      requestId,
-      CONVERSATION_ID,
-      TOOL_NAME,
-      TOOL_INPUT,
-    );
-
-    const beforeTime = Date.now();
-
-    const result = await handleApprovalInterception({
-      conversationId: "guardian-conv-9",
-      callbackData: `apr:${requestId}:approve_once`,
-      content: "",
-      conversationExternalId: GUARDIAN_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: GUARDIAN_USER,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: makeTrustContext(),
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.type).toBe("guardian_decision_applied");
-
-    const grant = getLatestGrant();
-    expect(grant).not.toBeNull();
-
-    const expiresAt = grant!.expires_at as number;
-    const expectedMin = beforeTime + GRANT_TTL_MS - 1000; // 1s tolerance
-    const expectedMax = beforeTime + GRANT_TTL_MS + 5000; // 5s tolerance
-    expect(expiresAt).toBeGreaterThanOrEqual(expectedMin);
-    expect(expiresAt).toBeLessThanOrEqual(expectedMax);
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-});
-
-describe("approval interception trust-class regression coverage", () => {
-  let deliverSpy: ReturnType<typeof spyOn>;
-  let composeSpy: ReturnType<typeof spyOn>;
-
-  beforeEach(() => {
-    resetTables();
-    deliverSpy = spyOn(gatewayClient, "deliverChannelReply").mockResolvedValue({
-      ok: true,
-    });
-    composeSpy = spyOn(
-      approvalMessageComposer,
-      "composeApprovalMessageGenerative",
-    ).mockResolvedValue("test message");
-  });
-
-  test("identity-known unknown sender does not auto-deny pending approval", async () => {
-    const requestId = "req-unknown-no-auto-deny-1";
-    const sessionMock = registerPendingInteraction(
-      requestId,
-      CONVERSATION_ID,
-      TOOL_NAME,
-      TOOL_INPUT,
-    );
-    createTestGuardianApproval(requestId);
-
-    const result = await handleApprovalInterception({
-      conversationId: CONVERSATION_ID,
-      content: "approve",
-      conversationExternalId: REQUESTER_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: "intruder-user-1",
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: {
-        sourceChannel: "telegram",
-        trustClass: "unknown",
-        requesterExternalUserId: "intruder-user-1",
-        guardianExternalUserId: "guardian-1",
-      },
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("assistant_turn");
-    expect(sessionMock).not.toHaveBeenCalled();
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-
-  test("legacy unverified sender still auto-denies pending approval", async () => {
-    const requestId = "req-unknown-auto-deny-1";
-    const sessionMock = registerPendingInteraction(
-      requestId,
-      CONVERSATION_ID,
-      TOOL_NAME,
-      TOOL_INPUT,
-    );
-    createTestGuardianApproval(requestId);
-
-    const result = await handleApprovalInterception({
-      conversationId: CONVERSATION_ID,
-      content: "approve",
-      conversationExternalId: REQUESTER_CHAT,
-      sourceChannel: "telegram",
-      actorExternalId: undefined,
-      replyCallbackUrl: "https://gateway.test/deliver",
-      trustCtx: {
-        sourceChannel: "telegram",
-        trustClass: "unknown",
-      },
-      assistantId: ASSISTANT_ID,
-    });
-
-    expect(result.handled).toBe(true);
-    expect(result.type).toBe("decision_applied");
-    expect(sessionMock).toHaveBeenCalled();
-    expect(sessionMock.mock.calls[0]?.[0]).toBe(requestId);
-    expect(sessionMock.mock.calls[0]?.[1]).toBe("deny");
-
-    deliverSpy.mockRestore();
-    composeSpy.mockRestore();
-  });
-});