@vellumai/assistant 0.8.1 → 0.8.2

package/src/__tests__/title-generate-pipeline.test.ts

@@ -80,11 +80,6 @@ describe("titleGenerate pipeline", () => {
       manifest: {
         name: "default-title-generate",
         version: "1.0.0",
-        provides: { titleGenerate: "v1" },
-        requires: {
-          pluginRuntime: "v1",
-          titleGenerateApi: "v1",
-        },
       },
     });
   });
@@ -149,10 +144,6 @@ describe("titleGenerate pipeline", () => {
       manifest: {
         name: "custom-deterministic-title",
         version: "0.0.1",
-        requires: {
-          pluginRuntime: "v1",
-          titleGenerateApi: "v1",
-        },
       },
       middleware: { titleGenerate: deterministicMw },
     };
@@ -201,10 +192,6 @@ describe("titleGenerate pipeline", () => {
       manifest: {
         name: "observer",
         version: "0.0.1",
-        requires: {
-          pluginRuntime: "v1",
-          titleGenerateApi: "v1",
-        },
       },
       middleware: { titleGenerate: passthroughMw },
     });

package/src/__tests__/token-estimate-pipeline.test.ts

@@ -338,7 +338,6 @@ describe("tokenEstimate pipeline — custom override", () => {
       manifest: {
         name: "custom-token-estimate",
         version: "1.0.0",
-        requires: { pluginRuntime: "v1", tokenEstimateApi: "v1" },
       },
       middleware: { tokenEstimate: override },
     };
@@ -376,7 +375,6 @@ describe("tokenEstimate pipeline — custom override", () => {
       manifest: {
         name: "doubling-token-estimate",
         version: "1.0.0",
-        requires: { pluginRuntime: "v1", tokenEstimateApi: "v1" },
       },
       middleware: { tokenEstimate: doubler },
     };
@@ -420,7 +418,6 @@ describe("tokenEstimate pipeline — default does not shadow late plugins", () =
       manifest: {
         name: "late-registered-observer",
         version: "1.0.0",
-        requires: { pluginRuntime: "v1", tokenEstimateApi: "v1" },
       },
       middleware: { tokenEstimate: observer },
     };

package/src/__tests__/tool-error-pipeline.test.ts

@@ -150,7 +150,6 @@ describe("toolError pipeline", () => {
         manifest: {
           name: "custom-tool-error",
           version: "0.0.1",
-          requires: { pluginRuntime: "v1", toolErrorApi: "v1" },
         },
         middleware: { toolError: customMiddleware },
       };
@@ -176,7 +175,6 @@ describe("toolError pipeline", () => {
         manifest: {
           name: "no-nudge",
           version: "0.0.1",
-          requires: { pluginRuntime: "v1", toolErrorApi: "v1" },
         },
         middleware: { toolError: suppressingMiddleware },
       };
@@ -227,7 +225,6 @@ describe("toolError pipeline", () => {
         manifest: {
           name: "late-user-plugin",
           version: "0.0.1",
-          requires: { pluginRuntime: "v1", toolErrorApi: "v1" },
         },
         middleware: { toolError: userMiddleware },
       });

package/src/__tests__/tool-execute-pipeline.test.ts

@@ -249,7 +249,6 @@ describe("ToolExecutor.execute → toolExecute pipeline", () => {
       manifest: {
         name: "spy-tool-execute",
         version: "0.0.1",
-        requires: { pluginRuntime: "v1" },
       },
       middleware: { toolExecute: spyMiddleware },
     };
@@ -308,7 +307,6 @@ describe("ToolExecutor.execute → toolExecute pipeline", () => {
       manifest: {
         name: "short-circuit-tool-execute",
         version: "0.0.1",
-        requires: { pluginRuntime: "v1" },
       },
       middleware: { toolExecute: shortCircuit },
     };
@@ -342,7 +340,6 @@ describe("ToolExecutor.execute → toolExecute pipeline", () => {
       manifest: {
         name: "slow-tool-execute",
         version: "0.0.1",
-        requires: { pluginRuntime: "v1" },
       },
       middleware: { toolExecute: slow },
     });
@@ -388,7 +385,6 @@ describe("ToolExecutor.execute → toolExecute pipeline", () => {
       manifest: {
         name: "outer-tool-execute",
         version: "0.0.1",
-        requires: { pluginRuntime: "v1" },
       },
       middleware: { toolExecute: outer },
     });
@@ -396,7 +392,6 @@ describe("ToolExecutor.execute → toolExecute pipeline", () => {
       manifest: {
         name: "inner-tool-execute",
         version: "0.0.1",
-        requires: { pluginRuntime: "v1" },
       },
       middleware: { toolExecute: inner },
     });

package/src/__tests__/tool-executor-lifecycle-events.test.ts

@@ -242,7 +242,7 @@ describe("ToolExecutor lifecycle events", () => {
     );
 
     expect(result.isError).toBe(true);
-    expect(result.content).toContain("Permission denied by user");
+    expect(result.content).toContain("Permission denied");
     expect(events.map((event) => event.type)).toEqual([
       "start",
       "permission_prompt",

package/src/__tests__/tool-executor.test.ts

@@ -462,6 +462,7 @@ describe("isSideEffectTool", () => {
       "web_fetch",
       "document_create",
       "document_update",
+      "document_delete",
       "schedule_create",
       "schedule_update",
       "schedule_delete",
@@ -670,6 +671,7 @@ describe("ToolExecutor forcePromptSideEffects enforcement", () => {
       { name: "web_fetch", input: { url: "https://example.com" } },
       { name: "document_create", input: { title: "doc", content: "body" } },
       { name: "document_update", input: { id: "doc-1", content: "updated" } },
+      { name: "document_delete", input: { surface_id: "doc-1" } },
       {
         name: "credential_store",
         input: { action: "store", name: "api-key", value: "secret" },
@@ -842,8 +844,12 @@ describe("ToolExecutor forcePromptSideEffects enforcement", () => {
 // ---------------------------------------------------------------------------
 
 // Import the real buildSanitizedEnv (not mocked) for baseline credential tests
-const { buildSanitizedEnv, SAFE_ENV_VARS, ALWAYS_INJECTED_ENV_VARS } =
-  await import("../tools/terminal/safe-env.js");
+const {
+  buildSanitizedEnv,
+  KATA_SAFE_ENV_VARS,
+  SAFE_ENV_VARS,
+  ALWAYS_INJECTED_ENV_VARS,
+} = await import("../tools/terminal/safe-env.js");
 
 describe("buildSanitizedEnv — baseline: credential exclusion", () => {
   // Credential-like env vars that must never appear in the sanitized env.
@@ -901,7 +907,11 @@ describe("buildSanitizedEnv — baseline: credential exclusion", () => {
   });
 
   test("sanitized env only contains keys from the allowlist", () => {
-    const allowed: string[] = [...SAFE_ENV_VARS, ...ALWAYS_INJECTED_ENV_VARS];
+    const allowed: string[] = [
+      ...SAFE_ENV_VARS,
+      ...KATA_SAFE_ENV_VARS,
+      ...ALWAYS_INJECTED_ENV_VARS,
+    ];
     const env = buildSanitizedEnv();
     for (const key of Object.keys(env)) {
       expect(allowed).toContain(key);
@@ -1211,7 +1221,9 @@ describe("ToolExecutionResult includes risk metadata from classifier assessment"
     cachedAssessmentOverride = {
       riskLevel: "low",
       reason: "GET request to public URL",
-      scopeOptions: [{ pattern: "https://example.com/.*", label: "example.com" }],
+      scopeOptions: [
+        { pattern: "https://example.com/.*", label: "example.com" },
+      ],
       // allowlistOptions intentionally omitted — some classifiers don't emit them.
       matchType: "registry",
     };

package/src/__tests__/tool-result-truncate-pipeline.test.ts

@@ -236,10 +236,6 @@ describe("toolResultTruncate pipeline", () => {
         manifest: {
           name: "short-circuit",
           version: "1.0.0",
-          requires: {
-            pluginRuntime: "v1",
-            toolResultTruncateApi: "v1",
-          },
         },
         middleware: { toolResultTruncate: shortCircuit },
       });
@@ -274,10 +270,6 @@ describe("toolResultTruncate pipeline", () => {
         manifest: {
           name: "prefixer",
           version: "1.0.0",
-          requires: {
-            pluginRuntime: "v1",
-            toolResultTruncateApi: "v1",
-          },
         },
         middleware: { toolResultTruncate: prefixer },
       });
@@ -323,10 +315,6 @@ describe("toolResultTruncate pipeline", () => {
         manifest: {
           name: "late-user-plugin",
           version: "0.0.1",
-          requires: {
-            pluginRuntime: "v1",
-            toolResultTruncateApi: "v1",
-          },
         },
         middleware: { toolResultTruncate: userMiddleware },
       });

package/src/__tests__/turn-events-store.test.ts

@@ -0,0 +1,256 @@
+/**
+ * Tests for queryUnreportedTurnEvents.
+ *
+ * Verifies that the JOIN to `conversations` correctly attaches the parent
+ * conversation's `conversationType` to each returned turn, so analytics
+ * downstream (DAU) can filter out background/scheduled prompts.
+ */
+
+import { beforeEach, describe, expect, mock, test } from "bun:test";
+
+import { makeMockLogger } from "./helpers/mock-logger.js";
+
+mock.module("../util/logger.js", () => ({
+  getLogger: () => makeMockLogger(),
+}));
+
+import { addMessage, createConversation } from "../memory/conversation-crud.js";
+import { getDb } from "../memory/db-connection.js";
+import { initializeDb } from "../memory/db-init.js";
+import { messages } from "../memory/schema.js";
+import { queryUnreportedTurnEvents } from "../memory/turn-events-store.js";
+
+initializeDb();
+
+function purge(): void {
+  const db = getDb();
+  // Wipe messages then conversations to keep tests independent. FK cascade
+  // also removes any dependent rows.
+  db.run("DELETE FROM messages");
+  db.run("DELETE FROM conversations");
+}
+
+beforeEach(() => {
+  purge();
+});
+
+/**
+ * Insert a user message at a specific epoch-millis timestamp so we can
+ * assert ordering and the watermark cursor.
+ */
+async function insertUserMessageAt(
+  conversationId: string,
+  content: string,
+  createdAt: number,
+): Promise<string> {
+  const message = await addMessage(conversationId, "user", content);
+  // addMessage stamps createdAt to monotonic now(); rewrite for
+  // deterministic ordering in the test.
+  const db = getDb();
+  db.run(
+    `UPDATE messages SET created_at = ${createdAt} WHERE id = '${message.id}'`,
+  );
+  return message.id;
+}
+
+describe("queryUnreportedTurnEvents", () => {
+  test("attaches conversationId, conversationType, and turnIndex for a standard conversation", async () => {
+    const conv = createConversation({ conversationType: "standard" });
+    await insertUserMessageAt(conv.id, "hello", 1000);
+
+    const events = queryUnreportedTurnEvents(0, undefined, 100);
+    expect(events.length).toBe(1);
+    expect(events[0].conversationId).toBe(conv.id);
+    expect(events[0].conversationType).toBe("standard");
+    // The single user message in this conversation is turn 1.
+    expect(events[0].turnIndex).toBe(1);
+  });
+
+  test("attaches conversationType for background and scheduled conversations", async () => {
+    const bg = createConversation({ conversationType: "background" });
+    const sched = createConversation({ conversationType: "scheduled" });
+    const std = createConversation({ conversationType: "standard" });
+
+    const bgMsg = await insertUserMessageAt(bg.id, "bg prompt", 1000);
+    const schedMsg = await insertUserMessageAt(sched.id, "sched prompt", 2000);
+    const stdMsg = await insertUserMessageAt(std.id, "user typed this", 3000);
+
+    const events = queryUnreportedTurnEvents(0, undefined, 100);
+    const byId = Object.fromEntries(events.map((e) => [e.id, e]));
+
+    expect(byId[bgMsg].conversationType).toBe("background");
+    expect(byId[schedMsg].conversationType).toBe("scheduled");
+    expect(byId[stdMsg].conversationType).toBe("standard");
+  });
+
+  test("excludes tool_result and web_search_tool_result content patterns", async () => {
+    const conv = createConversation({ conversationType: "standard" });
+
+    // Real user turn
+    const realId = await insertUserMessageAt(conv.id, "real user text", 1000);
+
+    // Synthetic "user" rows that wrap tool results: these should be filtered
+    const db = getDb();
+    db.insert(messages)
+      .values({
+        id: "tool-result-id",
+        conversationId: conv.id,
+        role: "user",
+        content: JSON.stringify([
+          { type: "tool_result", tool_use_id: "x", content: "" },
+        ]),
+        createdAt: 2000,
+      })
+      .run();
+    db.insert(messages)
+      .values({
+        id: "web-search-tool-result-id",
+        conversationId: conv.id,
+        role: "user",
+        content: JSON.stringify([
+          { type: "web_search_tool_result", tool_use_id: "y", content: "" },
+        ]),
+        createdAt: 3000,
+      })
+      .run();
+
+    const events = queryUnreportedTurnEvents(0, undefined, 100);
+    expect(events.length).toBe(1);
+    expect(events[0].id).toBe(realId);
+    expect(events[0].conversationType).toBe("standard");
+    // Tool-result rows must NOT increment the turn index: this is the
+    // only real user turn in the conversation, so its index is 1 even
+    // though two synthetic role="user" rows exist with later timestamps.
+    expect(events[0].turnIndex).toBe(1);
+  });
+
+  test("turnIndex counts real user turns in (createdAt, id) order within a conversation", async () => {
+    const conv = createConversation({ conversationType: "standard" });
+    const m1 = await insertUserMessageAt(conv.id, "first", 1000);
+    const m2 = await insertUserMessageAt(conv.id, "second", 2000);
+    const m3 = await insertUserMessageAt(conv.id, "third", 3000);
+
+    const events = queryUnreportedTurnEvents(0, undefined, 100);
+    const byId = Object.fromEntries(events.map((e) => [e.id, e]));
+    expect(byId[m1].turnIndex).toBe(1);
+    expect(byId[m2].turnIndex).toBe(2);
+    expect(byId[m3].turnIndex).toBe(3);
+  });
+
+  test("turnIndex resets across conversations", async () => {
+    const a = createConversation({ conversationType: "standard" });
+    const b = createConversation({ conversationType: "standard" });
+    const a1 = await insertUserMessageAt(a.id, "a first", 1000);
+    const b1 = await insertUserMessageAt(b.id, "b first", 1500);
+    const a2 = await insertUserMessageAt(a.id, "a second", 2000);
+
+    const events = queryUnreportedTurnEvents(0, undefined, 100);
+    const byId = Object.fromEntries(events.map((e) => [e.id, e]));
+    // Conversation A: 1, 2. Conversation B: 1 (independent of A).
+    expect(byId[a1].turnIndex).toBe(1);
+    expect(byId[a2].turnIndex).toBe(2);
+    expect(byId[b1].turnIndex).toBe(1);
+    // And the conversation_id labels them correctly.
+    expect(byId[a1].conversationId).toBe(a.id);
+    expect(byId[b1].conversationId).toBe(b.id);
+  });
+
+  test("turnIndex skips tool_result rows even when they're interleaved", async () => {
+    const conv = createConversation({ conversationType: "standard" });
+    const real1 = await insertUserMessageAt(conv.id, "first real", 1000);
+
+    // Inject a tool_result row between the two real turns. The
+    // correlated subquery must use the same content filter as the
+    // outer query, otherwise turn_index would jump by 2 instead of 1.
+    const db = getDb();
+    db.insert(messages)
+      .values({
+        id: "interleaved-tool-result",
+        conversationId: conv.id,
+        role: "user",
+        content: JSON.stringify([
+          { type: "tool_result", tool_use_id: "x", content: "" },
+        ]),
+        createdAt: 1500,
+      })
+      .run();
+
+    const real2 = await insertUserMessageAt(conv.id, "second real", 2000);
+
+    const events = queryUnreportedTurnEvents(0, undefined, 100);
+    const byId = Object.fromEntries(events.map((e) => [e.id, e]));
+    expect(events.length).toBe(2);
+    expect(byId[real1].turnIndex).toBe(1);
+    expect(byId[real2].turnIndex).toBe(2);
+  });
+
+  test("respects the cursor and returns events in (createdAt, id) order", async () => {
+    const conv = createConversation({ conversationType: "standard" });
+    const m1 = await insertUserMessageAt(conv.id, "first", 1000);
+    const m2 = await insertUserMessageAt(conv.id, "second", 2000);
+    const m3 = await insertUserMessageAt(conv.id, "third", 3000);
+
+    // Cursor at (1500, undefined) should skip m1 and return m2, m3 in order.
+    const after = queryUnreportedTurnEvents(1500, undefined, 100);
+    expect(after.map((e) => e.id)).toEqual([m2, m3]);
+
+    // Sanity: pulling from the start returns all three with correct types.
+    const all = queryUnreportedTurnEvents(0, undefined, 100);
+    expect(all.map((e) => e.id)).toEqual([m1, m2, m3]);
+    for (const e of all) {
+      expect(e.conversationType).toBe("standard");
+    }
+  });
+
+  test("extracts interfaceId, channelId, and clientMetadata from messages.metadata", async () => {
+    // Three user messages with different metadata shapes to cover the
+    // realistic spectrum of attribution carried on messages.metadata:
+    //  - Full interactive: macOS surface, vellum channel, with a client
+    //    block stamped by the (forthcoming) HTTP header middleware.
+    //  - Channel inbound: Slack surface, no client headers (channel
+    //    inbound paths don't carry browser context).
+    //  - Legacy: a message persisted before metadata threading existed.
+    const conv = createConversation({ conversationType: "standard" });
+
+    const macOsMsg = await addMessage(conv.id, "user", "hi from desktop", {
+      userMessageInterface: "macos",
+      userMessageChannel: "vellum",
+      client: {
+        os: "darwin",
+        interface_version: "0.8.2",
+      },
+    });
+    const slackMsg = await addMessage(conv.id, "user", "hi from slack", {
+      userMessageInterface: "slack",
+      userMessageChannel: "slack",
+    });
+    const legacyMsg = await addMessage(conv.id, "user", "hi from the past");
+
+    const events = queryUnreportedTurnEvents(0, undefined, 100);
+    const byId = Object.fromEntries(events.map((e) => [e.id, e]));
+
+    expect(byId[macOsMsg.id]).toBeDefined();
+    expect(byId[macOsMsg.id].interfaceId).toBe("macos");
+    expect(byId[macOsMsg.id].channelId).toBe("vellum");
+    // clientMetadata is returned as raw JSON text -- the reporter parses it.
+    // We verify it round-trips back to the input object.
+    expect(byId[macOsMsg.id].clientMetadata).not.toBeNull();
+    expect(
+      JSON.parse(byId[macOsMsg.id].clientMetadata as string),
+    ).toMatchObject({
+      os: "darwin",
+      interface_version: "0.8.2",
+    });
+
+    expect(byId[slackMsg.id].interfaceId).toBe("slack");
+    expect(byId[slackMsg.id].channelId).toBe("slack");
+    expect(byId[slackMsg.id].clientMetadata).toBeNull();
+
+    // Legacy rows (no metadata threading) return null for all three
+    // attribution fields -- downstream analytics treats these as
+    // "unknown" without breaking the batch.
+    expect(byId[legacyMsg.id].interfaceId).toBeNull();
+    expect(byId[legacyMsg.id].channelId).toBeNull();
+    expect(byId[legacyMsg.id].clientMetadata).toBeNull();
+  });
+});

package/src/__tests__/twilio-routes.test.ts

@@ -1392,6 +1392,10 @@ describe("twilio webhook routes", () => {
 
       expect(result.success).toBe(true);
       expect(result.hasCredentials).toBe(true);
+      expect(mockRawConfigStore.twilio).toEqual({
+        accountSid: "AC_new_credentials",
+        setupStarted: true,
+      });
     });
 
     test("clearing credentials succeeds", async () => {

package/src/__tests__/user-plugin-loader.test.ts

@@ -81,7 +81,6 @@ registerPlugin({
   manifest: {
     name: "my-plugin",
     version: "0.0.1",
-    requires: { pluginRuntime: "v1" },
   },
 });
 `,
@@ -111,7 +110,6 @@ registerPlugin({
   manifest: {
     name: "good-plugin",
     version: "0.0.1",
-    requires: { pluginRuntime: "v1" },
   },
 });
 `,
@@ -151,7 +149,6 @@ registerPlugin({
   manifest: {
     name: "hanging-plugin",
     version: "0.0.1",
-    requires: { pluginRuntime: "v1" },
   },
 });
 `,
@@ -163,7 +160,6 @@ registerPlugin({
   manifest: {
     name: "healthy-plugin",
     version: "0.0.1",
-    requires: { pluginRuntime: "v1" },
   },
 });
 `,
@@ -198,7 +194,6 @@ registerPlugin({
   manifest: {
     name: "slow-late-plugin",
     version: "0.0.1",
-    requires: { pluginRuntime: "v1" },
   },
 });
 `,
@@ -305,7 +300,6 @@ registerPlugin({
   manifest: {
     name: "healthy-legacy",
     version: "0.0.1",
-    requires: { pluginRuntime: "v1" },
   },
 });
 `,
@@ -330,7 +324,6 @@ registerPlugin({
   manifest: {
     name: "old-style",
     version: "0.0.1",
-    requires: { pluginRuntime: "v1" },
   },
 });
 `,