fix: add audio capability to OpenAI provider

The OpenAI provider had the `transcribeAudio` method implemented but
didn't declare "audio" in its capabilities array. This caused audio
transcription to fail when using OpenAI for Telegram voice messages.

Fixes #3539

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

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

src/media-understanding/providers/providers.test.ts

@@ -0,0 +1,65 @@
+import { describe, expect, it } from "vitest";
+
+import { buildMediaUnderstandingRegistry } from "./index.js";
+
+describe("media understanding providers", () => {
+  const registry = buildMediaUnderstandingRegistry();
+
+  it("providers declare capabilities matching their implemented methods", () => {
+    for (const [id, provider] of registry) {
+      const declared = provider.capabilities ?? [];
+
+      if (provider.transcribeAudio) {
+        expect(
+          declared.includes("audio"),
+          `Provider "${id}" has transcribeAudio but doesn't declare "audio" capability`,
+        ).toBe(true);
+      }
+
+      if (provider.describeImage) {
+        expect(
+          declared.includes("image"),
+          `Provider "${id}" has describeImage but doesn't declare "image" capability`,
+        ).toBe(true);
+      }
+
+      if (provider.describeVideo) {
+        expect(
+          declared.includes("video"),
+          `Provider "${id}" has describeVideo but doesn't declare "video" capability`,
+        ).toBe(true);
+      }
+    }
+  });
+
+  it("openai provider declares both image and audio capabilities", () => {
+    const openai = registry.get("openai");
+    expect(openai).toBeDefined();
+    expect(openai?.capabilities).toContain("image");
+    expect(openai?.capabilities).toContain("audio");
+    expect(openai?.describeImage).toBeDefined();
+    expect(openai?.transcribeAudio).toBeDefined();
+  });
+
+  it("groq provider declares audio capability", () => {
+    const groq = registry.get("groq");
+    expect(groq).toBeDefined();
+    expect(groq?.capabilities).toContain("audio");
+    expect(groq?.transcribeAudio).toBeDefined();
+  });
+
+  it("google provider declares image, audio, and video capabilities", () => {
+    const google = registry.get("google");
+    expect(google).toBeDefined();
+    expect(google?.capabilities).toContain("image");
+    expect(google?.capabilities).toContain("audio");
+    expect(google?.capabilities).toContain("video");
+  });
+
+  it("deepgram provider declares audio capability", () => {
+    const deepgram = registry.get("deepgram");
+    expect(deepgram).toBeDefined();
+    expect(deepgram?.capabilities).toContain("audio");
+    expect(deepgram?.transcribeAudio).toBeDefined();
+  });
+});

src/media-understanding/resolve.test.ts

@@ -4,7 +4,7 @@ import type { MoltbotConfig } from "../config/config.js";
 import { resolveEntriesWithActiveFallback, resolveModelEntries } from "./resolve.js";
 
 const providerRegistry = new Map([
-  ["openai", { capabilities: ["image"] }],
+  ["openai", { capabilities: ["image", "audio"] }],
   ["groq", { capabilities: ["audio"] }],
 ]);
 
@@ -30,7 +30,7 @@ describe("resolveModelEntries", () => {
       capability: "audio",
       providerRegistry,
     });
-    expect(audioEntries).toHaveLength(0);
+    expect(audioEntries).toHaveLength(1);
   });
 
   it("keeps per-capability entries even without explicit caps", () => {