refactor: extend media understanding

src/media-understanding/providers/anthropic/index.ts

@@ -3,5 +3,6 @@ import { describeImageWithModel } from "../image.js";
 
 export const anthropicProvider: MediaUnderstandingProvider = {
   id: "anthropic",
+  capabilities: ["image"],
   describeImage: describeImageWithModel,
 };

src/media-understanding/providers/google/index.ts

@@ -4,6 +4,7 @@ import { describeGeminiVideo } from "./video.js";
 
 export const googleProvider: MediaUnderstandingProvider = {
   id: "google",
+  capabilities: ["image", "audio", "video"],
   describeImage: describeImageWithModel,
   describeVideo: describeGeminiVideo,
 };

src/media-understanding/providers/groq/index.ts

@@ -5,6 +5,7 @@ const DEFAULT_GROQ_AUDIO_BASE_URL = "https://api.groq.com/openai/v1";
 
 export const groqProvider: MediaUnderstandingProvider = {
   id: "groq",
+  capabilities: ["audio"],
   transcribeAudio: (req) =>
     transcribeOpenAiCompatibleAudio({
       ...req,

src/media-understanding/providers/index.ts

@@ -29,7 +29,16 @@ export function buildMediaUnderstandingRegistry(
   }
   if (overrides) {
     for (const [key, provider] of Object.entries(overrides)) {
-      registry.set(normalizeMediaProviderId(key), provider);
+      const normalizedKey = normalizeMediaProviderId(key);
+      const existing = registry.get(normalizedKey);
+      const merged = existing
+        ? {
+            ...existing,
+            ...provider,
+            capabilities: provider.capabilities ?? existing.capabilities,
+          }
+        : provider;
+      registry.set(normalizedKey, merged);
     }
   }
   return registry;

src/media-understanding/providers/minimax/index.ts

@@ -3,5 +3,6 @@ import { describeImageWithModel } from "../image.js";
 
 export const minimaxProvider: MediaUnderstandingProvider = {
   id: "minimax",
+  capabilities: ["image"],
   describeImage: describeImageWithModel,
 };

src/media-understanding/providers/openai/index.ts

@@ -4,6 +4,7 @@ import { transcribeOpenAiCompatibleAudio } from "./audio.js";
 
 export const openaiProvider: MediaUnderstandingProvider = {
   id: "openai",
+  capabilities: ["image"],
   describeImage: describeImageWithModel,
   transcribeAudio: transcribeOpenAiCompatibleAudio,
 };