@writepanda/mcp 1.99.0 → 1.100.0

package/bin/server.mjs

@@ -229,6 +229,40 @@ const TOOLS = [
 		inputSchema: { type: "object", properties: {} },
 		command: "system.isWhisperModelDownloaded",
 	},
+	{
+		name: "system_get_narration_engine",
+		description:
+			"Read the active narration (TTS) engine for the current workspace. Returns { engine } — 'local-kokoro' (default; Kokoro-82M on-device, English, no key/cloud) or 'replicate' (cloud TTS, needs the user's Replicate key). Use before recommending a switch or when disclosing which engine narration uses.",
+		inputSchema: { type: "object", properties: {} },
+		command: "system.getNarrationEngine",
+	},
+	{
+		name: "system_set_narration_engine",
+		description:
+			"Set the DEFAULT narration engine for the current workspace. Pass `engine` = 'local-kokoro' (on-device English, no key) or 'replicate' (cloud, needs the user's Replicate key). Individual media_generate_narration calls can still override this via their `model` arg. Switching to local triggers a one-time ~330 MB model download on first use — call system_is_kokoro_model_downloaded to check.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				engine: { type: "string", description: "'local-kokoro' | 'replicate'." },
+			},
+			required: ["engine"],
+		},
+		command: "system.setNarrationEngine",
+	},
+	{
+		name: "system_is_kokoro_model_downloaded",
+		description:
+			"Check whether the local Kokoro-82M TTS model (English on-device narration) has been downloaded. Returns { downloaded: boolean }. The model (~330 MB) auto-downloads on first local narration; calling this first lets you surface the download instead of a long first-run wait.",
+		inputSchema: { type: "object", properties: {} },
+		command: "system.isKokoroModelDownloaded",
+	},
+	{
+		name: "system_download_kokoro_model",
+		description:
+			"Pre-download the local Kokoro-82M TTS model + voices (~330 MB) now so the first narration doesn't pay the wait. Idempotent; returns { downloaded: true }. Optional — narration also auto-downloads on first use.",
+		inputSchema: { type: "object", properties: {} },
+		command: "system.downloadKokoroModel",
+	},
 
 	// ── workspaces (v1.19) ──────────────────────────────────────────
 	// Multi-workspace isolation: each workspace has its own projects,
@@ -2432,27 +2466,29 @@ const TOOLS = [
 	{
 		name: "media_generate_narration",
 		description:
-			"Generate a voiceover / narration audio clip via Replicate TTS, for promo videos and explainers. Project-agnostic; writes audio to <userData>/narration/ and returns { audioPath, durationMs, model, voice, predictionId }. Canonical workflow: media_generate_narration → project_add_audio (place at startMs, pass the returned durationMs so the overlay is sized to the speech). Models: 'elevenlabs-v3' (default, most expressive — embed inline delivery tags in the text like [excited] or [whispers]) | 'gemini-flash-tts' (30 voices, strong multilingual; 'style' sets tone) | 'minimax-turbo' (fast; 'style' maps to an emotion, 'speed' 0.5-2). Voices — gemini e.g. Kore/Puck/Charon; minimax e.g. Friendly_Person/Wise_Woman/Casual_Guy; omit voice for the model default. Requires the user's Replicate API key (Settings → Integrations).",
+			"Generate a voiceover / narration audio clip. Project-agnostic; writes audio to <userData>/narration/ and returns { audioPath, durationMs, model, voice, predictionId }. Canonical workflow: media_generate_narration → project_add_audio (place at startMs, pass the returned durationMs so the overlay is sized to the speech). DEFAULT is the LOCAL, on-device Kokoro engine (English) — no API key, no cloud, runs on the user's machine; the model auto-downloads (~330 MB) on first use. Kokoro voices (pass in `voice`): af_heart (default), af_bella, am_michael, bf_emma, bm_george, etc. (a*=American, b*=British; f=female, m=male). To use CLOUD TTS instead (more voices/languages, needs the user's Replicate key), set `model` to a Replicate model: 'elevenlabs-v3' (most expressive — embed inline tags like [excited]/[whispers]) | 'gemini-flash-tts' (30 voices, multilingual, 'style' sets tone) | 'minimax-turbo' (fast, 'style' maps to an emotion, 'speed' 0.5-2). Cloud voices — gemini Kore/Puck/Charon; minimax Friendly_Person/Wise_Woman. Omit `model` to honour the workspace default (system_get_narration_engine).",
 		inputSchema: {
 			type: "object",
 			properties: {
 				text: {
 					type: "string",
 					description:
-						"The script to speak. For elevenlabs-v3 you can embed delivery tags inline, e.g. 'Welcome [excited] to the future of editing.'",
+						"The script to speak. For the cloud elevenlabs-v3 model you can embed delivery tags inline, e.g. 'Welcome [excited] to the future of editing.' (Local Kokoro ignores such tags.)",
 				},
 				model: {
 					type: "string",
-					description: "'elevenlabs-v3' (default) | 'gemini-flash-tts' | 'minimax-turbo'.",
+					description:
+						"Engine/model selector. Omit for the workspace default (local Kokoro). 'kokoro-local' forces on-device English. Cloud (needs Replicate key): 'elevenlabs-v3' | 'gemini-flash-tts' | 'minimax-turbo'.",
 				},
 				voice: {
 					type: "string",
 					description:
-						"Voice id/name for the chosen model (gemini: Kore/Puck/...; minimax: Friendly_Person/...). Omit for the model default.",
+						"Voice for the chosen engine. Local Kokoro: af_heart/af_bella/am_michael/bf_emma/bm_george/... Cloud: gemini Kore/Puck/...; minimax Friendly_Person/... Omit for the engine's default voice.",
 				},
 				language: {
 					type: "string",
-					description: "Optional BCP-47 language hint, e.g. 'en-US', 'es-ES'.",
+					description:
+						"Optional BCP-47 language hint for the cloud models, e.g. 'en-US', 'es-ES'. Local Kokoro is English-only and ignores this.",
 				},
 				style: {
 					type: "string",
@@ -3054,7 +3090,7 @@ const TOOLS = [
 const server = new Server(
 	{
 		name: "pandastudio",
-		version: "1.72.0",
+		version: "1.73.0",
 	},
 	{
 		capabilities: {

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@writepanda/mcp",
-	"version": "1.99.0",
+	"version": "1.100.0",
 	"description": "Model Context Protocol server for PandaStudio. Exposes the desktop video editor's automation surface to Cursor, Continue, Cline, Claude Desktop, and any MCP-compliant client.",
 	"keywords": [
 		"pandastudio",