@vellumai/assistant 0.10.4 → 0.10.5-dev.202607022329.d89bf2f

package/src/config/acp-defaults.test.ts

@@ -10,7 +10,6 @@ describe("DEFAULT_ACP_AGENT_PROFILES", () => {
     expect(Object.keys(DEFAULT_ACP_AGENT_PROFILES).sort()).toEqual([
       "claude",
       "codex",
-      "gemini",
     ]);
   });
 
@@ -30,14 +29,6 @@ describe("DEFAULT_ACP_AGENT_PROFILES", () => {
     });
   });
 
-  test("gemini profile speaks native ACP via gemini --acp (no adapter binary)", () => {
-    expect(DEFAULT_ACP_AGENT_PROFILES.gemini).toEqual({
-      command: "gemini",
-      args: ["--acp"],
-      description: "Google Gemini CLI (native ACP via gemini --acp)",
-    });
-  });
-
   test("is deeply frozen so mutation throws in strict mode", () => {
     expect(Object.isFrozen(DEFAULT_ACP_AGENT_PROFILES)).toBe(true);
     for (const profile of Object.values(DEFAULT_ACP_AGENT_PROFILES)) {
@@ -55,7 +46,6 @@ describe("DEFAULT_AGENT_NPM_PACKAGES", () => {
     expect(DEFAULT_AGENT_NPM_PACKAGES).toEqual({
       "claude-agent-acp": "@agentclientprotocol/claude-agent-acp",
       "codex-acp": "@zed-industries/codex-acp",
-      gemini: "@google/gemini-cli",
     });
   });
 

package/src/config/acp-defaults.ts

@@ -30,11 +30,6 @@ export const DEFAULT_ACP_AGENT_PROFILES: Readonly<
     args: FROZEN_EMPTY_ARGS,
     description: "OpenAI Codex CLI (via @zed-industries/codex-acp)",
   }),
-  gemini: Object.freeze({
-    command: "gemini",
-    args: Object.freeze(["--acp"]) as unknown as string[],
-    description: "Google Gemini CLI (native ACP via gemini --acp)",
-  }),
 });
 
 /**
@@ -49,5 +44,4 @@ export const DEFAULT_AGENT_NPM_PACKAGES: Readonly<Record<string, string>> =
   Object.freeze({
     "claude-agent-acp": "@agentclientprotocol/claude-agent-acp",
     "codex-acp": "@zed-industries/codex-acp",
-    gemini: "@google/gemini-cli",
   });

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

@@ -8,26 +8,26 @@ metadata:
     display-name: "ACP"
     category: "development"
     activation-hints:
-      - "User asks to use Claude Code, Codex, or Gemini to do something"
-      - "User wants to delegate a coding task to Claude Code, Codex, Gemini, or another ACP agent"
+      - "User asks to use Claude Code or Codex to do something"
+      - "User wants to delegate a coding task to Claude Code, Codex, or another ACP agent"
       - "User wants to hand a coding task to another agent and check on it later"
       - "User wants to spawn an external coding agent that runs autonomously and streams results back"
-      - "User mentions ACP, claude-agent-acp, codex-acp, gemini --acp, or running multiple coding agents in parallel"
+      - "User mentions ACP, claude-agent-acp, codex-acp, or running multiple coding agents in parallel"
     avoid-when:
       - "Task is small enough to do inline with the assistant's own tools - no need for an external agent"
 ---
 
-ACP agent orchestration - spawn external coding agents (Claude Code, Codex, Gemini) to work on tasks via the Agent Client Protocol. Each agent runs as its own subprocess speaking ACP over stdio and streams results back into the conversation.
+ACP agent orchestration - spawn external coding agents (Claude Code, Codex) to work on tasks via the Agent Client Protocol. Each agent runs as its own subprocess speaking ACP over stdio and streams results back into the conversation.
 
 ## Usage
 
 Use `acp_spawn` to delegate a coding task to an external agent. The agent runs as a subprocess speaking the ACP protocol over stdio and streams results back.
 
-Users can refer to agents by natural names: "claude code", "codex cli", "openai codex", "gemini cli", and "google gemini" all resolve to the canonical `claude`, `codex`, and `gemini` ids (unless the user's config defines an agent literally keyed by that name, which always wins).
+Users can refer to agents by natural names: "claude code", "codex cli", and "openai codex" all resolve to the canonical `claude` and `codex` ids (unless the user's config defines an agent literally keyed by that name, which always wins).
 
 ## First-time setup
 
-ACP is always available - default profiles for `claude`, `codex`, and `gemini` ship out-of-box, so no config edit is needed to start. First-time setup is just making the adapter binary available, then spawning:
+ACP is always available - default profiles for `claude` and `codex` ship out-of-box, so no config edit is needed to start. First-time setup is just making the adapter binary available, then spawning:
 
 1. Install the adapter binary if it's missing. This happens automatically: when `acp_spawn` finds the agent's binary missing from PATH, the assistant installs it once via a sandboxed bun global install and proceeds in the same call (see "Automatic adapter availability" below).
 
@@ -35,16 +35,15 @@ ACP is always available - default profiles for `claude`, `codex`, and `gemini` s
 
 ## Automatic adapter availability
 
-When `acp_spawn` finds the agent's binary missing from PATH, the assistant installs it once via a sandboxed bun global install and then runs the real installed binary. The install runs in a fresh empty temporary directory (never the task's project directory), with the Claude/Gemini secrets stripped from the installer environment and the registry pinned to the public npm registry, so a malicious project directory cannot hijack package resolution or capture a token. After this one-time install, the adapter is a normal trusted binary on PATH and every later spawn (and resume) uses it directly.
+When `acp_spawn` finds the agent's binary missing from PATH, the assistant installs it once via a sandboxed bun global install and then runs the real installed binary. The install runs in a fresh empty temporary directory (never the task's project directory), with known secrets stripped from the installer environment and the registry pinned to the public npm registry, so a malicious project directory cannot hijack package resolution or capture a token. After this one-time install, the adapter is a normal trusted binary on PATH and every later spawn (and resume) uses it directly.
 
-Only the allowlisted out-of-box packages are ever installed this way (`@agentclientprotocol/claude-agent-acp`, `@zed-industries/codex-acp`, `@google/gemini-cli`); user-configured agents with custom commands are never installed automatically.
+Only the allowlisted out-of-box packages are ever installed this way (`@agentclientprotocol/claude-agent-acp`, `@zed-industries/codex-acp`); user-configured agents with custom commands are never installed automatically.
 
 Manual installation is fallback guidance for unusual setups: bun unavailable, restricted global installs, or an auto-install failure (the failure reason is surfaced in the tool result).
 
 ```bash
 bun add -g @agentclientprotocol/claude-agent-acp   # claude
 bun add -g @zed-industries/codex-acp               # codex
-bun add -g @google/gemini-cli                       # gemini
 ```
 
 ## Claude setup
@@ -68,43 +67,14 @@ The `codex-acp` adapter is installed automatically when missing, but it shells o
    - `CODEX_API_KEY` environment variable
    - `OPENAI_API_KEY` environment variable
 
-## Gemini setup
-
-Gemini CLI speaks ACP natively (`gemini --acp`) - there is no separate adapter binary. The CLI itself is installed from `@google/gemini-cli` when missing.
-
-**Authenticate** with an API key through the credential store (the primary path):
-
-```bash
-assistant credentials set --service acp --field gemini_api_key <key>
-```
-
-Or collect the key via the secure prompt: `assistant credentials prompt --service acp --field gemini_api_key --label "Gemini API Key"`. Either way the key never appears in chat or workspace config, and every spawn injects it as `GEMINI_API_KEY` automatically. The key is optional - a spawn proceeds without it when the vault has no entry.
-
-The alternative is browser OAuth: run `gemini` once interactively and complete the sign-in flow. This is impractical on hosted assistants (no browser), so prefer the credential store there.
-
 Do NOT put API keys (or any secret) in the workspace config file - secrets never belong in the workspace directory. Use the credential store instead.
 
-A workspace `acp.agents.gemini` override is only for non-secret customization (custom binary path, extra args, non-secret env vars). It must spell out the full `command` and `args` - see "Critical: correct agent command" below for the replace-not-merge rule:
-  ```json
-  {
-    "acp": {
-      "agents": {
-        "gemini": {
-          "command": "gemini",
-          "args": ["--acp"],
-          "env": { "NO_COLOR": "1" }
-        }
-      }
-    }
-  }
-  ```
-
 ## Critical: correct agent command
 
-- Three agents are supported out-of-box: `claude` (via the `claude-agent-acp` adapter), `codex` (via the `codex-acp` adapter), and `gemini` (via `gemini --acp` - Gemini speaks ACP natively, no adapter binary).
-- NEVER use `claude`, `claude -p`, `claude --acp`, or the bare `codex` CLI as the ACP `command`. Claude and Codex only speak the protocol through their dedicated `*-acp` adapters. Gemini is the exception: the `gemini` CLI itself speaks ACP when launched with `--acp`.
-- Default profiles for all three ship out-of-box. Users only need an `agents.<id>` entry in config if they want to override the defaults (e.g. point to a custom binary path or pass extra args/env). An `acp.agents.<id>` entry replaces the bundled default entirely (no field merge), so any override must spell out the full `command` and `args`, not just the field being changed.
-- NEVER change an existing ACP config to use a different command. If the config already has `claude-agent-acp`, `codex-acp`, or `gemini`, leave it alone.
+- Two agents are supported out-of-box: `claude` (via the `claude-agent-acp` adapter) and `codex` (via the `codex-acp` adapter).
+- NEVER use `claude`, `claude -p`, `claude --acp`, or the bare `codex` CLI as the ACP `command`. Claude and Codex only speak the protocol through their dedicated `*-acp` adapters.
+- Default profiles for both ship out-of-box. Users only need an `agents.<id>` entry in config if they want to override the defaults (e.g. point to a custom binary path or pass extra args/env). An `acp.agents.<id>` entry replaces the bundled default entirely (no field merge), so any override must spell out the full `command` and `args`, not just the field being changed.
+- NEVER change an existing ACP config to use a different command. If the config already has `claude-agent-acp` or `codex-acp`, leave it alone.
 
 ## Updating an adapter
 
@@ -114,8 +84,6 @@ Adapters are installed once via a bun global install. To update one to the lates
 bun add -g @agentclientprotocol/claude-agent-acp@latest
 # or
 bun add -g @zed-industries/codex-acp@latest
-# or
-bun add -g @google/gemini-cli@latest
 ```
 
 Then retry the `acp_spawn` call.
@@ -128,7 +96,7 @@ Then retry the `acp_spawn` call.
 
 ## Discoverability
 
-Use `acp_list_agents` to see what's set up and what's missing. It returns each available agent profile, whether the agent's binary is on PATH (missing binaries are installed automatically on first spawn), and an install hint if not. This is the right tool to call when deciding between `claude`, `codex`, and `gemini`, or when the user asks "what coding agents do I have?"
+Use `acp_list_agents` to see what's set up and what's missing. It returns each available agent profile, whether the agent's binary is on PATH (missing binaries are installed automatically on first spawn), and an install hint if not. This is the right tool to call when deciding between `claude` and `codex`, or when the user asks "what coding agents do I have?"
 
 ## Working directory
 

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

@@ -3,7 +3,7 @@
   "tools": [
     {
       "name": "acp_spawn",
-      "description": "Spawn an external coding agent (e.g. Claude Code, Codex, Gemini) via ACP to work on a task. Default profiles ship for `claude` (`claude-agent-acp`), `codex` (`codex-acp`), and `gemini` (`gemini --acp`); the assistant resolves the agent id to the right binary. The agent runs as a subprocess and streams results back. Use this when you want to delegate a coding task to an external agent that has its own tools, file editing, and terminal access. If a default agent's binary is missing, the assistant installs it once via a sandboxed bun global install and proceeds in the same call; if that fails (e.g. bun unavailable), an actionable install hint is returned - do NOT alter `agents.<id>.command` to swap binaries.",
+      "description": "Spawn an external coding agent (e.g. Claude Code, Codex) via ACP to work on a task. Default profiles ship for `claude` (`claude-agent-acp`) and `codex` (`codex-acp`); the assistant resolves the agent id to the right binary. The agent runs as a subprocess and streams results back. Use this when you want to delegate a coding task to an external agent that has its own tools, file editing, and terminal access. If a default agent's binary is missing, the assistant installs it once via a sandboxed bun global install and proceeds in the same call; if that fails (e.g. bun unavailable), an actionable install hint is returned - do NOT alter `agents.<id>.command` to swap binaries.",
       "category": "orchestration",
       "risk": "high",
       "input_schema": {
@@ -11,7 +11,7 @@
         "properties": {
           "agent": {
             "type": "string",
-            "description": "Which agent to spawn (e.g. 'claude', 'codex', 'gemini'). Natural names like 'claude code' or 'gemini cli' also resolve. Defaults to 'claude'."
+            "description": "Which agent to spawn (e.g. 'claude', 'codex'). Natural names like 'claude code' or 'openai codex' also resolve. Defaults to 'claude'."
           },
           "task": {
             "type": "string",
@@ -87,7 +87,7 @@
     },
     {
       "name": "acp_list_agents",
-      "description": "Lists ACP coding agents available to spawn. Each entry includes whether the agent's binary is on PATH (missing binaries are installed automatically on first spawn), and an install command if not. Use this to decide between 'claude', 'codex', and 'gemini' or to surface setup steps to the user.",
+      "description": "Lists ACP coding agents available to spawn. Each entry includes whether the agent's binary is on PATH (missing binaries are installed automatically on first spawn), and an install command if not. Use this to decide between 'claude' and 'codex' or to surface setup steps to the user.",
       "category": "orchestration",
       "risk": "low",
       "input_schema": {

package/src/config/bundled-skills/app-builder/SKILL.md

@@ -72,9 +72,9 @@ All new apps use `formatVersion: 2` (multi-file TSX). No root-level `index.html`
 
 ## Responsive & design system
 
-Every app works phone (~360px) to desktop (~1400px+). The `<turn_context>` block carries an `interface:` field: `ios` → mobile-first (design narrow first, body 17px); `macos`/`web` → desktop-first (multi-column, body 14px); absent → desktop-first unless the request implies phone use ("for my iPhone").
+Every app works phone (~360px) to desktop (~1400px+). The `<turn_context>` block carries a `client_os:` field (the OS the user is on): `ios`/`android` → mobile-first (design narrow first, body 17px); `macos`/`web` → desktop-first (multi-column, body 14px); absent → desktop-first unless the request implies phone use ("for my iPhone"). (The sibling `interface:` field is the transport surface — always `web` for the apps — so key layout off `client_os`, not `interface`.)
 
-**Universal baseline — every build, regardless of interface:**
+**Universal baseline — every build, regardless of client_os:**
 - Viewport meta: `width=device-width, initial-scale=1, viewport-fit=cover`. Never `user-scalable=no` (blocks accessibility zoom).
 - Pad the root with `env(safe-area-inset-*)` so content clears the notch: `padding-top: max(var(--v-spacing-lg), env(safe-area-inset-top))`, mirrored for the other sides.
 - Full-height containers use `100dvh`, not `100vh`.
@@ -82,7 +82,7 @@ Every app works phone (~360px) to desktop (~1400px+). The `<turn_context>` block
 - Interactive elements ≥44×44pt (`.v-button` already complies; custom controls set `min-height: 44px`). Gate hover behind `@media (hover: hover)`.
 - Fluid widths only — `%`, `fr`, `minmax`, `clamp()`, never fixed `px` on containers. Size chart containers in `vw`/`%`. At narrow widths, collapse tables into stacked label-value cards.
 
-**Mobile-first extras (`interface: ios`):** body `--v-font-size-lg` (17px); one column by default, multi-column only above `@media (min-width: 720px)`; bottom-anchor the primary action (`position: sticky; bottom: env(safe-area-inset-bottom)`); bottom sheets instead of side modals.
+**Mobile-first extras (`client_os: ios` / `android`):** body `--v-font-size-lg` (17px); one column by default, multi-column only above `@media (min-width: 720px)`; bottom-anchor the primary action (`position: sticky; bottom: env(safe-area-inset-bottom)`); bottom sheets instead of side modals.
 
 Full detail when reachable: `{baseDir}/references/RESPONSIVE.md`.
 

package/src/config/bundled-skills/app-builder/references/RESPONSIVE.md

@@ -4,20 +4,20 @@ Every app must work across phone (~360px) to desktop (~1400px+).
 
 ## Mode selection
 
-The conversation context's `<turn_context>` block carries an `interface:` field.
+The conversation context's `<turn_context>` block carries a `client_os:` field — the OS the user is on. (Key layout off this, not the sibling `interface:` field: `interface:` is the transport surface and is always `web` for the web/iOS/macOS apps.)
 
-**If `interface: ios`** (or any future mobile-web / android identifier):
-  → Mobile-first build. Design the narrow viewport first, enhance upward.
+**If `client_os: ios` or `android`** (any mobile-web surface):
+→ Mobile-first build. Design the narrow viewport first, enhance upward.
 
-**If `interface: macos` or `web`**:
-  → Desktop-first build. Design larger composition first; narrow fallback still meets the universal baseline below.
+**If `client_os: macos` or `web`**:
+→ Desktop-first build. Design larger composition first; narrow fallback still meets the universal baseline below.
 
 **If field is absent or ambiguous**:
-  → Default to desktop-first unless the request implies phone use ("for my iPhone", "a tap-tracker I'll use on the go").
+→ Default to desktop-first unless the request implies phone use ("for my iPhone", "a tap-tracker I'll use on the go").
 
 ---
 
-## Universal baseline (every build, regardless of interface)
+## Universal baseline (every build, regardless of client_os)
 
 ### Viewport & safe areas
 
@@ -44,7 +44,7 @@ The conversation context's `<turn_context>` block carries an `interface:` field.
 
 ---
 
-## Mobile-first priorities (`interface: ios`)
+## Mobile-first priorities (`client_os: ios` / `android`)
 
 - Default body text to `--v-font-size-lg` (17px), not `--v-font-size-base` (14px).
 - Bump default vertical rhythm one step (e.g. `--v-spacing-md` → `--v-spacing-lg`).
@@ -52,6 +52,6 @@ The conversation context's `<turn_context>` block carries an `interface:` field.
 - Bottom-anchor the primary action: `position: sticky; bottom: env(safe-area-inset-bottom)`.
 - Replace side modals and popovers with bottom sheets.
 
-## Desktop-first priorities (`interface: macos` / `web`)
+## Desktop-first priorities (`client_os: macos` / `web`)
 
 Multi-column composition, hover-rich affordances, denser information, side modals, inline primary actions. The universal baseline above is still the floor — narrow view must still work.

package/src/config/bundled-skills/contacts/tools/contact-merge.ts

@@ -88,7 +88,7 @@ export async function executeContactMerge(
       ``,
       `Surviving contact (${merged.id}):`,
       `  Name: ${displayName}`,
-      `  Interactions: ${merged.interactionCount}`,
+      `  Interactions: ${merged.interactionCount ?? 0}`,
       merged.notes ? `  Notes: ${merged.notes}` : null,
       merged.channels.length > 0 ? `  Channels:\n${channelList}` : null,
       ``,

package/src/config/bundled-skills/contacts/tools/contact-search.ts

@@ -21,7 +21,7 @@ function formatContactSummary(c: SearchContact): string {
     c.role === "guardian" ? resolveGuardianName(c.displayName) : c.displayName;
   const parts = [`- **${displayName}** (ID: ${c.id})`];
   if (c.notes) parts.push(`  Notes: ${c.notes}`);
-  if (c.interactionCount > 0)
+  if ((c.interactionCount ?? 0) > 0)
     parts.push(`  Interactions: ${c.interactionCount}`);
   if (c.channels.length > 0) {
     const channelList = c.channels

package/src/config/bundled-skills/phone-calls/SKILL.md

@@ -26,7 +26,7 @@ When speaking on behalf of your user during calls, refer to yourself as an "assi
 
 # Overview
 
-The calling system uses Twilio's ConversationRelay for both **outbound** and **inbound** voice calls. The text-to-speech voice is provided by the globally configured TTS provider (set via `services.tts.provider`, default: ElevenLabs). After Twilio setup, the assistant prompts the user to choose a voice from a curated list of supported options.
+The calling system connects both **outbound** and **inbound** voice calls over Twilio Media Streams: the assistant transcribes and synthesizes call audio itself using the configured STT provider (`services.stt.provider`) and TTS provider (`services.tts.provider`, default: ElevenLabs), so working API keys for both are required before calls can connect. After Twilio setup, the assistant prompts the user to choose a voice from a curated list of supported options.
 
 # Initial Setup
 

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

@@ -14,22 +14,22 @@ All call-related settings can be managed via `assistant config`:
 | `calls.callerIdentity.allowPerCallOverride` | Allow per-call caller identity selection                                                                                                                                                                                                  | `true`                                                                                                   |
 | `calls.callerIdentity.userNumber`           | E.164 phone number for user-number mode                                                                                                                                                                                                   | _(empty)_                                                                                                |
 | `calls.voice.language`                      | Language code for TTS and transcription                                                                                                                                                                                                   | `en-US`                                                                                                  |
-| `services.stt.provider`                     | STT provider for transcription and telephony. Determines the Twilio integration path at call setup time (Deepgram/Google use native ConversationRelay; OpenAI Whisper and xAI use media-stream).                                          | `deepgram`                                                                                               |
+| `services.stt.provider`                     | STT provider for transcription and telephony. The assistant transcribes call audio itself over the Twilio media stream (streaming when the provider supports it, batch otherwise), so calls require a working API key for this provider.  | `deepgram`                                                                                               |
 | `services.tts.provider`                     | Active TTS provider for speech synthesis. Must be a provider ID from the catalog (`elevenlabs`, `fish-audio`, `deepgram`, `xai`). New providers can be added via the catalog without code changes to call routing.                        | `elevenlabs`                                                                                             |
 | `services.tts.providers.<id>.*`             | Provider-specific settings block. Each catalog provider has its own settings namespace under `services.tts.providers.<id>`. See voice settings in the desktop/iOS app or run `assistant config list` for available settings per provider. | _(per-provider defaults)_                                                                                |
 
 ## TTS provider call-path behavior
 
-Each TTS provider uses one of two call-path modes during phone calls:
+During phone calls the assistant synthesizes audio server-side via the configured provider's HTTP API, transcodes it to mu-law, and streams it to Twilio over the media stream. Each provider's catalog entry declares its playback format:
 
-| Provider     | Call mode          | Description                                                                                                                                                              |
-| ------------ | ------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| `elevenlabs` | `native-twilio`    | Text tokens are forwarded to Twilio's ConversationRelay, which synthesizes audio via its built-in ElevenLabs integration.                                                |
-| `fish-audio` | `synthesized-play` | The assistant synthesizes audio server-side via Fish Audio's HTTP API and streams chunks to Twilio via play messages.                                                    |
-| `deepgram`   | `synthesized-play` | The assistant synthesizes audio server-side via Deepgram's HTTP API and streams chunks to Twilio via play messages. Uses the same API key as Deepgram speech-to-text.    |
-| `xai`        | `synthesized-play` | The assistant synthesizes audio server-side via xAI's HTTP API and streams chunks to Twilio via play messages. No native Twilio fallback (`allowNativeFallback: false`). |
+| Provider     | Playback format | Description                                                                                                                       |
+| ------------ | --------------- | ---------------------------------------------------------------------------------------------------------------------------------- |
+| `elevenlabs` | `pcm`           | Synthesizes PCM audio via the ElevenLabs API.                                                                                     |
+| `fish-audio` | `wav`           | Synthesizes WAV audio via Fish Audio's HTTP API. Requires a configured `referenceId`.                                             |
+| `deepgram`   | `pcm`           | Synthesizes PCM audio via Deepgram's HTTP API. Uses the same API key as Deepgram speech-to-text.                                  |
+| `xai`        | `pcm`           | Synthesizes PCM audio via xAI's HTTP API.                                                                                         |
 
-Providers using `synthesized-play` add a small amount of latency compared to `native-twilio` because audio must be synthesized server-side before playback. The assistant handles chunk buffering and streaming automatically.
+Calls require a media-stream-playable TTS provider with a working API key. When the configured provider is not playable (unsupported format or missing credentials), the assistant falls back to a credentialed playable provider rather than letting the call go silent; if none exists, calls are blocked up front with a setup-required message.
 
 ## Adjusting settings
 

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

@@ -27,7 +27,7 @@ If it reports an available platform assistant, do not install or start ngrok for
 
 ## Call connects but no audio / one-way audio
 
-- The ConversationRelay WebSocket may not be connecting. If you are using ngrok or a custom tunnel, check that `ingress.publicBaseUrl` is correct and the tunnel is active. If you are using Velay, check `assistant platform status --json`; a 503 from `https://velay.../<assistant-id>/...` usually means the assistant tunnel is not connected or the gateway did not complete the WebSocket open, not that ngrok is required.
+- The media-stream WebSocket may not be connecting. If you are using ngrok or a custom tunnel, check that `ingress.publicBaseUrl` is correct and the tunnel is active. If you are using Velay, check `assistant platform status --json`; a 503 from `https://velay.../<assistant-id>/...` usually means the assistant tunnel is not connected or the gateway did not complete the WebSocket open, not that ngrok is required.
 - Verify the assistant is running
 
 ## "Number not eligible for caller identity"
@@ -68,8 +68,8 @@ Do not rotate ngrok to work around a managed Velay WebSocket failure. Fix the Ve
 ## Local Twilio Velay smoke tests
 
 - HTTP bridge: request `${VELAY_PUBLIC_BASE_URL}/<assistant-id>/healthz` and `${VELAY_PUBLIC_BASE_URL}/<assistant-id>/schema`. When testing a JSON webhook route under active development, POST a small JSON body through the same Velay public URL and confirm the gateway receives it.
-- Synthetic WebSocket: connect a local WebSocket client to `${VELAY_PUBLIC_BASE_URL}/<assistant-id>/webhooks/twilio/relay?callSessionId=session-123&token=<edge-token>` and confirm the upgrade reaches the gateway.
-- Real Twilio call: wait for the gateway to register with Velay, then place a call and confirm Twilio fetches `/webhooks/twilio/voice` and opens the relay or media-stream WebSocket through the Velay URL.
+- Synthetic WebSocket: connect a local WebSocket client to `${VELAY_PUBLIC_BASE_URL}/<assistant-id>/webhooks/twilio/media-stream/session-123/<edge-token>` and confirm the upgrade reaches the gateway.
+- Real Twilio call: wait for the gateway to register with Velay, then place a call and confirm Twilio fetches `/webhooks/twilio/voice` and opens the media-stream WebSocket through the Velay URL.
 
 ## Call drops after 30 seconds of silence
 
@@ -82,7 +82,10 @@ The system has a 30-second silence timeout. If nobody speaks for 30 seconds duri
 
 ## Twilio says "application error" right after answer
 
-- This often means ConversationRelay rejected voice configuration after TwiML fetch
-- Keep `services.tts.providers.elevenlabs.voiceModelId` empty first (bare `voiceId` mode)
-- If you set `voiceModelId`, try clearing it and retesting:
-  `assistant config set services.tts.providers.elevenlabs.voiceModelId ""`
+- This often means the voice webhook failed or returned invalid TwiML, or the media-stream WebSocket could not connect
+- Check the assistant logs for voice-webhook errors and confirm `ingress.publicBaseUrl` (or the Velay tunnel) is reachable from Twilio
+
+## Call answers with a spoken "setup required" message and hangs up
+
+- The credential preflight found a missing STT or TTS provider credential — the spoken message names what is missing
+- Store the required API key for the configured `services.stt.provider` / `services.tts.provider` and retry

package/src/config/bundled-skills/skill-management/TOOLS.json

@@ -42,6 +42,16 @@
             "items": { "type": "string" },
             "description": "Optional list of child skill IDs that this skill includes (metadata only, no auto-activation)."
           },
+          "activation_hints": {
+            "type": "array",
+            "items": { "type": "string" },
+            "description": "Optional trigger phrases describing the situations where this skill should activate, phrased as the observed intent (e.g. \"user asks to deploy staging\"). Surfaced in memory as a \"Use when: …\" clause so the skill is retrievable by intent, not just by name."
+          },
+          "avoid_when": {
+            "type": "array",
+            "items": { "type": "string" },
+            "description": "Optional situations where this skill should NOT be used. Surfaced in memory as an \"Avoid when: …\" clause to steer retrieval away from the wrong contexts."
+          },
           "files": {
             "type": "array",
             "items": {

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

@@ -66,8 +66,6 @@ export const CALL_SITE_DEFAULTS: Record<LLMCallSite, CallSiteDefaultConfig> = {
   approvalConversation: { profile: "cost-optimized" },
   trustRuleSuggestion: { profile: "cost-optimized" },
   styleAnalyzer: { profile: "cost-optimized" },
-  meetConsentMonitor: { profile: "cost-optimized" },
-  meetChatOpportunity: { profile: "cost-optimized" },
   inference: { profile: "cost-optimized" },
   // Vision captioning for the image-fallback plugin. No pinned profile — the
   // plugin resolves a vision-capable profile itself via `doesSupportVision` and

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

@@ -218,14 +218,6 @@
       "description": "Replace the knowledge graph top-level tab with a new Home page showing relationship progression, facts, and capability tiers",
       "defaultEnabled": false
     },
-    {
-      "id": "meet",
-      "scope": "assistant",
-      "key": "meet",
-      "label": "Google Meet",
-      "description": "Enables the Google Meet joining bot and the meet-join skill.",
-      "defaultEnabled": false
-    },
     {
       "id": "scroll-debug-overlay",
       "scope": "client",
@@ -362,6 +354,14 @@
       "description": "Expose the developer-only Memory Router Playground tab in macOS Settings and the /assistant/memory-router-playground web page for dry-running v4 router config overrides against the live page index. Dev-only; default off.",
       "defaultEnabled": false
     },
+    {
+      "id": "memory-v3-injection-gate",
+      "scope": "assistant",
+      "key": "memory-v3-injection-gate",
+      "label": "Memory v3 injection gate",
+      "description": "Per-turn injection gate for memory v3: after the BM25F needle + dense cosine finder lanes run, skip the selectPool LLM call (inject nothing) when finder-lane scores fall below the configured thresholds (memory.v3.gate). Default on (flipped remotely July 2026); this registry value is the fallback when no platform snapshot is available.",
+      "defaultEnabled": true
+    },
     {
       "id": "self-intro-greeting",
       "scope": "both",
@@ -378,6 +378,14 @@
       "description": "Spike: replace pre-chat onboarding with a web-research \"here's what I know about you\" flow, including a Google Calendar \"Let's chat tomorrow\" step shown over the streaming research. Gates the /assistant/onboarding/research route and the research mock harness. Off by default; enable locally via the feature-flags panel.",
       "defaultEnabled": false
     },
+    {
+      "id": "personality-onboarding",
+      "scope": "client",
+      "key": "personality-onboarding",
+      "label": "Personality onboarding step (spike)",
+      "description": "Spike: show the \"Create my personality\" trait-slider step in the research-onboarding flow (between the pitch and the integration step). On continue it applies the sliders to the assistant's persona on a throwaway side conversation. When off, the step is skipped entirely. Off by default; enable locally via the feature-flags panel.",
+      "defaultEnabled": false
+    },
     {
       "id": "channel-trust-floors",
       "scope": "assistant",

package/src/config/loader.ts

@@ -779,6 +779,17 @@ export function loadConfig(): AssistantConfig {
         // Drop the deployment-context embedding provider so it is never
         // persisted (see above); the schema default re-applies in memory.
         delete (seed.memory.embeddings as { provider?: unknown }).provider;
+        // Memory-v3 tuning knobs are globally-shipped defaults, not per-assistant
+        // config: persist only `live` (genuine per-assistant state — some
+        // workspaces predate the v3 migration and must not be flipped on) and let
+        // every tuning knob resolve from the schema on load. This way a shipped
+        // schema-default change reaches all assistants (mirrors the
+        // embedding-provider strip above); migration
+        // 119-strip-persisted-memory-v3-tuning-defaults handles already-seeded
+        // configs.
+        seed.memory.v3 = {
+          live: seed.memory.v3.live,
+        } as (typeof seed.memory)["v3"];
         // Strip dataDir (runtime-derived) from the persisted config
         const { dataDir: _, ...persistable } = seed;
         writeFileSync(configPath, JSON.stringify(persistable, null, 2) + "\n");

package/src/config/schema.ts

@@ -43,12 +43,14 @@ import {
 import { McpConfigSchema } from "./schemas/mcp.js";
 import { MemoryConfigSchema } from "./schemas/memory.js";
 import { MigrationsConfigSchema } from "./schemas/migrations.js";
+import { MonitoringConfigSchema } from "./schemas/monitoring.js";
 import { NotificationsConfigSchema } from "./schemas/notifications.js";
 import {
   DaemonConfigSchema,
   PlatformConfigSchema,
   UiConfigSchema,
 } from "./schemas/platform.js";
+import { SchedulesConfigSchema } from "./schemas/schedules.js";
 import { SecretDetectionConfigSchema } from "./schemas/security.js";
 import { ServicesSchema } from "./schemas/services.js";
 import { SkillsConfigSchema } from "./schemas/skills.js";
@@ -64,7 +66,13 @@ export const AssistantConfigSchema = z
   .object({
     services: ServicesSchema.default(ServicesSchema.parse({})),
     memory: MemoryConfigSchema.default(MemoryConfigSchema.parse({})),
-    migrations: MigrationsConfigSchema.default(MigrationsConfigSchema.parse({})),
+    monitoring: MonitoringConfigSchema.default(
+      MonitoringConfigSchema.parse({}),
+    ),
+    schedules: SchedulesConfigSchema.default(SchedulesConfigSchema.parse({})),
+    migrations: MigrationsConfigSchema.default(
+      MigrationsConfigSchema.parse({}),
+    ),
     dataDir: z
       .string({ error: "dataDir must be a string" })
       .default(getDataDir())

package/src/config/schemas/__tests__/memory-v2.test.ts

@@ -22,7 +22,7 @@ describe("MemoryV2ConfigSchema", () => {
       sparse_weight: 0.15,
       bm25_k1: 1.2,
       bm25_b: 0.4,
-      consolidation_interval_hours: 4,
+      consolidation_interval_hours: 8,
       consolidation_max_buffer_lines: 100,
       consolidation_max_entries_per_run: 150,
       max_page_chars: 5000,
@@ -222,7 +222,7 @@ describe("MemoryConfigSchema integration with v2 block", () => {
     expect(parsed.v2.d).toBe(0.3);
     expect(parsed.v2.dense_weight).toBe(0.85);
     expect(parsed.v2.sparse_weight).toBe(0.15);
-    expect(parsed.v2.consolidation_interval_hours).toBe(4);
+    expect(parsed.v2.consolidation_interval_hours).toBe(8);
     expect(parsed.v2.max_page_chars).toBe(5000);
   });
 

package/src/config/schemas/__tests__/memory-v3.test.ts

@@ -26,6 +26,17 @@ describe("MemoryV3ConfigSchema", () => {
       selectorPromptPath: null,
       edge: { hubDegree: 30, seedCount: 18, perSeed: 6, cap: 45 },
       entity: { enabled: true, idfFloor: 4, cap: 8 },
+      gate: {
+        enabled: true,
+        denseThreshold: 0.66,
+        sparseThreshold: 0.35,
+        sparseOnlyThreshold: 0.75,
+        denseClusterThreshold: 0.6,
+        denseClusterMaxDelta: 0.02,
+        topK: 5,
+        bm25NormK: null,
+        bypassForCore: false,
+      },
     });
   });
 
@@ -121,4 +132,57 @@ describe("MemoryV3ConfigSchema", () => {
     ).toThrow();
     expect(() => MemoryV3ConfigSchema.parse({ edge: { cap: -1 } })).toThrow();
   });
+
+  test("accepts a partial gate override, defaulting the rest", () => {
+    const parsed = MemoryV3ConfigSchema.parse({
+      gate: { denseThreshold: 0.6 },
+    });
+    expect(parsed.gate).toEqual({
+      enabled: true,
+      denseThreshold: 0.6,
+      sparseThreshold: 0.35,
+      sparseOnlyThreshold: 0.75,
+      denseClusterThreshold: 0.6,
+      denseClusterMaxDelta: 0.02,
+      topK: 5,
+      bm25NormK: null,
+      bypassForCore: false,
+    });
+  });
+
+  test("gate.enabled defaults true, accepts false, rejects non-booleans", () => {
+    expect(MemoryV3ConfigSchema.parse({}).gate.enabled).toBe(true);
+    expect(
+      MemoryV3ConfigSchema.parse({ gate: { enabled: false } }).gate.enabled,
+    ).toBe(false);
+    expect(() =>
+      MemoryV3ConfigSchema.parse({ gate: { enabled: "yes" } }),
+    ).toThrow();
+  });
+
+  test("gate.topK rejects zero and non-integers", () => {
+    expect(() => MemoryV3ConfigSchema.parse({ gate: { topK: 0 } })).toThrow();
+    expect(() => MemoryV3ConfigSchema.parse({ gate: { topK: 2.5 } })).toThrow();
+  });
+
+  test("gate thresholds reject out-of-range values", () => {
+    expect(() =>
+      MemoryV3ConfigSchema.parse({ gate: { denseThreshold: 1.5 } }),
+    ).toThrow();
+  });
+
+  test("gate.bm25NormK accepts null and a positive number, rejects non-positive", () => {
+    expect(
+      MemoryV3ConfigSchema.parse({ gate: { bm25NormK: null } }).gate.bm25NormK,
+    ).toBe(null);
+    expect(
+      MemoryV3ConfigSchema.parse({ gate: { bm25NormK: 1.2 } }).gate.bm25NormK,
+    ).toBe(1.2);
+    expect(() =>
+      MemoryV3ConfigSchema.parse({ gate: { bm25NormK: 0 } }),
+    ).toThrow();
+    expect(() =>
+      MemoryV3ConfigSchema.parse({ gate: { bm25NormK: -1 } }),
+    ).toThrow();
+  });
 });

package/src/config/schemas/call-site-catalog.ts

@@ -282,18 +282,6 @@ const CATALOG_RECORD: CatalogRecord = {
     description: "Infers the category of a skill from its description.",
     domain: "skills",
   },
-  meetConsentMonitor: {
-    id: "meetConsentMonitor",
-    displayName: "Meet Consent Monitor",
-    description: "Monitors meeting consent signals during live calls.",
-    domain: "skills",
-  },
-  meetChatOpportunity: {
-    id: "meetChatOpportunity",
-    displayName: "Meet Chat Opportunity",
-    description: "Identifies opportunities to engage in meeting chat.",
-    domain: "skills",
-  },
   inference: {
     id: "inference",
     displayName: "Inference",