@vellumai/assistant 0.5.13 → 0.5.14

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

@@ -56,16 +56,16 @@ When the user asks to "connect my email", "set up email", "manage my email", or
 
 ### Gmail
 
-1. **Try connecting directly first.** Call `credential_store` with `action: "oauth2_connect"` and `service: "google"`. The tool auto-fills Google's OAuth endpoints and looks up any previously stored client credentials - so this single call may be all that's needed.
-2. **If it fails because no client_id is found:** The user needs to create Google Cloud OAuth credentials first. Load the **google-oauth-app-setup** skill:
-   - Call `skill_load` with `skill: "google-oauth-app-setup"` to load the dependency skill.
-   - Tell the user Gmail isn't connected yet and briefly explain what the setup involves, then use `ui_show` with `surface_type: "confirmation"` to ask for permission to start:
+1. **Try connecting directly first.** Run `assistant oauth status google`. This will show whether or not the user had previously connected their google account. If so, they are ready to go.
+2. **If no connections are found:** The user needs to either use Vellum's managed google integration or set up their own google oauth app.
+   - Call `skill_load` with `skill: "vellum-oauth-integrations"` with `provider-key: google` throughout.
+   - To use `your-own` mode, you will need to call `skill_load` with `skill: google-oauth-app-setup`. In this case:
+     - Tell the user Gmail isn't connected yet and briefly explain what the setup involves, then use `ui_show` with `surface_type: "confirmation"` to ask for permission to start:
      - **message:** "Ready to set up Gmail?"
      - **detail:** "I'll open a few pages in your browser and walk you through setting up Google Cloud credentials - creating a project, enabling APIs, and connecting your account. Takes about 5 minutes."
      - **confirmLabel:** "Get Started"
      - **cancelLabel:** "Not Now"
-   - If the user confirms, briefly acknowledge (e.g., "Setting up Gmail now...") and proceed with the setup guide. If they decline, acknowledge and let them know they can set it up later.
-3. **If the user provides a client_id directly in chat:** Call `credential_store` with `action: "oauth2_connect"`, `service: "google"`, and `client_id: "<their value>"`. If a `client_secret` is also needed, collect it securely via `credential_store` with `action: "prompt"` — never accept it pasted in chat. Everything else is auto-filled.
+     - If the user confirms, briefly acknowledge (e.g., "Setting up Gmail now...") and proceed with the setup guide. If they decline, acknowledge and let them know they can set it up later.
 
 ### Slack
 
@@ -95,7 +95,7 @@ The guardian-verify-setup skill handles the full outbound verification flow for
 
 When a messaging tool fails with a token or authorization error:
 
-1. **Try to reconnect silently.** Call `credential_store` with `action: "oauth2_connect"` and the appropriate `service`. This often resolves expired tokens automatically.
+1. **Try to reconnect silently.** Run `assistant oauth ping <provider>`. This often resolves expired tokens automatically.
 2. **If reconnection fails, go straight to setup.** Don't present options, ask which route the user prefers, or explain what went wrong technically. Just tell the user briefly (e.g., "Gmail needs to be reconnected - let me set that up") and immediately follow the connection setup flow for that platform (e.g., install and load **google-oauth-app-setup** for Gmail). The user came to you to get something done, not to troubleshoot OAuth - make it seamless.
 3. **Never try alternative approaches.** Don't use bash, curl, browser automation, or any workaround. If the messaging tools can't do it, the reconnection flow is the answer.
 4. **Never expose error details.** The user doesn't need to see error messages about tokens, OAuth, or API failures. Translate errors into plain language.

package/src/config/bundled-skills/messaging/tools/messaging-send.ts

@@ -1,6 +1,12 @@
 import { readFile } from "node:fs/promises";
 import { basename } from "node:path";
 
+import {
+  addMessage,
+  getConversation,
+} from "../../../../memory/conversation-crud.js";
+import { syncMessageToDisk } from "../../../../memory/conversation-disk-view.js";
+import { getBindingByChannelChat } from "../../../../memory/external-conversation-store.js";
 import {
   batchGetMessages,
   createDraft,
@@ -13,6 +19,7 @@ import type {
   ToolContext,
   ToolExecutionResult,
 } from "../../../../tools/types.js";
+import { getLogger } from "../../../../util/logger.js";
 import { guessMimeType } from "./gmail-mime-helpers.js";
 import {
   err,
@@ -24,6 +31,8 @@ import {
   resolveProvider,
 } from "./shared.js";
 
+const log = getLogger("messaging-send");
+
 export async function run(
   input: Record<string, unknown>,
   context: ToolContext,
@@ -219,6 +228,34 @@ export async function run(
     const threadSuffix = result.threadId
       ? `, "thread_id": "${result.threadId}"`
       : "";
+
+    // Cross-post to the conversation bound to this channel so replies have context.
+    try {
+      const binding = getBindingByChannelChat(provider.id, conversationId);
+      if (binding && binding.conversationId !== context.conversationId) {
+        const boundConv = getConversation(binding.conversationId);
+        if (boundConv) {
+          const crossPosted = await addMessage(
+            binding.conversationId,
+            "assistant",
+            JSON.stringify([{ type: "text", text }]),
+            { automated: true, crossPostedFrom: context.conversationId },
+            { skipIndexing: true },
+          );
+          syncMessageToDisk(
+            binding.conversationId,
+            crossPosted.id,
+            boundConv.createdAt,
+          );
+        }
+      }
+    } catch (e) {
+      log.warn(
+        { err: e, provider: provider.id, externalChatId: conversationId },
+        "Failed to cross-post outbound message to bound conversation",
+      );
+    }
+
     return ok(`Message sent (ID: ${result.id}${threadSuffix}).`);
   } catch (e) {
     return err(e instanceof Error ? e.message : String(e));

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

@@ -62,6 +62,20 @@ When the user asks to scan or catch up on Slack:
 3. Offer to drill into specific threads with `conversations.replies`
 4. Summarize with attribution: who said what, decisions made, open questions
 
+## User Resolution
+
+When you need to send a DM or look up a Slack user by name, check contacts first to avoid redundant API calls:
+
+1. **Before calling `users.list`**: Use `contact_search` with `query: "<name>"` and `channel_type: "slack"`. If a matching contact has `externalUserId` (Slack user ID) and `externalChatId` (DM channel ID), skip the API lookups and use those IDs directly with `chat.postMessage`.
+
+2. **After resolving via API**: When you had to call `users.list` or `conversations.open` to resolve a user, save the mapping with `contact_upsert` so future lookups are instant:
+   ```
+   contact_upsert {
+     display_name: "<display name>"
+     channels: [{ type: "slack", address: "<slack handle>", external_user_id: "<U...>", external_chat_id: "<D...>" }]
+   }
+   ```
+
 ## Privacy Rules
 
 **Channel privacy must be respected at all times:**
@@ -86,6 +100,10 @@ When responding to messages from Slack channels, replies should be threaded. Pas
 
 Before making any Slack API calls, verify that Slack is connected. If not connected, load the **slack-app-setup** skill (`skill_load` with `skill: "slack-app-setup"`) and follow its guided flow. Do NOT improvise setup instructions — the `slack-app-setup` skill is the single source of truth. Slack uses Socket Mode and does not require redirect URLs or any OAuth flow.
 
+## Error Handling
+
+If a Slack API call fails due to missing or invalid credentials — for example, an error indicating that `slack_channel/bot_token` is not found, the credential is missing, or the token is invalid — do NOT attempt to fix the credentials manually. Instead, load the **slack-app-setup** skill (`skill_load` with `skill: "slack-app-setup"`) and follow its guided flow to set up or reconnect Slack. Tell the user something like "Slack needs to be reconnected" and start the setup skill.
+
 ## Communication Style
 
 - **Be action-oriented.** When the user asks to check Slack, start scanning immediately.

package/src/config/env-registry.ts

@@ -27,15 +27,6 @@ function flag(name: string): boolean {
 // ── Registry ─────────────────────────────────────────────────────────────────
 // Each entry documents the env var name, type, default, and purpose.
 
-/**
- * BASE_DATA_DIR — string, default: os.homedir()
- * Overrides the home directory used as the base for ~/.vellum and lockfiles.
- * Primarily used in tests to isolate filesystem state.
- */
-export function getBaseDataDir(): string | undefined {
-  return str("BASE_DATA_DIR");
-}
-
 /**
  * DEBUG_STDOUT_LOGS — boolean, default: false
  * Enables additional log output to stdout (alongside file logging).
@@ -47,13 +38,26 @@ export function getDebugStdoutLogs(): boolean {
 /**
  * IS_CONTAINERIZED — boolean, default: false
  * When true, indicates the assistant is running inside a container (e.g. Docker).
- * Any new data that needs to survive restarts must be written to BASE_DATA_DIR,
- * which is mapped to a persistent volume.
+ * Persistent data is stored in VELLUM_WORKSPACE_DIR (mapped to a dedicated volume).
  */
 export function getIsContainerized(): boolean {
   return flag("IS_CONTAINERIZED");
 }
 
+/**
+ * Whether this assistant is running as a platform-managed remote instance.
+ *
+ * Currently this is determined solely by the IS_CONTAINERIZED env var, which
+ * the platform sets when provisioning assistant containers. This is not ideal
+ * because any Docker environment could set it. We plan to replace this with a
+ * less spoof-able mechanism in the future — e.g. a signed platform token
+ * verified via asymmetric cryptography or an authenticated attestation
+ * endpoint on the platform.
+ */
+export function isPlatformRemote(): boolean {
+  return getIsContainerized();
+}
+
 /**
  * VELLUM_WORKSPACE_DIR — string, default: undefined
  * Overrides the default workspace directory.

package/src/config/env.ts

@@ -8,7 +8,7 @@
  * - Fail-fast validation via validateEnv() at startup
  * - Shared derived values (e.g. gateway base URL) instead of duplicated logic
  *
- * Bootstrap-level env vars (BASE_DATA_DIR, DEBUG_STDOUT_LOGS) are defined
+ * Bootstrap-level env vars (IS_CONTAINERIZED, DEBUG_STDOUT_LOGS) are defined
  * in config/env-registry.ts which has no internal dependencies and can be
  * imported from platform/logger without circular imports.
  */
@@ -216,16 +216,6 @@ export function getPlatformInternalApiKey(): string {
   return str("PLATFORM_INTERNAL_API_KEY") ?? "";
 }
 
-// ── Telemetry ──────────────────────────────────────────────────────────────────
-
-export function getTelemetryPlatformUrl(): string {
-  return str("TELEMETRY_PLATFORM_URL") ?? "";
-}
-
-export function getTelemetryAppToken(): string {
-  return str("TELEMETRY_APP_TOKEN") ?? "";
-}
-
 // ── Startup validation ──────────────────────────────────────────────────────
 
 /**

package/src/config/feature-flag-registry.json

@@ -289,6 +289,14 @@
       "description": "Auto-expand completed tool call step groups instead of showing them collapsed",
       "defaultEnabled": false
     },
+    {
+      "id": "show-thinking-blocks",
+      "scope": "macos",
+      "key": "show-thinking-blocks",
+      "label": "Show Thinking Blocks",
+      "description": "Display the assistant's thinking/reasoning inline in chat messages as collapsible blocks",
+      "defaultEnabled": false
+    },
     {
       "id": "inline-skill-commands",
       "scope": "assistant",
@@ -344,6 +352,14 @@
       "label": "Voice Mode",
       "description": "Show the live voice conversation button in the composer",
       "defaultEnabled": false
+    },
+    {
+      "id": "fast-mode",
+      "scope": "assistant",
+      "key": "fast-mode",
+      "label": "Fast Mode",
+      "description": "Enable Anthropic fast mode for Opus 4.6, delivering up to 2.5x higher output tokens per second at premium pricing",
+      "defaultEnabled": false
     }
   ]
 }

package/src/config/schema.ts

@@ -49,6 +49,7 @@ export type {
   ContextWindowConfig,
   Effort,
   ModelPricingOverride,
+  Speed,
   ThinkingConfig,
 } from "./schemas/inference.js";
 export {
@@ -56,6 +57,7 @@ export {
   ContextWindowConfigSchema,
   EffortSchema,
   ModelPricingOverrideSchema,
+  SpeedSchema,
   ThinkingConfigSchema,
 } from "./schemas/inference.js";
 export type {
@@ -201,6 +203,7 @@ import {
   ContextWindowConfigSchema,
   EffortSchema,
   ModelPricingOverrideSchema,
+  SpeedSchema,
   ThinkingConfigSchema,
 } from "./schemas/inference.js";
 import { IngressConfigSchema } from "./schemas/ingress.js";
@@ -240,6 +243,7 @@ export const AssistantConfigSchema = z
       .default(16000)
       .describe("Maximum number of output tokens per LLM response"),
     effort: EffortSchema,
+    speed: SpeedSchema,
     thinking: ThinkingConfigSchema.default(ThinkingConfigSchema.parse({})),
     contextWindow: ContextWindowConfigSchema.default(
       ContextWindowConfigSchema.parse({}),

package/src/config/schemas/heartbeat.ts

@@ -1,10 +1,12 @@
 import { z } from "zod";
 
+import { SpeedSchema } from "./inference.js";
+
 export const HeartbeatConfigSchema = z
   .object({
     enabled: z
       .boolean({ error: "heartbeat.enabled must be a boolean" })
-      .default(false)
+      .default(true)
       .describe("Whether periodic heartbeat checks are enabled"),
     intervalMs: z
       .number({ error: "heartbeat.intervalMs must be a number" })
@@ -12,6 +14,9 @@ export const HeartbeatConfigSchema = z
       .positive("heartbeat.intervalMs must be a positive integer")
       .default(3_600_000)
       .describe("Time between heartbeat checks in milliseconds"),
+    speed: SpeedSchema.default("standard").describe(
+      "Inference speed mode for heartbeat conversations — defaults to standard to avoid inheriting the global fast mode multiplier",
+    ),
     activeHoursStart: z
       .number({ error: "heartbeat.activeHoursStart must be a number" })
       .int("heartbeat.activeHoursStart must be an integer")

package/src/config/schemas/inference.ts

@@ -12,13 +12,13 @@ export const ThinkingConfigSchema = z
   .object({
     enabled: z
       .boolean({ error: "thinking.enabled must be a boolean" })
-      .default(false)
+      .default(true)
       .describe(
         "Whether to enable extended thinking (chain-of-thought) for the model",
       ),
     streamThinking: z
       .boolean({ error: "thinking.streamThinking must be a boolean" })
-      .default(false)
+      .default(true)
       .describe(
         "Whether to stream the model's thinking tokens to the client in real time",
       ),
@@ -29,13 +29,24 @@ export const EffortSchema = z
   .enum(["low", "medium", "high", "max"], {
     error: 'effort must be "low", "medium", "high", or "max"',
   })
-  .default("high")
+  .default("max")
   .describe(
     "How much effort the model should put into its response — lower effort is faster and cheaper",
   );
 
 export type Effort = z.infer<typeof EffortSchema>;
 
+export const SpeedSchema = z
+  .enum(["standard", "fast"], {
+    error: 'speed must be "standard" or "fast"',
+  })
+  .default("standard")
+  .describe(
+    'Inference speed mode — "fast" enables higher output token throughput on supported models (Opus 4.6) at premium pricing',
+  );
+
+export type Speed = z.infer<typeof SpeedSchema>;
+
 export const ContextOverflowRecoveryConfigSchema = z
   .object({
     enabled: z

package/src/config/schemas/memory-processing.ts

@@ -8,14 +8,6 @@ export const MemoryExtractionConfigSchema = z
       .describe(
         "Whether to use an LLM for extracting structured memory items from conversations",
       ),
-    modelIntent: z
-      .enum(["latency-optimized", "quality-optimized", "vision-optimized"], {
-        error: "memory.extraction.modelIntent must be a valid model intent",
-      })
-      .default("quality-optimized")
-      .describe(
-        "Model selection strategy for extraction — trade off speed vs quality",
-      ),
     extractFromAssistant: z
       .boolean({
         error: "memory.extraction.extractFromAssistant must be a boolean",
@@ -24,6 +16,22 @@ export const MemoryExtractionConfigSchema = z
       .describe(
         "Whether to extract memory items from the assistant's own messages (in addition to user messages)",
       ),
+    batchSize: z
+      .number()
+      .int()
+      .positive()
+      .default(10)
+      .describe(
+        "Number of unextracted messages before triggering batch extraction",
+      ),
+    idleTimeoutMs: z
+      .number()
+      .int()
+      .positive()
+      .default(300000)
+      .describe(
+        "Milliseconds of idle time before triggering extraction of pending messages",
+      ),
   })
   .describe("Controls how memory items are extracted from conversations");
 

package/src/config/schemas/memory-retrieval.ts

@@ -19,7 +19,7 @@ export const MemoryDynamicBudgetConfigSchema = z
       .positive(
         "memory.retrieval.dynamicBudget.minInjectTokens must be a positive integer",
       )
-      .default(1200)
+      .default(2400)
       .describe(
         "Minimum number of tokens to inject from memory, even when context space is limited",
       ),
@@ -32,7 +32,7 @@ export const MemoryDynamicBudgetConfigSchema = z
       .positive(
         "memory.retrieval.dynamicBudget.maxInjectTokens must be a positive integer",
       )
-      .default(10000)
+      .default(16000)
       .describe(
         "Maximum number of tokens to inject from memory, even when plenty of context space is available",
       ),
@@ -190,7 +190,7 @@ export const MemoryRetrievalConfigSchema = z
       .number({ error: "memory.retrieval.maxInjectTokens must be a number" })
       .int("memory.retrieval.maxInjectTokens must be an integer")
       .positive("memory.retrieval.maxInjectTokens must be a positive integer")
-      .default(10000)
+      .default(16000)
       .describe(
         "Maximum number of tokens to inject from long-term memory into the conversation context",
       ),

package/src/config/skills.ts

@@ -18,7 +18,6 @@ import {
 
 import { z } from "zod";
 
-import { stripCommentLines } from "../prompts/system-prompt.js";
 import {
   extractAllText,
   getConfiguredProvider,
@@ -34,6 +33,7 @@ import {
   getWorkspaceDirDisplay,
   getWorkspaceSkillsDir,
 } from "../util/platform.js";
+import { stripCommentLines } from "../util/strip-comment-lines.js";
 import { isAssistantFeatureFlagEnabled } from "./assistant-feature-flags.js";
 import { getConfig } from "./loader.js";