@vellumai/plugin-api 0.12.2-dev.202609181515.7e76c92 → 0.12.2-dev.202609181615.645c0a5

package/index.d.ts

@@ -4740,11 +4740,13 @@ declare type DistributiveOmit<T, K extends PropertyKey> = T extends unknown ? Om
 /**
  * Whether the given model or profile can process image input.
  *
- * `modelOrProfile` may be a concrete model id, a profile key, or a
- * {@link ModelProfileInfo}. A bare string is resolved as a model id first and,
- * failing that, as a profile key. Returns `false` when nothing resolves.
+ * `modelOrProfile` may be a concrete model id, a profile key, a
+ * {@link ModelProfileInfo}, or a fully resolved target. A resolved target
+ * combines the final provider/model with the winning profile's modality
+ * policy. A bare string is resolved as a model id first and, failing that, as
+ * a profile key. Returns `false` when nothing resolves.
  */
-export declare function doesSupportVision(modelOrProfile: ModelProfileInfo | string): boolean;
+export declare function doesSupportVision(modelOrProfile: ModelProfileInfo | ResolvedVisionTarget | string): boolean;
 
 /**
  * Embed an input and return the raw dense vector (provider, model, and
@@ -5097,6 +5099,27 @@ export declare interface InitContext {
     assistantVersion: string;
 }
 
+declare type InputModalities = z.infer<typeof InputModalitiesSchema>;
+
+declare const InputModalitiesSchema: z.ZodObject<{
+    text: z.ZodOptional<z.ZodObject<{
+        enabled: z.ZodOptional<z.ZodBoolean>;
+        supported: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+    image: z.ZodOptional<z.ZodObject<{
+        enabled: z.ZodOptional<z.ZodBoolean>;
+        supported: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+    audio: z.ZodOptional<z.ZodObject<{
+        enabled: z.ZodOptional<z.ZodBoolean>;
+        supported: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+    video: z.ZodOptional<z.ZodObject<{
+        enabled: z.ZodOptional<z.ZodBoolean>;
+        supported: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>>;
+}, z.core.$strip>;
+
 /** Shape returned by {@link insertMessageCore} and its public wrappers. */
 declare interface InsertedMessage {
     id: string;
@@ -5219,8 +5242,9 @@ declare const LIVE_VOICE_TURN_DETECTION_MODES: readonly ["manual", "server_vad"]
  * the room) can otherwise only say "Thinking...". The label is composed here
  * rather than by each surface for the same reason phase wording is composed
  * once: the Live Activity has two independent drivers (this socket and an APNs
- * push the daemon dispatches), they must carry identical content, and the only
- * way to guarantee that is for both to be handed the same string.
+ * push the daemon dispatches), they must carry compatible content. Structured
+ * kinds let in-conversation surfaces use localized copy while system-level
+ * surfaces can suppress internal detail.
  *
  * An empty `label` means "no current activity", which is what a turn's end
  * sends. Emitted only on change, never per tool result.
@@ -5229,6 +5253,12 @@ declare interface LiveVoiceActivityServerFrame extends LiveVoiceServerFrameBase
     readonly type: "activity";
     readonly turnId: string;
     readonly label: string;
+    /** Structured reason for the activity, when a client needs custom display. */
+    readonly kind?: "escalation";
+    /** Selected inference profile for diagnostics, never default UI copy. */
+    readonly profile?: string;
+    /** Why the selected profile won for this leg. */
+    readonly profileSource?: VoiceEscalationProfileSource;
     /**
      * The confirmation this turn is blocked on, when the label describes a wait
      * rather than work in flight. Absent otherwise.
@@ -6705,6 +6735,13 @@ export declare interface ResolvedSkillEntry {
     installMeta?: SkillInstallMeta | null;
 }
 
+/** A fully resolved provider target plus the winning profile's policy. */
+declare interface ResolvedVisionTarget {
+    readonly provider?: string;
+    readonly model: string;
+    readonly inputModalities?: InputModalities | null;
+}
+
 /**
  * Resolve a media source to inline base64, reading a reference source back from
  * its workspace location. Returns `null` when a reference can no longer be
@@ -8348,6 +8385,8 @@ export declare interface VideoEmbeddingInput {
  */
 export declare const VOICE_ESCALATION_CONTINUATION_MESSAGE_KIND = "voice_escalation_continuation";
 
+declare type VoiceEscalationProfileSource = "conversation" | "turn_override" | "image_compatibility" | "pre_model_hook" | "call_site";
+
 declare interface WebFetchMetadata {
     url: string;
     finalUrl: string;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vellumai/plugin-api",
-  "version": "0.12.2-dev.202609181515.7e76c92",
+  "version": "0.12.2-dev.202609181615.645c0a5",
   "description": "Public TypeScript authoring contract for Vellum assistant plugins.",
   "license": "MIT",
   "type": "module",