@vellumai/assistant 0.8.4 → 0.8.6

package/src/config/assistant-feature-flags.ts

@@ -7,7 +7,7 @@
  * (in priority order):
  *   1. Override values from the gateway IPC socket
  *   2. defaults registry `defaultEnabled`         (for declared keys)
- *   3. `true`                                     (for undeclared keys)
+ *   3. `false`                                    (for undeclared keys)
  *
  * Key format:
  *   Canonical:  simple kebab-case string (e.g., "browser", "ces-tools")
@@ -18,6 +18,12 @@ import { dirname, join } from "node:path";
 
 import { ipcGetFeatureFlags } from "../ipc/gateway-client.js";
 import { getLogger } from "../util/logger.js";
+import {
+  clearCachedOverrides,
+  getCachedOverrides,
+  isCachedFromGateway,
+  setCachedOverrides,
+} from "./feature-flag-cache.js";
 import type { AssistantConfig } from "./schema.js";
 
 const log = getLogger("assistant-feature-flags");
@@ -112,22 +118,10 @@ function parseRegistryToDefaults(parsed: unknown): FeatureFlagDefaultsRegistry {
 // ---------------------------------------------------------------------------
 // Override loading — reads from gateway IPC socket
 // ---------------------------------------------------------------------------
-
-/**
- * Module-level cache of feature flag override values. Populated by
- * `initFeatureFlagOverrides()` at startup, invalidated by
- * `clearFeatureFlagOverridesCache()`.
- */
-let cachedOverrides: Record<string, boolean> | null = null;
-
-/**
- * True when `cachedOverrides` was populated by the gateway IPC fetch or
- * preseeded by a test via `_setOverridesForTesting()`. Guards
- * `initFeatureFlagOverrides()` from clobbering an existing populated cache
- * when called a second time (e.g. by a CLI entry point after the daemon
- * has already initialized).
- */
-let cachedOverridesFromGateway = false;
+//
+// The override cache lives in `feature-flag-cache.ts` (stdlib-only) so test
+// helpers can seed it without dragging the pino logger + gateway IPC client
+// transitively through their import chain. See that file's block comment.
 
 /**
  * Fetch override values from the gateway via IPC (Unix domain socket).
@@ -172,8 +166,8 @@ const DEFAULT_INIT_RETRY_BACKOFFS_MS: readonly number[] = [
  * Retries the gateway IPC fetch on empty/failed results — the gateway
  * may not have bound its IPC socket yet when the daemon races ahead at
  * startup. After exhausting retries, the cache is left unset so
- * subsequent sync calls return an empty override map (registry defaults
- * only).
+ * subsequent sync calls return an empty override map (registry defaults for
+ * declared flags, fail-closed for undeclared flags).
  *
  * Pass `retryBackoffsMs: []` to disable retries (used by unit tests that
  * intentionally simulate an unreachable gateway and want immediate
@@ -181,8 +175,9 @@ const DEFAULT_INIT_RETRY_BACKOFFS_MS: readonly number[] = [
  *
  * No-ops when the cache is already populated — callers that want to
  * refresh must call `clearFeatureFlagOverridesCache()` first. This lets
- * tests preseed flag state via `_setOverridesForTesting()` without the
- * gateway IPC call clobbering their setup.
+ * tests preseed flag state via `setOverridesForTesting()` (in
+ * `__tests__/feature-flag-test-helpers.ts`) without the gateway IPC call
+ * clobbering their setup.
  */
 export async function initFeatureFlagOverrides(options?: {
   retryBackoffsMs?: readonly number[];
@@ -194,7 +189,7 @@ export async function initFeatureFlagOverrides(options?: {
    */
   callTimeoutMs?: number;
 }): Promise<void> {
-  if (cachedOverridesFromGateway) return;
+  if (isCachedFromGateway()) return;
 
   const backoffs = options?.retryBackoffsMs ?? DEFAULT_INIT_RETRY_BACKOFFS_MS;
   const callTimeoutMs = options?.callTimeoutMs;
@@ -208,15 +203,14 @@ export async function initFeatureFlagOverrides(options?: {
       const delay = backoffs[attempt - 1]!;
       await new Promise((resolve) => setTimeout(resolve, delay));
       // Re-check after the wait: a concurrent caller (e.g. a test using
-      // `_setOverridesForTesting`) may have populated the cache while we
+      // `setOverridesForTesting`) may have populated the cache while we
       // were sleeping. Bail out so we don't clobber their setup.
-      if (cachedOverridesFromGateway) return;
+      if (isCachedFromGateway()) return;
     }
 
     const gatewayOverrides = await fetchOverridesFromGateway(callTimeoutMs);
     if (Object.keys(gatewayOverrides).length > 0) {
-      cachedOverrides = gatewayOverrides;
-      cachedOverridesFromGateway = true;
+      setCachedOverrides(gatewayOverrides, { fromGateway: true });
       if (attempt > 0) {
         log.info(
           { attempt: attempt + 1 },
@@ -229,12 +223,11 @@ export async function initFeatureFlagOverrides(options?: {
 
   // Exhausted retries — leave cache unset so loadOverrides() returns an
   // empty map on subsequent sync reads. Flag checks fall through to the
-  // registry default (`defaultEnabled`), which biases toward off for
-  // newer assistant-scope flags.
+  // registry default (`defaultEnabled`) or the fail-closed undeclared default.
   if (backoffs.length > 0) {
     log.warn(
       { attempts: backoffs.length + 1 },
-      "Feature flag overrides empty after all retries; falling back to registry defaults",
+      "Feature flag overrides empty after all retries; falling back to registry defaults and fail-closed undeclared flags",
     );
   }
 }
@@ -246,33 +239,33 @@ export async function initFeatureFlagOverrides(options?: {
  * called at startup, or an empty record otherwise.
  */
 function loadOverrides(): Record<string, boolean> {
-  return cachedOverrides ?? {};
+  return getCachedOverrides() ?? {};
 }
 
 /**
  * Invalidate the cached overrides so the next call to
  * `isAssistantFeatureFlagEnabled` re-reads from the gateway.
  *
- * Used by tests between cases to reset module state.
+ * Called by `refreshOverridesFromGateway()` when the gateway pushes a
+ * `feature_flags_changed` event, and by tests between cases to reset
+ * module state. (Tests typically call `setOverridesForTesting()` from
+ * `__tests__/feature-flag-test-helpers.ts`, which combines clear + seed.)
  */
 export function clearFeatureFlagOverridesCache(): void {
-  cachedOverrides = null;
-  cachedOverridesFromGateway = false;
+  clearCachedOverrides();
 }
 
 /**
- * Directly inject override values into the module-level cache.
+ * Re-fetch feature flag overrides from the gateway.
  *
- * **Test-only** — bypasses the gateway IPC fetch so unit tests can control
- * flag state without standing up a real gateway. Production code should
- * never call this; use `clearFeatureFlagOverridesCache()` instead and let
- * the resolver re-read from the gateway.
+ * Clears the cached overrides and re-runs the gateway IPC fetch without
+ * retries (the gateway is known to be up because it just pushed an event).
+ * Called by the gateway flag listener when a `feature_flags_changed` event
+ * arrives.
  */
-export function _setOverridesForTesting(
-  overrides: Record<string, boolean>,
-): void {
-  cachedOverrides = { ...overrides };
-  cachedOverridesFromGateway = true;
+export async function refreshOverridesFromGateway(): Promise<void> {
+  clearFeatureFlagOverridesCache();
+  await initFeatureFlagOverrides({ retryBackoffsMs: [] });
 }
 
 // ---------------------------------------------------------------------------
@@ -287,7 +280,7 @@ export function _setOverridesForTesting(
  *      values, which the gateway merges server-side: persisted > remote >
  *      registry)
  *   2. Registry `defaultEnabled` (for declared assistant-scope keys)
- *   3. `true` (for undeclared keys with no override)
+ *   3. `false` (for undeclared keys with no override)
  */
 export function isAssistantFeatureFlagEnabled(
   key: string,
@@ -304,6 +297,6 @@ export function isAssistantFeatureFlagEnabled(
   // 2. For declared keys, use the registry default.
   if (declared) return declared.defaultEnabled;
 
-  // 3. Undeclared keys with no override default to enabled.
-  return true;
+  // 3. Undeclared keys with no override fail closed.
+  return false;
 }

package/src/config/bundled-skills/document-editor/SKILL.md

@@ -16,6 +16,7 @@ Create and edit long-form documents using the built-in rich text editor. Documen
 
 ## Tools
 
+- **document_open** - Opens an existing document in the editor panel by `surface_id`. Use this when a document exists but isn't visible in the editor — for example after the user switches devices, refreshes the page, or when the editor panel was closed. Fetches the document from storage and sends it to the client.
 - **document_create** - Opens a new document editor with an optional title and initial Markdown content. Returns a `surface_id` for subsequent updates.
 - **document_update** - Updates content in an open document editor by `surface_id`. Supports `replace` (overwrite) and `append` (add to end) modes.
 - **document_read** - Reads the current content of a document by `surface_id` when it belongs to the current conversation, or when the current actor is the guardian/local user. Use to verify content before editing.
@@ -28,18 +29,24 @@ Create and edit long-form documents using the built-in rich text editor. Documen
 
 When the user asks to see, open, or pull up a document:
 
-1. Check the `<active_documents>` block in your context — it lists all documents in this conversation with their `surface_id` and title. If the document is there, call `document_read` with its `surface_id`. Done in one call.
+1. Check the `<active_documents>` block in your context — it lists all documents in this conversation with their `surface_id` and title.
 2. If the document is NOT in `<active_documents>`, call `document_list` with a `query` matching the document title. For guardian/local users, this searches across previous conversations and sessions.
-3. Once you have the `surface_id`, call `document_read` to retrieve the content.
+3. Once you have the `surface_id`, call `document_open` to open the editor panel. This both surfaces the editor on the client and returns the document content. If the user only needs the text (not the editor), use `document_read` instead.
 
 **Never** search the filesystem, conversation history, or archives to find a document. Always use `document_list` with a `query`.
 
+**If the user says they can't see a document you know exists** (e.g. after switching from macOS to web, or after a page refresh), call `document_open` with the `surface_id` to re-surface the editor panel on their current client.
+
 ## Creating a new document
 
-1. **Create the document**: Call `document_create` with a title (inferred from the request). Call the tool immediately, not after conversational preamble.
+1. **Create the document**: Call `document_create` with a title (inferred from the request). Call the tool immediately, not after conversational preamble. Capture the `surface_id` from the response — every subsequent `document_update` call must reference it.
 2. **Write content in Markdown**: Use proper structure (`#` for titles, `##` for sections), **bold**, _italic_, code blocks, tables, lists, blockquotes as appropriate.
 3. **CRITICAL - Stream content in chunks**: Call `document_update` MULTIPLE times, not just once. Break content into logical chunks (paragraphs, sections, or every 200-300 words). Call `document_update` with `mode: "append"` for EACH chunk separately. The user experiences real-time content appearing as you write.
 
+### Recovering from a failed update
+
+If a `document_update` call fails with an `Invalid input` error (for example because `surface_id` was missing), do NOT call `document_create` again. The `surface_id` you need is in the tool result of the most recent `document_create` call in this turn. Retry `document_update` with that `surface_id` and the same content. Creating a second document with the same title produces a duplicate for the user.
+
 ## Editing an existing document
 
 When the user requests changes to a document:
@@ -107,6 +114,12 @@ Users can leave inline comments on documents. Open comments are surfaced in a `<
 3. Call `comment_resolve` on comments you have addressed.
 4. If a comment is ambiguous, call `comment_reply` to ask for clarification instead of guessing.
 
+## Anti-Patterns
+
+- **Don't use `app_create` for blog posts, articles, or written content.** Use `document_create` — apps are for interactive content with state/data.
+- **Don't output the full content in chat.** The content goes in the document editor, not in the chat response. Acknowledge what you're doing and stream to the editor.
+- **Don't wait to generate everything before sending.** Stream content in chunks via `document_update` with `mode: "append"` so users see progress in real time.
+
 ## Usage Notes
 
 - The `mode` parameter on `document_update` defaults to `append`.

package/src/config/bundled-skills/document-editor/TOOLS.json

@@ -1,6 +1,24 @@
 {
   "version": 1,
   "tools": [
+    {
+      "name": "document_open",
+      "description": "Open an existing document in the editor panel. Use this when the user asks to see or pull up a document that exists but isn't currently visible in the editor — for example after switching devices, refreshing, or when the editor panel was closed.",
+      "category": "document-editor",
+      "risk": "low",
+      "input_schema": {
+        "type": "object",
+        "properties": {
+          "surface_id": {
+            "type": "string",
+            "description": "The ID of the document to open"
+          }
+        },
+        "required": ["surface_id"]
+      },
+      "executor": "tools/document-open.ts",
+      "execution_target": "host"
+    },
     {
       "name": "document_create",
       "description": "Create a new long-form document with a rich text editor. Use this when the user asks to write a blog post, article, or any long-form content. The editor opens in workspace mode with chat docked to the side.",

package/src/config/bundled-skills/document-editor/tools/document-open.ts

@@ -0,0 +1,12 @@
+import { executeDocumentOpen } from "../../../../tools/document/document-tool.js";
+import type {
+  ToolContext,
+  ToolExecutionResult,
+} from "../../../../tools/types.js";
+
+export async function run(
+  input: Record<string, unknown>,
+  context: ToolContext,
+): Promise<ToolExecutionResult> {
+  return executeDocumentOpen(input, context);
+}

package/src/config/bundled-skills/image-studio/SKILL.md

@@ -37,3 +37,7 @@ You are an image generation assistant. When the user asks you to create or edit
 - When editing images, clearly describe what changes you want made to the source image.
 - Use the `variants` parameter (1-4) to generate multiple options and pick the best one.
 - If no API key is configured for the selected model's provider (Gemini or OpenAI), the tool will return an error - ask the user to set one up.
+
+## Error handling
+
+When image generation fails, report the error to the user as-is. **Do not** attempt to fix the error by changing service configuration (e.g. switching between "managed" and "your-own" mode, or changing the provider/model). Service configuration changes should only be made at the user's explicit request via Settings.

package/src/config/bundled-skills/image-studio/tools/media-generate-image.ts

@@ -30,7 +30,7 @@ export async function run(
   });
   if (!credentials) {
     return {
-      content: errorHint ?? "Image generation is not configured.",
+      content: `${errorHint ?? "Image generation is not configured."}\n\nReport this error to the user. Do not change service configuration (mode, provider, or model) to try to fix it.`,
       isError: true,
     };
   }
@@ -131,7 +131,7 @@ export async function run(
     };
   } catch (error) {
     return {
-      content: mapImageGenError(provider, error),
+      content: `${mapImageGenError(provider, error)}\n\nReport this error to the user. Do not change service configuration (mode, provider, or model) to try to fix it.`,
       isError: true,
     };
   }

package/src/config/bundled-skills/media-processing/tools/ingest-media.ts

@@ -1,7 +1,10 @@
 import { access } from "node:fs/promises";
 import { basename, extname } from "node:path";
 
-import { enqueueMemoryJob } from "../../../../memory/jobs-store.js";
+import {
+  enqueueMemoryJob,
+  isMemoryEnabled,
+} from "../../../../memory/jobs-store.js";
 import {
   computeFileHashStreaming,
   createProcessingStage,
@@ -199,13 +202,15 @@ export async function run(
   updateMediaAssetStatus(asset.id, "processing");
 
   // Enqueue a processing job via the existing jobs framework
-  enqueueMemoryJob("media_processing", {
-    mediaAssetId: asset.id,
-    stage: "ingest",
-    filePath,
-    mimeType,
-    mediaType,
-  });
+  if (isMemoryEnabled()) {
+    enqueueMemoryJob("media_processing", {
+      mediaAssetId: asset.id,
+      stage: "ingest",
+      filePath,
+      mimeType,
+      mediaType,
+    });
+  }
 
   context.onOutput?.(`Registered media asset: ${asset.id}\n`);
 

package/src/config/bundled-skills/messaging/tools/messaging-analyze-style.ts

@@ -2,7 +2,10 @@ import { and, eq, sql } from "drizzle-orm";
 import { v4 as uuid } from "uuid";
 
 import { getDb } from "../../../../memory/db-connection.js";
-import { enqueueMemoryJob } from "../../../../memory/jobs-store.js";
+import {
+  enqueueMemoryJob,
+  isMemoryEnabled,
+} from "../../../../memory/jobs-store.js";
 import { memoryGraphNodes } from "../../../../memory/schema.js";
 import { clampUnitInterval } from "../../../../memory/validation.js";
 import { extractStylePatterns } from "../../../../messaging/style-analyzer.js";
@@ -56,7 +59,9 @@ function upsertMemoryItem(opts: {
       })
       .where(eq(memoryGraphNodes.id, existing.id))
       .run();
-    enqueueMemoryJob("embed_graph_node", { nodeId: existing.id });
+    if (isMemoryEnabled()) {
+      enqueueMemoryJob("embed_graph_node", { nodeId: existing.id });
+    }
   } else {
     const id = uuid();
     db.insert(memoryGraphNodes)
@@ -82,7 +87,9 @@ function upsertMemoryItem(opts: {
         scopeId: opts.scopeId,
       })
       .run();
-    enqueueMemoryJob("embed_graph_node", { nodeId: id });
+    if (isMemoryEnabled()) {
+      enqueueMemoryJob("embed_graph_node", { nodeId: id });
+    }
   }
 }
 

package/src/config/bundled-skills/phone-calls/references/TRANSCRIPTS.md

@@ -1,13 +1,15 @@
 # Retrieving Past Call Transcripts
 
-After a call ends, the full bidirectional transcript (caller speech, assistant responses, tool calls, and tool results) is stored in the SQLite database. The daemon logs (`vellum.log`) only contain caller-side transcripts and lifecycle events at the default log level, so they are **not sufficient** for full transcript reconstruction.
+After a call ends, the full bidirectional transcript (caller speech, assistant responses, tool calls, and tool results) is stored in the SQLite database. The assistant logs only contain caller-side transcripts and lifecycle events at the default log level, so they are **not sufficient** for full transcript reconstruction.
+
+Logs rotate daily as `$VELLUM_WORKSPACE_DIR/data/logs/assistant-YYYY-MM-DD.log`. The commands below use today's UTC log file; if a call happened on a previous day, swap the date in or grep across `assistant-*.log`.
 
 ## Finding the conversation
 
-1. **Get the call session ID and voice conversation ID** from `vellum.log` by searching for recent session creation entries:
+1. **Get the call session ID and voice conversation ID** from today's log by searching for recent session creation entries:
 
 ```bash
-grep "voiceConversationId" ~/.vellum/workspace/data/logs/vellum.log | tail -5
+grep "voiceConversationId" "$VELLUM_WORKSPACE_DIR/data/logs/assistant-$(date -u +%Y-%m-%d).log" | tail -5
 ```
 
 The `voiceConversationId` field in the `Created new inbound voice session` (or outbound equivalent) log line is the key you need.
@@ -15,7 +17,7 @@ The `voiceConversationId` field in the `Created new inbound voice session` (or o
 2. **Query the messages table** in the SQLite database using the voice conversation ID:
 
 ```bash
-sqlite3 ~/.vellum/workspace/data/db/assistant.db \
+sqlite3 "$VELLUM_WORKSPACE_DIR/data/db/assistant.db" \
   "SELECT role, content FROM messages WHERE conversation_id = '<voiceConversationId>' ORDER BY created_at ASC;"
 ```
 
@@ -27,9 +29,9 @@ This returns all messages in chronological order with:
 ## Quick one-liner for the most recent call
 
 ```bash
-CONV_ID=$(grep "voiceConversationId" ~/.vellum/workspace/data/logs/vellum.log | tail -1 | python3 -c "import sys,json; print(json.loads(sys.stdin.readline().strip())['voiceConversationId'])")
+CONV_ID=$(grep -h "voiceConversationId" "$VELLUM_WORKSPACE_DIR"/data/logs/assistant-*.log | tail -1 | python3 -c "import sys,json; print(json.loads(sys.stdin.readline().strip())['voiceConversationId'])")
 
-sqlite3 ~/.vellum/workspace/data/db/assistant.db \
+sqlite3 "$VELLUM_WORKSPACE_DIR/data/db/assistant.db" \
   "SELECT role, content FROM messages WHERE conversation_id = '$CONV_ID' ORDER BY created_at ASC;"
 ```
 
@@ -44,14 +46,14 @@ sqlite3 ~/.vellum/workspace/data/db/assistant.db \
 
 ## Key paths
 
-| Resource                                      | Path                                       |
-| --------------------------------------------- | ------------------------------------------ |
-| Assistant logs (caller-side transcripts only) | `~/.vellum/workspace/data/logs/vellum.log` |
-| Full conversation database                    | `~/.vellum/workspace/data/db/assistant.db` |
-| Messages table                                | `messages` (keyed by `conversation_id`)    |
-| Call sessions table                           | `call_sessions`                            |
-| Call events table                             | `call_events`                              |
+| Resource                                      | Path                                                  |
+| --------------------------------------------- | ----------------------------------------------------- |
+| Assistant logs (caller-side transcripts only) | `$VELLUM_WORKSPACE_DIR/data/logs/assistant-*.log`     |
+| Full conversation database                    | `$VELLUM_WORKSPACE_DIR/data/db/assistant.db`          |
+| Messages table                                | `messages` (keyed by `conversation_id`)               |
+| Call sessions table                           | `call_sessions`                                       |
+| Call events table                             | `call_events`                                         |
 
 ## Important
 
-`vellum.log` at the default log level does **not** contain assistant responses, TTS text, or LLM completions for voice calls. Always use the `messages` table in `assistant.db` as the source of truth for complete call transcripts.
+Assistant log files at the default log level do **not** contain assistant responses, TTS text, or LLM completions for voice calls. Always use the `messages` table in `assistant.db` as the source of truth for complete call transcripts.

package/src/config/bundled-skills/playbooks/tools/playbook-create.ts

@@ -3,7 +3,10 @@ import { sql } from "drizzle-orm";
 import { getDb } from "../../../../memory/db-connection.js";
 import { createNode, updateNode } from "../../../../memory/graph/store.js";
 import type { NewNode } from "../../../../memory/graph/types.js";
-import { enqueueMemoryJob } from "../../../../memory/jobs-store.js";
+import {
+  enqueueMemoryJob,
+  isMemoryEnabled,
+} from "../../../../memory/jobs-store.js";
 import { memoryGraphNodes } from "../../../../memory/schema.js";
 import type {
   Playbook,
@@ -116,7 +119,9 @@ export async function executePlaybookCreate(
       sourceConversations: [`playbook:${node.id}`],
     });
 
-    enqueueMemoryJob("embed_graph_node", { nodeId: node.id });
+    if (isMemoryEnabled()) {
+      enqueueMemoryJob("embed_graph_node", { nodeId: node.id });
+    }
 
     const autonomyLabel =
       autonomyLevel === "auto"

package/src/config/bundled-skills/playbooks/tools/playbook-update.ts

@@ -2,7 +2,10 @@ import { and, eq, sql } from "drizzle-orm";
 
 import { getDb } from "../../../../memory/db-connection.js";
 import { getNode, updateNode } from "../../../../memory/graph/store.js";
-import { enqueueMemoryJob } from "../../../../memory/jobs-store.js";
+import {
+  enqueueMemoryJob,
+  isMemoryEnabled,
+} from "../../../../memory/jobs-store.js";
 import { memoryGraphNodes } from "../../../../memory/schema.js";
 import type {
   Playbook,
@@ -123,7 +126,9 @@ export async function executePlaybookUpdate(
       lastAccessed: Date.now(),
     });
 
-    enqueueMemoryJob("embed_graph_node", { nodeId: existing.id });
+    if (isMemoryEnabled()) {
+      enqueueMemoryJob("embed_graph_node", { nodeId: existing.id });
+    }
 
     const autonomyLabel =
       updated.autonomyLevel === "auto"

package/src/config/bundled-skills/schedule/SKILL.md

@@ -100,7 +100,7 @@ Use `notify` for simple reminders ("remind me to take medicine at 9am"), `execut
 
 ## Conversation Reuse
 
-By default, each schedule run creates a new conversation. For recurring schedules that benefit from accumulating context across runs (e.g. polling-style jobs, daily digests that reference prior results), set `reuse_conversation: true`. When enabled, subsequent runs reuse the conversation from the last successful run instead of creating a new one.
+Recurring schedules reuse the same conversation across runs by default — subsequent runs continue the conversation from the last successful run, preserving context and channel thread continuity. Set `reuse_conversation: false` explicitly if each run should start with a fresh conversation (e.g. independent reports that shouldn't accumulate prior context). One-shot schedules always create a fresh conversation.
 
 - Only applies to **recurring** schedules; ignored for one-shot schedules.
 - If the prior conversation has been deleted, a new one is created automatically.

package/src/config/bundled-skills/schedule/TOOLS.json

@@ -62,7 +62,7 @@
           },
           "reuse_conversation": {
             "type": "boolean",
-            "description": "When true, reuse the same conversation across recurring schedule runs instead of creating a new one each time. Useful for polling-style schedules that accumulate context over time. Ignored for one-shot schedules. Defaults to false."
+            "description": "When true, reuse the same conversation across recurring schedule runs instead of creating a new one each time. Defaults to true for recurring schedules, false for one-shot. Set to false explicitly if each run should start fresh."
           },
           "max_retries": {
             "type": "integer",
@@ -161,7 +161,7 @@
           },
           "reuse_conversation": {
             "type": "boolean",
-            "description": "When true, reuse the same conversation across recurring schedule runs instead of creating a new one each time. Useful for polling-style schedules that accumulate context over time. Ignored for one-shot schedules."
+            "description": "When true, reuse the same conversation across recurring schedule runs instead of creating a new one each time. Defaults to true for recurring schedules. Set to false explicitly if each run should start fresh."
           },
           "max_retries": {
             "type": "integer",

package/src/config/bundled-skills/settings/tools/open-system-settings.ts

@@ -42,6 +42,7 @@ export async function run(
     context.sendToClient({
       type: "open_url",
       url: meta.url,
+      conversationId: context.conversationId,
     });
   }
 

package/src/config/bundled-tool-registry.ts

@@ -58,6 +58,7 @@ import * as documentCreate from "./bundled-skills/document-editor/tools/document
 import * as documentDelete from "./bundled-skills/document-editor/tools/document-delete.js";
 import * as documentFind from "./bundled-skills/document-editor/tools/document-find.js";
 import * as documentList from "./bundled-skills/document-editor/tools/document-list.js";
+import * as documentOpen from "./bundled-skills/document-editor/tools/document-open.js";
 import * as documentRead from "./bundled-skills/document-editor/tools/document-read.js";
 import * as documentReplaceText from "./bundled-skills/document-editor/tools/document-replace-text.js";
 import * as documentUpdate from "./bundled-skills/document-editor/tools/document-update.js";
@@ -175,6 +176,7 @@ export const bundledToolRegistry = new Map<string, SkillToolScript>([
   ["contacts:tools/google-contacts.ts", googleContacts],
 
   // document-editor
+  ["document-editor:tools/document-open.ts", documentOpen],
   ["document-editor:tools/document-create.ts", documentCreate],
   ["document-editor:tools/document-update.ts", documentUpdate],
   ["document-editor:tools/document-read.ts", documentRead],

package/src/config/call-site-defaults.ts

@@ -23,7 +23,7 @@ export const CALL_SITE_DEFAULTS: Record<LLMCallSite, CallSiteDefaultConfig> = {
   emptyStateGreeting: { profile: "balanced" },
 
   memoryRouter: {
-    profile: "balanced",
+    profile: "cost-optimized",
     contextWindow: { maxInputTokens: 1000000 },
   },
   recall: {
@@ -47,6 +47,13 @@ export const CALL_SITE_DEFAULTS: Record<LLMCallSite, CallSiteDefaultConfig> = {
   memoryV2Migration: { profile: "cost-optimized" },
   memoryV2Sweep: { profile: "cost-optimized" },
   memoryV2Consolidation: { profile: "balanced" },
+  // memory v3: cheap filter + descent, capable gate. All three are
+  // selection/classification calls, not generation — pin temperature to 0 so
+  // the same candidate set yields the same keep/descend/select decision instead
+  // of sampling-driven variance.
+  memoryV3Filter: { profile: "cost-optimized", temperature: 0 },
+  memoryV3Descent: { profile: "cost-optimized", temperature: 0 },
+  memoryV3Gate: { profile: "balanced", temperature: 0 },
   conversationSummarization: { profile: "cost-optimized" },
   conversationTitle: { profile: "cost-optimized" },
   approvalCopy: { profile: "cost-optimized" },
@@ -115,10 +122,4 @@ export const CALL_SITE_DEFAULTS: Record<LLMCallSite, CallSiteDefaultConfig> = {
     effort: "low",
     thinking: { enabled: false },
   },
-  queryComplexityRouter: {
-    profile: "cost-optimized",
-    maxTokens: 16,
-    effort: "low",
-    thinking: { enabled: false },
-  },
 };