From 8eece4506ab020e1e33183e6cf914d9c23ea7bf9 Mon Sep 17 00:00:00 2001
From: dev <dev@local>
Date: Sun, 9 Aug 2026 22:43:19 +0200
Subject: [PATCH] Aggiunto TTS via Gemini API: tool agy_tts, comando
 /agy:speak, notifiche vocali nel flusso F12

---
 extensions/index.ts | 122 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 122 insertions(+)

diff --git a/extensions/index.ts b/extensions/index.ts
index 7b51129..a020b45 100644
--- a/extensions/index.ts
+++ b/extensions/index.ts
@@ -386,6 +386,74 @@ function getConversationContext(ctx: any, maxEntries = 8): string {
 	}
 }
 
+// ---------------------------------------------------------------------------
+// TTS via Gemini API (nessun engine esterno)
+// ---------------------------------------------------------------------------
+async function ttsSpeak(text: string, outputDir?: string): Promise<string | null> {
+	let key = process.env.GEMINI_API_KEY ?? "";
+	if (!key) {
+		try {
+			key = fs.readFileSync(path.join(AGY_CHAT_DIR, "gemini-key"), "utf8").trim();
+		} catch {
+			/* ignora */
+		}
+	}
+	if (!key) return null;
+
+	const model = process.env.AGY_TTS_MODEL ?? "gemini-2.5-flash-preview-tts";
+	try {
+		const body = {
+			contents: [{ parts: [{ text }] }],
+			generationConfig: { responseModalities: ["AUDIO"] },
+		};
+		const res = await fetch(
+			`https://generativelanguage.googleapis.com/v1beta/models/${model}:generateContent?key=${key}`,
+			{
+				method: "POST",
+				headers: { "Content-Type": "application/json" },
+				body: JSON.stringify(body),
+				signal: AbortSignal.timeout(60_000),
+			},
+		);
+		if (!res.ok) return null;
+		const data: any = await res.json();
+		const parts = data?.candidates?.[0]?.content?.parts ?? [];
+		const audioPart = parts.find((p: any) => p.inlineData);
+		if (!audioPart) return null;
+
+		// salva PCM raw (16-bit, 24kHz) e converti in WAV
+		const pcmFile = path.join(os.tmpdir(), `tts-${Date.now()}.pcm`);
+		fs.writeFileSync(pcmFile, Buffer.from(audioPart.inlineData.data, "base64"));
+		const wavFile = pcmFile.replace(".pcm", ".wav");
+		await execFileAsync(
+			"ffmpeg",
+			["-y", "-f", "s16le", "-ar", "24000", "-ac", "1", "-i", pcmFile, wavFile],
+			{ timeout: 30_000 },
+		);
+		fs.rmSync(pcmFile, { force: true });
+
+		// copia in outputDir se richiesto
+		let finalFile = wavFile;
+		if (outputDir) {
+			try {
+				fs.mkdirSync(outputDir, { recursive: true });
+				const dest = path.join(outputDir, `tts-${Date.now()}.wav`);
+				fs.copyFileSync(wavFile, dest);
+				finalFile = dest;
+			} catch {
+				/* ignora */
+			}
+		}
+
+		// riproduci (fire-and-forget)
+		execFileAsync("paplay", [finalFile], { timeout: 60_000 }).catch(() => {});
+
+		return finalFile;
+	} catch {
+		return null;
+	}
+}
+
 // ---------------------------------------------------------------------------
 // Estensione
 // ---------------------------------------------------------------------------
@@ -827,6 +895,40 @@ export default function agyExtension(pi: ExtensionAPI) {
 		},
 	});
 
+	// =========================================================================
+	// TOOL: agy_tts — text-to-speech via Gemini API
+	// =========================================================================
+	pi.registerTool({
+		name: "agy_tts",
+		label: "agy TTS (text to speech)",
+		description:
+			"Converte un testo in audio (TTS) usando la Gemini API (gemini-2.5-flash-preview-tts) e lo riproduce. " +
+			"Utile per notifiche vocali, sintesi parlate e aggiornamenti di stato. " +
+			"Richiede la key Gemini in ~/.agy-chat/gemini-key o GEMINI_API_KEY.",
+		parameters: Type.Object({
+			text: Type.String({ description: "Il testo da pronunciare." }),
+			play: Type.Optional(Type.Boolean({ description: "true per riprodurre l'audio (default: true)." })),
+			outputDir: Type.Optional(Type.String({ description: "Cartella dove salvare il file audio." })),
+		}),
+		async execute(toolCallId, params, signal, onUpdate, ctx) {
+			const p = params as any;
+			onUpdate?.("agy_tts: sintesi vocale...");
+			return ttsSpeak(p.text, p.outputDir).then((file) => {
+				if (!file) {
+					return {
+						content: [{ type: "text", text: "TTS fallito: key Gemini mancante o errore API." }],
+						details: {},
+						isError: true,
+					};
+				}
+				return {
+					content: [{ type: "text", text: `Audio TTS generato: ${file}` }],
+					details: { audioFile: file },
+				};
+			});
+		},
+	});
+
 	// =========================================================================
 	// TOOL: agy_models — elenca i modelli disponibili
 	// =========================================================================
@@ -986,6 +1088,11 @@ export default function agyExtension(pi: ExtensionAPI) {
 			pi.sendUserMessage(finalText, { deliverAs: "followUp" });
 		}
 		ctx.ui.notify("✅ Trascrizione inviata come prompt a pi", "info");
+
+		// notifica vocale (disattivabile con AGY_TTS_NOTIFY=0)
+		if (process.env.AGY_TTS_NOTIFY !== "0") {
+			ttsSpeak("Trascrizione completata e inviata.").catch(() => {});
+		}
 	}
 
 	pi.registerShortcut("f12", {
@@ -1012,4 +1119,19 @@ export default function agyExtension(pi: ExtensionAPI) {
 			await handleRecordToggle(ctx);
 		},
 	});
+
+	pi.registerCommand("agy:speak", {
+		description: "Pronuncia un testo con TTS Gemini. Uso: /agy:speak <testo>",
+		handler: async (args, ctx) => {
+			if (!args?.trim()) {
+				ctx.ui.notify("Uso: /agy:speak <testo>", "error");
+				return;
+			}
+			ctx.ui.setStatus("agy-tts", "🔊 sintesi vocale...");
+			const file = await ttsSpeak(args.trim());
+			ctx.ui.setStatus("agy-tts", "");
+			if (file) ctx.ui.notify(`🔊 Audio: ${file}`, "info");
+			else ctx.ui.notify("TTS fallito (key Gemini mancante?)", "error");
+		},
+	});
 }