@vellumai/assistant 0.9.0-dev.202606162337.b5a0f93 → 0.9.0-dev.202606170116.ccd16ef

package/ARCHITECTURE.md

@@ -609,10 +609,9 @@ Audio-to-text conversion occurs in six distinct runtime boundaries, each with it
 | ---------------------------- | ----------------------------------------------------------------------------- | -------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ |
 | **Telephony (hybrid)**       | Twilio-native ConversationRelay or daemon media-stream (provider-conditional) | Configured STT provider (via `services.stt`) | `src/calls/telephony-stt-routing.ts`                                                                                                                                                                                                                       | `src/calls/twilio-routes.ts`                                                   |
 | **Daemon batch**             | Daemon process (REST API to provider)                                         | Configured STT provider (via `services.stt`) | `src/stt/daemon-batch-transcriber.ts`                                                                                                                                                                                                                      | `src/runtime/routes/inbound-stages/transcribe-audio.ts`                        |
-| **Conversation streaming**   | Daemon process (WebSocket-based)                                              | Configured STT provider (via `services.stt`) | `src/stt/stt-stream-session.ts`, `src/providers/speech-to-text/deepgram-realtime.ts`, `src/providers/speech-to-text/google-gemini-live-stream.ts`, `src/providers/speech-to-text/openai-whisper-stream.ts`, `src/providers/speech-to-text/xai-realtime.ts` | `VoiceInputManager` (macOS conversation) via gateway WS proxy                  |
-| **Live voice channel**       | Assistant process (gateway-authenticated WebSocket)                           | Configured STT provider (via `services.stt`) | `src/runtime/http-server.ts`, `src/live-voice/live-voice-session-manager.ts`, `src/live-voice/live-voice-session.ts`, `src/providers/speech-to-text/resolve.ts`, streaming provider adapters                                                               | `LiveVoiceChannelManager` (macOS voice mode) via `/v1/live-voice`              |
-| **Client service-first**     | macOS via gateway → daemon                                                    | Configured STT provider (via `services.stt`) | `src/runtime/routes/stt-routes.ts`, `clients/shared/Network/STTClient.swift`                                                                                                                                                                               | `VoiceInputManager` (macOS dictation), `OpenAIVoiceService` (macOS voice mode) |
-| **Client-native (fallback)** | macOS on-device                                                               | Apple Speech (`SFSpeechRecognizer`)          | `clients/macos/.../SpeechRecognizerAdapter.swift`                                                                                                                                                                                                          | Fallback when STT service is unconfigured or fails                             |
+| **Conversation streaming**   | Daemon process (WebSocket-based)                                              | Configured STT provider (via `services.stt`) | `src/stt/stt-stream-session.ts`, `src/providers/speech-to-text/deepgram-realtime.ts`, `src/providers/speech-to-text/google-gemini-live-stream.ts`, `src/providers/speech-to-text/openai-whisper-stream.ts`, `src/providers/speech-to-text/xai-realtime.ts` | Web/Electron dictation client via gateway WS proxy                             |
+| **Live voice channel**       | Assistant process (gateway-authenticated WebSocket)                           | Configured STT provider (via `services.stt`) | `src/runtime/http-server.ts`, `src/live-voice/live-voice-session-manager.ts`, `src/live-voice/live-voice-session.ts`, `src/providers/speech-to-text/resolve.ts`, streaming provider adapters                                                               | Web/Electron live voice client via `/v1/live-voice`                            |
+| **Client service-first**     | Web/Electron via gateway → daemon                                            | Configured STT provider (via `services.stt`) | `src/runtime/routes/stt-routes.ts`                                                                                                                                                                                                                         | Web/Electron dictation and voice clients                                       |
 
 **Telephony boundary (hybrid routing):**
 
@@ -679,16 +678,14 @@ Two provider adapters are supported, each implementing the `StreamingTranscriber
 
 **Session lifecycle (client side):**
 
-- `STTStreamingClient` (`clients/shared/Network/STTStreamingClient.swift`) manages the WebSocket session using `URLSessionWebSocketTask`. It builds the gateway WebSocket URL via `GatewayHTTPClient.buildWebSocketRequest(path: "stt/stream", params:)`.
-- `STTProviderRegistry` (`clients/shared/Utilities/STTProviderRegistry.swift`) exposes `isStreamingAvailable` (checks the configured provider's `conversationStreamingMode` from the `GET /v1/stt/providers` API) and `isServiceConfigured` (checks whether any STT provider is set).
-- macOS: `VoiceInputManager.startStreamingSession()` creates a fresh `STTStreamingClient` per recording session. Streaming partials take priority over `SFSpeechRecognizer` partials while the stream is active and healthy. When recording stops, if the stream delivered at least one `final` event (`streamingReceivedFinal`) and has not failed (`streamingFailed`), the streaming final text is used directly. Otherwise, the batch STT path (`STTClient.transcribe()`) provides the fallback.
+The web client streaming dictation client lives at `clients/web/src/domains/chat/voice/dictation-stream.ts`; it opens the gateway WebSocket to `/v1/stt/stream`, parses `partial`/`final` frames, and reports failures so the caller can fall back to batch transcription. Before opening a session the client checks the configured provider's `conversationStreamingMode` from the `GET /v1/stt/providers` API. When the stream delivers at least one `final` event and has not failed, the streaming final text is used directly; otherwise the batch STT path (`clients/web/src/domains/chat/voice/stt-api.ts`) provides the fallback.
 
 **Fallback semantics:**
 
 The conversation streaming path degrades gracefully to the existing batch STT path:
 
-1. **Unsupported provider** (a hypothetical provider with `conversationStreamingMode: "none"`): The client checks `STTProviderRegistry.isStreamingAvailable` before attempting a streaming session. When `false`, recording proceeds with the batch-only flow (no WebSocket is opened). On the daemon side, if a streaming session is somehow opened for an unsupported provider, the session sends an `error` event followed by `closed` and closes the socket with code 1000.
-2. **Connection failure** (network error, gateway down, auth failure): The `STTStreamingClient` reports an `STTStreamFailure` to the client's `onFailure` callback. macOS sets `streamingFailed = true` and falls through to batch STT resolution when recording stops.
+1. **Unsupported provider** (a hypothetical provider with `conversationStreamingMode: "none"`): The client checks streaming availability (the configured provider's `conversationStreamingMode` from `GET /v1/stt/providers`) before attempting a streaming session. When unavailable, recording proceeds with the batch-only flow (no WebSocket is opened). On the daemon side, if a streaming session is somehow opened for an unsupported provider, the session sends an `error` event followed by `closed` and closes the socket with code 1000.
+2. **Connection failure** (network error, gateway down, auth failure): The streaming client reports the failure to its caller, which marks the stream as failed and falls through to batch STT resolution when recording stops.
 3. **Mid-session provider error** (provider WebSocket disconnect, timeout, rate limit): The daemon session emits an `error` event (with a normalized `SttErrorCategory`) followed by `closed`. The client marks the stream as failed and defers to batch STT.
 4. **Missing credentials**: `resolveStreamingTranscriber()` returns `null` when the API key is not configured. The session sends an `error`+`closed` pair and the client falls back to batch.
 
@@ -713,9 +710,8 @@ The conversation streaming path degrades gracefully to the existing batch STT pa
 | `src/providers/speech-to-text/resolve.ts`                   | `resolveStreamingTranscriber()`: credential-aware factory for streaming adapters; `resolveConversationStreamingSttCapability()`: capability validator |
 | `src/runtime/http-server.ts`                                | Runtime WebSocket upgrade handler for `/v1/stt/stream`, session registry (`activeSttStreamSessions`), graceful shutdown                               |
 | `gateway/src/http/routes/stt-stream-websocket.ts`           | Gateway WebSocket proxy: authenticates client, opens upstream WS to daemon with service token                                                         |
-| `clients/shared/Network/STTStreamingClient.swift`           | Shared Swift WebSocket client: `URLSessionWebSocketTask`-based, event parsing, failure reporting                                                      |
-| `clients/shared/Utilities/STTProviderRegistry.swift`        | Client-side provider catalog: `isStreamingAvailable`, `conversationStreamingMode` per provider                                                        |
-| `clients/macos/.../VoiceInputManager.swift`                 | macOS integration: `startStreamingSession()`, streaming/batch priority, fallback on failure                                                           |
+| `clients/web/src/domains/chat/voice/dictation-stream.ts`       | Web streaming dictation client: WebSocket session, event parsing, failure reporting                                                                   |
+| `clients/web/src/domains/chat/voice/voice-recording-store.ts`  | Web voice recording state: streaming/batch priority, fallback on failure                                                                              |
 
 **Live voice channel boundary:**
 
@@ -742,32 +738,20 @@ V1 is local/gateway-scoped. Managed/cloud WebSocket proxy support, cross-region
 
 **Client service-first boundary:**
 
-All product-facing dictation and voice-streaming paths on macOS use a service-first STT strategy. Clients record audio, encode it to WAV via `AudioWavEncoder` (shared utility in `clients/shared/Utilities/AudioWavEncoder.swift`), and POST it through the gateway to the daemon's `POST /v1/stt/transcribe` endpoint via `STTClient` (`clients/shared/Network/STTClient.swift`). The daemon resolves the configured STT provider through `resolveBatchTranscriber()` and returns the transcribed text.
+All product-facing dictation and voice-streaming paths use a service-first STT strategy. Clients record audio, encode it to WAV, and POST it through the gateway to the daemon's `POST /v1/stt/transcribe` endpoint. The daemon resolves the configured STT provider through `resolveBatchTranscriber()` and returns the transcribed text.
 
-- `STTClient` conforms to `STTClientProtocol` and returns a typed `STTResult` enum (`success`, `notConfigured`, `serviceUnavailable`, `error`). Callers pattern-match on the result to deterministically trigger native fallback.
+- The client receives a typed result distinguishing success from `notConfigured`, `serviceUnavailable`, and `error`, so callers can deterministically trigger their fallback path.
 - The gateway proxies the request via assistant-scoped path rewriting: `/v1/assistants/:id/stt/transcribe` is rewritten to `/v1/stt/transcribe` on the daemon.
 - `stt-routes.ts` (`src/runtime/routes/stt-routes.ts`) defines the HTTP endpoint, validates the audio payload, and delegates to `resolveBatchTranscriber()`.
 
-Product-facing flows using service-first STT:
-
-| Flow                          | Client | Entry point                                                                                                                                                                                         |
-| ----------------------------- | ------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| **Push-to-talk dictation**    | macOS  | `VoiceInputManager.resolveTranscription()` — encodes accumulated PCM buffers to WAV, calls `sttClient.transcribe()`, falls back to native text on failure                                           |
-| **Conversation chat capture** | macOS  | `VoiceInputManager.handleFinalTranscription()` — prefers streaming final when available; falls back to batch `sttClient.transcribe()` when streaming was not used, failed, or produced no finals    |
-| **Voice mode (streaming)**    | macOS  | `OpenAIVoiceService.stopRecordingAndGetTranscription()` — encodes per-turn PCM to WAV, calls `sttClient.transcribe()` for turn-final transcript resolution, falls back to SFSpeechRecognizer result |
-
-**Client-native fallback boundary:**
-
-Apple-native on-device recognition via `SFSpeechRecognizer` serves two roles in all three product-facing flows above: (1) it provides low-latency partial transcriptions for real-time display during recording, and (2) it provides the fallback final transcription when the STT service is unconfigured (HTTP 503), temporarily unavailable (HTTP 5xx), or returns an empty result. The `SpeechRecognizerAdapter` protocols on each platform abstract Apple Speech for **testability and dependency injection**.
-
-The macOS `SpeechRecognizerAdapter` protocol in `clients/macos/vellum-assistant/Features/Voice/SpeechRecognizerAdapter.swift` abstracts `SFSpeechRecognizer` static APIs and instance creation. `AppleSpeechRecognizerAdapter` is the production implementation. `OpenAIVoiceService` and `VoiceInputManager` consume the adapter via dependency injection. **Note:** The protocol leaks Apple Speech types through its surface — `authorizationStatus()` returns `SFSpeechRecognizerAuthorizationStatus` and `makeRecognizer(locale:)` returns `SFSpeechRecognizer?` directly. This means callers depend on the Speech framework at compile time.
+The web client implements these flows in `clients/web/src/domains/chat/voice/`: `stt-api.ts` (batch transcribe), `dictation-stream.ts` (streaming dictation), and `voice-recording-store.ts` (recording state, streaming/batch priority, fallback). Push-to-talk dictation and conversation chat capture prefer a streaming final when available and fall back to batch transcription when streaming was not used, failed, or produced no finals.
 
 **Cross-boundary notes:**
 
 - The `services.stt` config block is the single source of truth for STT provider selection across the daemon batch boundary, the conversation streaming boundary, the client service-first boundary, and the telephony boundary. The batch and streaming resolvers (`resolveBatchTranscriber()`, `resolveStreamingTranscriber()`) both read from `services.stt.provider` and resolve credentials through the same catalog; the telephony boundary uses `resolveTelephonySttRouting()` to determine the Twilio integration strategy. The daemon provider catalog (`src/providers/speech-to-text/provider-catalog.ts`) is the authoritative registry of supported providers. Native clients fetch display metadata via `GET /v1/stt/providers`.
 - Conversation streaming does not replace the client service-first batch path. When streaming is available, it runs concurrently during recording and provides real-time partials and finals. The batch path remains the fallback for providers that do not support streaming, when streaming fails mid-session, or when streaming produces no final transcript.
 - Credential mapping is catalog-driven: `provider-secret-catalog.ts` derives STT API-key provider names from the daemon catalog via `listCredentialProviderNames()`, deduplicating against the LLM/search provider list. Adding a provider to the catalog automatically includes its credential name in `API_KEY_PROVIDERS`.
-- Terminology: "STT" and "transcription" refer to the same operation (converting audio to text). "Speech recognition" is used in client-native contexts where Apple's Speech framework terminology is canonical. All three terms map to the same conceptual operation.
+- Terminology: "STT", "transcription", and "speech recognition" all refer to the same operation (converting audio to text).
 - **Onboarding**: For a step-by-step guide to adding a new STT provider, see `docs/stt-provider-onboarding.md`.
 
 ### On-Demand Home Content Generation

package/docs/architecture/security.md

@@ -227,14 +227,14 @@ sequenceDiagram
     participant Route as credentials prompt route
     participant Prompter as SecretPrompter
     participant HTTP as HTTP Transport
-    participant UI as SecretPromptManager (Swift)
+    participant UI as Secret prompt UI (client)
     participant Store as Credential Store (CES / encrypted file)
 
     Model->>Route: assistant credentials prompt --service --field --label
     Route->>Prompter: requestSecretStandalone(service, field, label, ...)
     Prompter->>HTTP: secret_request {requestId, service, field, label, allowOneTimeSend}
-    HTTP->>UI: Show SecretPromptView (floating panel)
-    UI->>UI: User enters value in SecureField
+    HTTP->>UI: Show secret prompt
+    UI->>UI: User enters value in a masked field
     alt Store (default)
         UI->>HTTP: secret_response {requestId, value, delivery: "store"}
         HTTP->>Prompter: resolve(value, "store")
@@ -298,7 +298,7 @@ The `allowOneTimeSend` config gate (default: `false`) enables a secondary "Send
 | `assistant/src/security/secret-scanner.ts`                  | Prefix + shape-based secret regex detection (used by display-time `redactSecrets`) |
 | `assistant/src/security/secret-ingress.ts`                  | Prefix-only ingress check on user messages                                         |
 | `assistant/src/util/log-redact.ts`                          | Pino log serializers — prefix-based redaction for logs                             |
-| `clients/macos/.../SecretPromptManager.swift`               | Floating panel UI for secure credential entry                                      |
+| `clients/web/src/domains/chat/components/secret-prompt-card.tsx` | UI for secure credential entry                                                    |
 
 ---
 

package/docs/stt-provider-onboarding.md

@@ -72,13 +72,11 @@ Native clients fetch this metadata at launch via `GET /v1/stt/providers`. No sep
 | `google-gemini`  | `gemini`             | shared        |
 | `xai`            | `xai`                | exclusive     |
 
-When the provider ID differs from the credential provider name (e.g. `google-gemini` maps to `gemini`), the key is **shared** with other services that use the same credential. The `sttKeyIsExclusive` / `sttKeyIsShared` helpers in the macOS settings layer derive this automatically from the catalog.
+When the provider ID differs from the credential provider name (e.g. `google-gemini` maps to `gemini`), the key is **shared** with other services that use the same credential.
 
-### macOS settings key behavior
+### Client settings key behavior
 
-**File:** `clients/macos/vellum-assistant/Features/Settings/SettingsStore.swift`
-
-The `sttKeyIsExclusive(for:)` / `sttKeyIsShared(for:)` helpers derive shared-vs-exclusive key behavior from the catalog automatically: if `apiKeyProviderName == id`, the key is exclusive; otherwise it is shared. No new conditionals are needed unless the provider has a non-standard key-ownership model.
+Clients derive shared-vs-exclusive key behavior from the catalog automatically: if `apiKeyProviderName == id`, the key is exclusive; otherwise it is shared. No new conditionals are needed unless the provider has a non-standard key-ownership model. The web settings UI lives in `clients/web/src/domains/settings/ai/speech-to-text-card.tsx`.
 
 ## 7. Verify unified STT architecture
 

package/docs/workflows-testing.md

@@ -2,14 +2,13 @@
 
 This runbook covers the parts of the workflow engine that automated tests do not:
 real provider calls, real journaled resume across a restart, capability
-containment under a live model, flag-off inertness, and persona-leaf voice. Unit
+containment under a live model, and persona-leaf voice. Unit
 and integration coverage lives under `assistant/src/workflows/*.test.ts` and
 `assistant/src/__tests__/`; this is the human-in-the-loop pass that runs **last**,
 on throwaway instances first, and only touches a real inbox or a production
 instance at the very end.
 
-Work through the steps in order. The `workflows` flag is **off by default**, so
-every step that exercises the engine first enables it on a throwaway instance.
+Work through the steps in order, on throwaway instances first.
 
 > **Safety rule:** Nothing in this runbook touches a real inbox, a real workspace,
 > or a production instance until every throwaway-instance step has passed. The
@@ -17,7 +16,7 @@ every step that exercises the engine first enables it on a throwaway instance.
 
 ---
 
-## 1. Hatch a throwaway instance and enable the flag
+## 1. Hatch a throwaway instance
 
 Hatch a disposable Docker instance built from local source:
 
@@ -26,25 +25,14 @@ vellum hatch --remote docker --source .
 ```
 
 Note the instance name it prints (e.g. `vellum-<adjective>-<animal>`). Use it as
-`--assistant <name>` for everything below, or set it active.
-
-Enable the `workflows` flag via that instance's feature-flag override file. The
-override lives in the instance's `protected/feature-flags.json` (overrides here
-win over the registry default — see the feature-flag-overrides gotcha):
-
-```jsonc
-// .../<instance>/protected/feature-flags.json
-{ "workflows": true }
-```
-
-Restart the instance so it re-reads the override, then confirm it is up:
+`--assistant <name>` for everything below, or set it active. Confirm it is up:
 
 ```
 vellum ps
 ```
 
-Sanity-check that the surface is now live: `vellum workflows runs --assistant <name>`
-should return an (empty) table rather than a 404.
+Sanity-check that the workflow surface is live: `vellum workflows runs --assistant <name>`
+should return an (empty) table.
 
 ---
 
@@ -168,27 +156,9 @@ or a send tool, with `capabilities.tools` left empty). Run it and confirm:
 
 ---
 
-## 6. Flag-off inertness
-
-On a **default** instance (no `workflows` override, or set back to `false` and
-restart), confirm the whole surface is gone:
-
-- `run_workflow` and `manage_workflows` are **absent** from the tool set (the
-  assistant cannot call them).
-- The routes 404:
-
-  ```
-  vellum workflows runs --assistant <default-instance>   # request fails (404)
-  ```
-
-- A scheduler `workflow`-mode job is **rejected** (the engine gate throws before
-  any run is launched).
-
----
-
-## 7. Real-data smoke (throwaway TEST account)
+## 6. Real-data smoke (throwaway TEST account)
 
-Only after steps 1–6 pass: run the full path against a **throwaway TEST Google or
+Only after steps 1–5 pass: run the full path against a **throwaway TEST Google or
 Slack account** with about a dozen messages — never a real account. Drive the
 end-to-end flow:
 
@@ -203,17 +173,16 @@ real content, and every side effect was one the manifest declared.
 
 ---
 
-## 8. Production / persona instance — LAST
+## 7. Production / persona instance — LAST
 
-Flip the `workflows` flag on a real (e.g. persona) instance **only after** the
-throwaway instances above pass, and only after the instance is on current code:
+Run on a real (e.g. persona) instance **only after** the throwaway instances
+above pass, and only after the instance is on current code:
 
 1. `git pull` and **restart** the instance so it runs the merged code (a restart
    alone re-runs stale code — see the deploy gotcha).
-2. Enable the flag and restart.
-3. Start with the **smallest real slice** — a few items, dry-run actions where the
+2. Start with the **smallest real slice** — a few items, dry-run actions where the
    tool supports it — before any larger or side-effecting run.
-4. **Human-eval the persona-leaf voice**: read a persona leaf's output and confirm
+3. **Human-eval the persona-leaf voice**: read a persona leaf's output and confirm
    it reads as the assistant, not as a generic worker.
 
 Stop and reassess if any run's `agentsSpawned` approaches `maxAgentsPerRun`, if

package/docs/workflows.md

@@ -8,11 +8,9 @@ from each of a hundred documents, draft-then-verify a batch — and you want the
 results orchestrated deterministically and reported back when the whole run
 finishes.
 
-Workflows are gated behind the `workflows` feature flag (default **off**). The
-`run_workflow` / `manage_workflows` tools are served by the flag-gated `workflows`
+The `run_workflow` / `manage_workflows` tools are served by the `workflows`
 bundled skill rather than as always-on tools — load it with `skill_load` and invoke
-its tools via `skill_execute`. When the flag is off, the skill is absent, the
-management routes 404, and the scheduler rejects `workflow`-mode jobs.
+its tools via `skill_execute`.
 
 - Engine code: `assistant/src/workflows/`
 - Skill (tool surface): `assistant/src/config/bundled-skills/workflows/`
@@ -437,7 +435,7 @@ Reached via the skill (`skill_load` then `skill_execute`), not as always-on tool
   `list_profiles` returns `{ profiles, activeProfile }` — the defined LLM profile
   names plus the workspace active profile, used to pick a valid leaf `profile`.
 
-### Routes (read/abort/resume, all 404 when the flag is off)
+### Routes (read/abort/resume)
 
 | Method | Path                            | Purpose                                                                                              |
 | ------ | ------------------------------- | ---------------------------------------------------------------------------------------------------- |

package/node_modules/@vellumai/environments/src/seeds.ts

@@ -25,11 +25,8 @@ function portBlock(base: number): PortMap {
 }
 
 /**
- * Built-in environment definitions and the TS-side source of truth for the
- * set of known environment names. The Swift client mirrors this list in
- * `clients/macos/vellum-assistant/App/VellumEnvironment.swift`; since Swift
- * can't import TypeScript, drift between the two is caught at test time by
- * `cli/src/__tests__/env-drift.test.ts`.
+ * Built-in environment definitions and the source of truth for the
+ * set of known environment names.
  *
  * Custom environments via a user config file are a future phase — see the
  * "Coexisting environments" design doc. Until then, a call site that needs a

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vellumai/assistant",
-  "version": "0.9.0-dev.202606162337.b5a0f93",
+  "version": "0.9.0-dev.202606170116.ccd16ef",
   "license": "MIT",
   "type": "module",
   "exports": {

package/src/__tests__/assistant-feature-flag-guard.test.ts

@@ -60,10 +60,7 @@ const CANONICAL_KEY_RE = /^[a-z0-9][a-z0-9-]*$/;
  * Keep this list minimal — only files that genuinely need to reference
  * the legacy format for backward compatibility.
  */
-const LEGACY_KEY_ALLOWLIST = new Set([
-  // macOS client: fallback reads from legacy config section
-  "clients/macos/vellum-assistant/Features/Settings/SettingsAccountTab.swift",
-]);
+const LEGACY_KEY_ALLOWLIST = new Set<string>([]);
 
 function isTestFile(filePath: string): boolean {
   return (

package/src/__tests__/assistant-feature-flag-guardrails.test.ts

@@ -58,8 +58,6 @@ function loadRegistry(): Registry {
 const LEGACY_KEY_ALLOWLIST = new Set([
   // Type definitions documenting the legacy format
   "assistant/src/config/types.ts",
-  // macOS client: fallback reads from legacy config section
-  "clients/macos/vellum-assistant/Features/Settings/SettingsAccountTab.swift",
 ]);
 
 function isTestFile(filePath: string): boolean {

package/src/__tests__/dynamic-page-surface.test.ts

@@ -5,7 +5,7 @@ import type {
   UiSurfaceShowDynamicPage,
 } from "../daemon/message-protocol.js";
 import { INTERACTIVE_SURFACE_TYPES } from "../daemon/message-protocol.js";
-import { uiShowTool } from "../tools/ui-surface/definitions.js";
+import { uiShowTool, uiUpdateTool } from "../tools/ui-surface/definitions.js";
 
 // ---------------------------------------------------------------------------
 // DynamicPageSurfaceData shape
@@ -300,6 +300,106 @@ describe("ui_show empty card guard", () => {
   });
 });
 
+describe("ui_update empty payload guard", () => {
+  function makeCtx(onProxy: () => void) {
+    return {
+      conversationId: "conversation-123",
+      workingDir: "/tmp",
+      trustClass: "guardian" as const,
+      proxyToolResolver: async () => {
+        onProxy();
+        return { content: "Surface updated", isError: false };
+      },
+    };
+  }
+
+  test("rejects an empty data object and does not proxy", async () => {
+    let proxied = false;
+    const result = await uiUpdateTool.execute(
+      { surface_id: "surf-1", data: {} },
+      makeCtx(() => {
+        proxied = true;
+      }),
+    );
+
+    expect(result.isError).toBe(true);
+    expect(result.content).toContain("empty `data` payload");
+    expect(proxied).toBe(false);
+  });
+
+  test("rejects missing data and does not proxy", async () => {
+    let proxied = false;
+    const result = await uiUpdateTool.execute(
+      { surface_id: "surf-1" },
+      makeCtx(() => {
+        proxied = true;
+      }),
+    );
+
+    expect(result.isError).toBe(true);
+    expect(proxied).toBe(false);
+  });
+
+  test("rejects data containing only nested empty objects", async () => {
+    let proxied = false;
+    const result = await uiUpdateTool.execute(
+      { surface_id: "surf-1", data: { templateData: {} } },
+      makeCtx(() => {
+        proxied = true;
+      }),
+    );
+
+    expect(result.isError).toBe(true);
+    expect(proxied).toBe(false);
+  });
+
+  test("rejects a task_progress update with an empty steps array", async () => {
+    let proxied = false;
+    const result = await uiUpdateTool.execute(
+      { surface_id: "surf-1", data: { templateData: { steps: [] } } },
+      makeCtx(() => {
+        proxied = true;
+      }),
+    );
+
+    expect(result.isError).toBe(true);
+    expect(proxied).toBe(false);
+  });
+
+  test("allows a task_progress step update", async () => {
+    let proxied = false;
+    const result = await uiUpdateTool.execute(
+      {
+        surface_id: "surf-1",
+        data: {
+          templateData: {
+            steps: [{ label: "Read memo", status: "completed" }],
+          },
+        },
+      },
+      makeCtx(() => {
+        proxied = true;
+      }),
+    );
+
+    expect(result.isError).toBe(false);
+    expect(proxied).toBe(true);
+  });
+
+  test("allows a status-only update without resending steps", async () => {
+    let proxied = false;
+    const result = await uiUpdateTool.execute(
+      { surface_id: "surf-1", data: { templateData: { status: "completed" } } },
+      makeCtx(() => {
+        proxied = true;
+      }),
+    );
+
+    expect(result.isError).toBe(false);
+    expect(proxied).toBe(true);
+  });
+});
+
 // ---------------------------------------------------------------------------
 // task_progress ui_show appends the update hint to its return value
 // ---------------------------------------------------------------------------

package/src/__tests__/gateway-only-guard.test.ts

@@ -23,11 +23,7 @@ const ALLOWLIST = new Set([
 
   // --- Intentional local daemon-control paths ---
   "assistant/src/cli/commands/conversations.ts", // CLI wipe talks to runtime directly
-  "clients/shared/Network/DaemonClient.swift",
-  "clients/shared/App/Auth/PlatformOAuthService.swift", // comment explaining runtimeUrl vs platformUrl
-  "clients/macos/vellum-assistant/App/AppDelegate.swift",
-  "clients/macos/vellum-assistant/Features/Settings/SettingsConnectTab.swift",
-  "apps/macos/src/main/bundle-flow.ts", // Electron main calls the local gateway (gatewayPort) with a Guardian token to scan bundles
+  "clients/macos/src/main/bundle-flow.ts", // Electron main calls the local gateway (gatewayPort) with a Guardian token to scan bundles
   ".claude/skills/update/SKILL.md", // daemon health check script
 
   // --- Test fixtures that poll the daemon directly (gateway may require auth) ---
@@ -87,12 +83,12 @@ function isGatewayInternal(filePath: string): boolean {
 
 /** Additional files allowed for the interpolated-port check only (use gateway port, not runtime). */
 const INTERPOLATED_PORT_ALLOWLIST = new Set([
-  "apps/macos/src/main/bundle-flow.ts",
+  "clients/macos/src/main/bundle-flow.ts",
   // Electron main bridges the host proxy to a local assistant's gateway
   // (gatewayPort) over loopback with a Guardian-minted gateway token — same
   // class as bundle-flow.ts. Added with #34049 (host proxy for cloud/managed
   // assistants); the allowlist entry was missed there.
-  "apps/macos/src/main/host-proxy-router.ts",
+  "clients/macos/src/main/host-proxy-router.ts",
 ]);
 
 /** Shared violation filter: exempt test files, gateway internals, and allowlisted paths. */

package/src/__tests__/llm-callsite-catalog.test.ts

@@ -54,9 +54,7 @@ describe("LLM call-site catalog", () => {
 
   test("CALL_SITE_DEFAULTS covers every LLMCallSite enum value", () => {
     const defaultIds = new Set(Object.keys(CALL_SITE_DEFAULTS));
-    const missing = LLMCallSiteEnum.options.filter(
-      (id) => !defaultIds.has(id),
-    );
+    const missing = LLMCallSiteEnum.options.filter((id) => !defaultIds.has(id));
     expect(missing).toEqual([]);
   });
 
@@ -68,11 +66,12 @@ describe("LLM call-site catalog", () => {
     expect(extra).toEqual([]);
   });
 
-  test("every CALL_SITE_DEFAULTS entry has a profile field", () => {
+  test("every CALL_SITE_DEFAULTS profile, when present, is a non-empty string", () => {
     for (const [, config] of Object.entries(CALL_SITE_DEFAULTS)) {
-      expect(config.profile).toBeDefined();
+      // A call site may omit `profile` to inherit the workspace default config.
+      if (config.profile === undefined) continue;
       expect(typeof config.profile).toBe("string");
-      expect(config.profile!.length).toBeGreaterThan(0);
+      expect(config.profile.length).toBeGreaterThan(0);
     }
   });
 });

package/src/__tests__/llm-resolver.test.ts

@@ -670,30 +670,6 @@ describe("resolveCallSiteConfig", () => {
     expect(resolved.effort).toBe("medium");
   });
 
-  test("workflowLeaf defaults to the cost-optimized profile", () => {
-    const llm = LLMSchema.parse({
-      default: fullDefault,
-      profiles: {
-        balanced: {
-          model: "claude-sonnet-4-7",
-          effort: "medium",
-        },
-        "cost-optimized": {
-          model: "claude-haiku-4-5-20251001",
-          effort: "low",
-        },
-      },
-    });
-
-    expect(resolveDefaultProfileKey("workflowLeaf", llm)).toBe(
-      "cost-optimized",
-    );
-    const resolved = resolveCallSiteConfig("workflowLeaf", llm);
-    expect(resolved.model).toBe("claude-haiku-4-5-20251001");
-    expect(resolved.effort).toBe("low");
-    expect(resolved.thinking.enabled).toBe(false);
-  });
-
   test("explicit callSites config overrides CALL_SITE_DEFAULTS", () => {
     const llm = LLMSchema.parse({
       default: fullDefault,
@@ -1437,3 +1413,52 @@ describe("resolveCallSiteConfig logitBias provenance", () => {
     expect(resolveCallSiteConfig("mainAgent", llm).logitBias).toBeUndefined();
   });
 });
+
+describe("resolveCallSiteConfig — workflowLeaf default", () => {
+  test("inherits the workspace default config rather than pinning cost-optimized", () => {
+    const llm = LLMSchema.parse({
+      default: fullDefault,
+      profiles: {
+        "cost-optimized": {
+          provider: "anthropic",
+          model: "claude-haiku-4-5-20251001",
+        },
+      },
+    });
+    const resolved = resolveCallSiteConfig("workflowLeaf", llm);
+    // No pinned profile → the model comes from llm.default, NOT the
+    // `cost-optimized` profile (which is uncredentialed on a BYOK install).
+    expect(resolved.model).toBe("claude-opus-4-7");
+    // Call-site tuning still applies.
+    expect(resolved.effort).toBe("low");
+    expect(resolved.thinking?.enabled).toBe(false);
+    // The call site has no implicit default profile key.
+    expect(resolveDefaultProfileKey("workflowLeaf", llm)).toBeUndefined();
+  });
+
+  test("honors an explicit workflowLeaf call-site override", () => {
+    const llm = LLMSchema.parse({
+      default: fullDefault,
+      profiles: {
+        cheap: { provider: "anthropic", model: "claude-haiku-4-5-20251001" },
+      },
+      callSites: { workflowLeaf: { profile: "cheap" } },
+    });
+    expect(resolveCallSiteConfig("workflowLeaf", llm).model).toBe(
+      "claude-haiku-4-5-20251001",
+    );
+  });
+
+  test("honors a per-call override profile (an explicit per-leaf profile)", () => {
+    const llm = LLMSchema.parse({
+      default: fullDefault,
+      profiles: {
+        fancy: { provider: "anthropic", model: "claude-sonnet-4-7" },
+      },
+    });
+    expect(
+      resolveCallSiteConfig("workflowLeaf", llm, { overrideProfile: "fancy" })
+        .model,
+    ).toBe("claude-sonnet-4-7");
+  });
+});

package/src/__tests__/oauth-provider-seed-logos.test.ts

@@ -3,17 +3,15 @@ import { describe, expect, test } from "bun:test";
 import { PROVIDER_SEED_DATA } from "../oauth/seed-providers.js";
 
 /**
- * Allowed CDN prefixes for ``logoUrl``. Mirrors the source registry in
- * ``clients/shared/Resources/integration-logos-manifest.json``:
+ * Allowed CDN prefixes for the ``logoUrl`` field on ``PROVIDER_SEED_DATA``
+ * (``assistant/src/oauth/seed-providers.ts``):
  *
  * - Simple Icons (CC0) is the default for most providers.
  * - thesvg via jsDelivr is the documented fallback for brands Simple Icons
  *   doesn't host (e.g. Salesforce, which Simple Icons removed for
- *   trademark reasons). Same source is already used for the bundled PDFs
- *   of figma/github/gmail/linear/notion/outlook/slack.
+ *   trademark reasons).
  *
- * Adding another CDN should be a deliberate choice — extend this list
- * and update the manifest in tandem.
+ * Adding another CDN should be a deliberate choice — extend this list.
  */
 const ALLOWED_LOGO_URL_PREFIXES = [
   "https://cdn.simpleicons.org/",