@kuralle-syrinx/gemini 2.1.1 → 3.1.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kuralle-syrinx/gemini",
-  "version": "2.1.1",
+  "version": "3.1.0",
   "private": false,
   "type": "module",
   "license": "MIT",
@@ -8,7 +8,7 @@
   "types": "./src/index.ts",
   "dependencies": {
     "@google/genai": "^1.0.0",
-    "@kuralle-syrinx/core": "2.1.1"
+    "@kuralle-syrinx/core": "3.1.0"
   },
   "devDependencies": {
     "typescript": "^5.7.0",

package/src/index.test.ts

@@ -73,4 +73,43 @@ describe("GeminiTTSPlugin", () => {
     bus.stop();
     await started;
   });
+
+  it("sends the raw text to Gemini by default (no hardcoded persona lead-in)", async () => {
+    generateContent.mockResolvedValue({
+      candidates: [{ content: { parts: [{ inlineData: { mimeType: "audio/pcm", data: Buffer.from(new Uint8Array([1, 2, 3, 4])).toString("base64") } }] } }],
+    });
+    const bus = new PipelineBusImpl();
+    const started = bus.start();
+    const plugin = new GeminiTTSPlugin();
+    await plugin.initialize(bus, { api_key: "test-gemini-key" });
+    bus.push(Route.Main, { kind: "tts.done", contextId: "t1", timestampMs: Date.now(), text: "Add Biology 101." });
+    await new Promise((resolve) => setTimeout(resolve, 50));
+
+    expect(generateContent).toHaveBeenCalledTimes(1);
+    const arg = generateContent.mock.calls[0]![0] as { contents: Array<{ parts: Array<{ text: string }> }> };
+    expect(arg.contents[0]!.parts[0]!.text).toBe("Add Biology 101.");
+
+    await plugin.close();
+    bus.stop();
+    await started;
+  });
+
+  it("prepends a configured instruction lead-in when provided", async () => {
+    generateContent.mockResolvedValue({
+      candidates: [{ content: { parts: [{ inlineData: { mimeType: "audio/pcm", data: Buffer.from(new Uint8Array([1, 2, 3, 4])).toString("base64") } }] } }],
+    });
+    const bus = new PipelineBusImpl();
+    const started = bus.start();
+    const plugin = new GeminiTTSPlugin();
+    await plugin.initialize(bus, { api_key: "test-gemini-key", instruction: "Read aloud warmly" });
+    bus.push(Route.Main, { kind: "tts.done", contextId: "t2", timestampMs: Date.now(), text: "Hi." });
+    await new Promise((resolve) => setTimeout(resolve, 50));
+
+    const arg = generateContent.mock.calls[0]![0] as { contents: Array<{ parts: Array<{ text: string }> }> };
+    expect(arg.contents[0]!.parts[0]!.text).toBe("Read aloud warmly: Hi.");
+
+    await plugin.close();
+    bus.stop();
+    await started;
+  });
 });

package/src/index.ts

@@ -59,6 +59,7 @@ export class GeminiTTSPlugin implements VoicePlugin {
   private apiKey: string = "";
   private model: string = DEFAULT_MODEL;
   private voiceName: string = DEFAULT_VOICE;
+  private instruction = "";
   private timeoutMs = 45_000;
   private abortController: AbortController | null = null;
   private textByContextId = new Map<string, string>();
@@ -71,6 +72,7 @@ export class GeminiTTSPlugin implements VoicePlugin {
     this.apiKey = requireStringConfig(config, "api_key");
     this.model = optionalStringConfig(config, "model") ?? DEFAULT_MODEL;
     this.voiceName = optionalStringConfig(config, "voice_name") ?? DEFAULT_VOICE;
+    this.instruction = optionalStringConfig(config, "instruction") ?? "";
     this.timeoutMs = readPositiveInteger(config["timeout_ms"], 45_000);
     this.retryConfig = readRetryConfig(config);
     assertAudioFormat(this.audioFormat);
@@ -171,7 +173,7 @@ export class GeminiTTSPlugin implements VoicePlugin {
     const response = await withTimeout(
       client.models.generateContent({
         model: this.model,
-        contents: [{ parts: [{ text: `Read this aloud in a natural student-support phone voice: ${text}` }] }],
+        contents: [{ parts: [{ text: this.instruction ? `${this.instruction}: ${text}` : text }] }],
         config: {
           responseModalities: ["AUDIO"],
           speechConfig: {