@writepanda/mcp 1.87.0 → 1.92.0

package/bin/server.mjs

@@ -165,7 +165,7 @@ const TOOLS = [
 	{
 		name: "system_get_transcription_language",
 		description:
-			"Read the active transcription engine selection for the current workspace. Returns { language } where language is one of: 'auto' (Parakeet TDT v3, default — auto-detects English + 25 European languages including Russian/Ukrainian), 'chinese' / 'japanese' / 'korean' / 'hindi' / 'arabic' / 'thai' (each routes to Whisper Large-v3-turbo). Use this when surfacing the active engine to the user or before recommending a switch.",
+			"Read the active transcription engine selection for the current workspace. Returns { language } where language is one of: 'auto' (Parakeet TDT v3, default — auto-detects English + 25 European languages including Russian/Ukrainian), or a Whisper Large-v3-turbo language: 'chinese' / 'japanese' / 'korean' / 'hindi' / 'arabic' / 'thai' / 'tamil' / 'telugu' / 'kannada' / 'malayalam' / 'bengali' / 'marathi' / 'gujarati' / 'punjabi'. Use this when surfacing the active engine to the user or before recommending a switch.",
 		inputSchema: { type: "object", properties: {} },
 		command: "system.getTranscriptionLanguage",
 	},
@@ -178,7 +178,23 @@ const TOOLS = [
 			properties: {
 				language: {
 					type: "string",
-					enum: ["auto", "chinese", "japanese", "korean", "hindi", "arabic", "thai"],
+					enum: [
+						"auto",
+						"chinese",
+						"japanese",
+						"korean",
+						"hindi",
+						"arabic",
+						"thai",
+						"tamil",
+						"telugu",
+						"kannada",
+						"malayalam",
+						"bengali",
+						"marathi",
+						"gujarati",
+						"punjabi",
+					],
 					description: "Target language. 'auto' = Parakeet (default); the others route to Whisper.",
 				},
 			},
@@ -444,6 +460,54 @@ const TOOLS = [
 		},
 		command: "export.publish-youtube",
 	},
+	{
+		name: "instagram_connect",
+		description:
+			"Start connecting an Instagram account. Opens the user's browser to the Instagram consent page and returns `{ connectionId, redirectUrl }`. INTERACTIVE: the user must finish consent in the browser, then poll `instagram_status` until active before publishing. Only Business/Creator accounts can publish (a Meta restriction). Requires an activated PandaStudio license.",
+		inputSchema: { type: "object", properties: {} },
+		command: "instagram.connect",
+	},
+	{
+		name: "instagram_status",
+		description:
+			"Check whether the Instagram connection is active yet. Returns `{ status, active }`. Poll after `instagram_connect` until `active` is true.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				connectionId: { type: "string", description: "From instagram_connect (optional)." },
+			},
+		},
+		command: "instagram.status",
+	},
+	{
+		name: "instagram_account",
+		description:
+			"Read the connected Instagram account: `username`, `account_type`, and `publishable` (true only for Business/Creator accounts). Call before `export_publish_instagram` to warn the user if their account type can't publish.",
+		inputSchema: { type: "object", properties: {} },
+		command: "instagram.account",
+	},
+	{
+		name: "instagram_disconnect",
+		description:
+			"Disconnect the user's Instagram account. Already-published Reels stay on Instagram.",
+		inputSchema: { type: "object", properties: {} },
+		command: "instagram.disconnect",
+	},
+	{
+		name: "export_publish_instagram",
+		description:
+			"Publish an export-library entry to Instagram as a Reel. The video uploads directly to Composio storage (never through our server) and publishes via the broker. Returns `{ mediaId, permalink }`. LONG-RUNNING: Instagram processes the Reel for ~30-120 s before it goes live. Requires a connected Business/Creator account (check `instagram_account` first). Instagram only accepts short-form video — export a 9:16 clip up to ~90 s.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				id: { type: "string", description: "Export entry id." },
+				caption: { type: "string", description: "Reel caption, max 2200 chars; hashtags go here." },
+				shareToFeed: { type: "boolean", description: "Also show in the feed grid. Default true." },
+			},
+			required: ["id"],
+		},
+		command: "export.publish-instagram",
+	},
 	{
 		name: "export_list_youtube",
 		description:
@@ -668,13 +732,18 @@ const TOOLS = [
 	{
 		name: "project_add_clip",
 		description:
-			"Append (or insert) a video clip into the project's main track. Probes duration via FFmpeg.",
+			"Append (or insert) a video clip into the project's main track. Probes duration via FFmpeg. Pass `webcam` (a second, separately-recorded source) to make the clip a two-speaker podcast composite (media=host, webcam=guest, kind=podcast) — for 3+ speakers or audio auto-sync use project_add_podcast_clip instead.",
 		inputSchema: {
 			type: "object",
 			properties: {
 				id: { type: "string" },
 				path: { type: "string" },
-				media: { type: "string", description: "Absolute path to video file" },
+				media: { type: "string", description: "Absolute path to video file (the host source)" },
+				webcam: {
+					type: "string",
+					description:
+						"Optional absolute path to a SECOND, separately-recorded source (the guest). When set, the clip becomes a two-speaker podcast composite. Use project_add_podcast_clip for 3+ participants or audio auto-sync.",
+				},
 				atIndex: {
 					type: "number",
 					description: "Insert before this zero-based clip index. Omit to append at the end.",
@@ -685,6 +754,83 @@ const TOOLS = [
 		},
 		command: "project.add-clip",
 	},
+	{
+		name: "project_add_podcast_clip",
+		description:
+			"Assemble 2-4 separately-recorded sources (e.g. host + guest exported from Zoom/Riverside) into ONE podcast composite clip — the import equivalent of an in-app/remote podcast recording. media=host (the authoritative clock); guests=ordered guest paths (1-3). Speakers are assigned host/guest/guest-2/guest-3 and stored as participants[]. By default auto-syncs each guest to the host via audio cross-correlation (set autoSync=false to skip); fine-tune later with project_set_participant_offset. After this, run transcript_transcribe (it transcribes each speaker separately and merges into one speaker-attributed transcript), then edit by transcript exactly like a recorded podcast. Preview + export render every participant via the existing N-party podcast layout.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				id: { type: "string" },
+				path: { type: "string" },
+				media: { type: "string", description: "Absolute path to the HOST video (reference clock)" },
+				guests: {
+					type: "array",
+					items: { type: "string" },
+					description:
+						"Absolute paths to the 1-3 guest videos in participant order (guest, guest-2, guest-3).",
+				},
+				labels: {
+					type: "array",
+					items: { type: "string" },
+					description:
+						"Optional display names parallel to [host, ...guests] (index 0 = host). Falls back to the speaker id.",
+				},
+				autoSync: {
+					type: "boolean",
+					description:
+						"Audio-align each guest to the host before assembling (default true). Set false to assemble with zero offsets.",
+				},
+				atIndex: {
+					type: "number",
+					description: "0-based insert position in the clips array. Omit to append.",
+				},
+				expectedRevision: { type: "number" },
+			},
+			required: ["media", "guests"],
+		},
+		command: "project.add-podcast-clip",
+	},
+	{
+		name: "project_set_participant_offset",
+		description:
+			"Fine-tune ONE participant's A/V sync on a multi-party podcast composite. speaker = guest | guest-2 | guest-3 (the host is the reference clock and can't be offset). offsetMs shifts that participant's video AND audio vs the host: positive delays, negative advances, 0 clears (±10000). The N-party analogue of project_set_webcam_offset (which only reaches guest-1). Honored in preview + export.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				id: { type: "string" },
+				path: { type: "string" },
+				clipId: { type: "string", description: "Clip id of the podcast composite" },
+				speaker: {
+					type: "string",
+					description: "Which participant to nudge: guest | guest-2 | guest-3",
+				},
+				offsetMs: {
+					type: "number",
+					description: "Time nudge in ms (±10000). Positive delays this guest; 0 clears.",
+				},
+				expectedRevision: { type: "number" },
+			},
+			required: ["clipId", "speaker", "offsetMs"],
+		},
+		command: "project.set-participant-offset",
+	},
+	{
+		name: "project_auto_sync_podcast",
+		description:
+			"Re-run audio auto-sync on an existing podcast composite clip: cross-correlates each guest's audio against the host and stores the resulting offset on each participant. Use after project_add_podcast_clip with autoSync=false, or to re-align if a manual nudge drifted. Returns the per-guest offsets + confidence; fine-tune afterward with project_set_participant_offset.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				id: { type: "string" },
+				path: { type: "string" },
+				clipId: { type: "string", description: "Clip id of the podcast composite" },
+				expectedRevision: { type: "number" },
+			},
+			required: ["clipId"],
+		},
+		command: "project.auto-sync-podcast",
+	},
 	{
 		name: "project_add_motion_graphic",
 		description:
@@ -1301,19 +1447,22 @@ const TOOLS = [
 	{
 		name: "project_save",
 		description:
-			"Overwrite a project with the full JSON you supply. Use this when you've built or mutated a project object client-side and want to persist it. Always pass `expectedRevision` (the revision you read) — the server will reject the write if a concurrent edit happened between your read and save, returning { ok: false, details: { code: 'revision_conflict' } }.",
+			"Persist a project. Two modes: (1) pass a full `project` JSON to OVERWRITE the file (use after building/mutating a project object client-side) — always pass `expectedRevision` so a concurrent edit returns { ok:false, details:{ code:'revision_conflict' } }; (2) OMIT `project` to simply confirm the current on-disk state. The add_*/set_* verbs already autosave, so a bodyless save is a no-op confirmation that returns the current revision — NOT an error. You usually don't need an explicit save at all.",
 		inputSchema: {
 			type: "object",
 			properties: {
 				id: { type: "string" },
 				path: { type: "string" },
-				project: { type: "object", description: "Full project JSON to persist" },
+				project: {
+					type: "object",
+					description:
+						"Full project JSON to persist (overwrite mode). Omit to just confirm current on-disk state — edits via add_*/set_* verbs already autosave.",
+				},
 				expectedRevision: {
 					type: "number",
-					description: "Revision from your last project_read. Strongly recommended.",
+					description: "Revision from your last project_read. Strongly recommended in overwrite mode.",
 				},
 			},
-			required: ["project"],
 		},
 		command: "project.save",
 	},
@@ -1686,6 +1835,16 @@ const TOOLS = [
 					description: "Fallback duration when neither endMs nor maxDurationMs is given.",
 				},
 				volume: { type: "number", description: "Volume multiplier 0–2. Default 0.8." },
+				fadeIn: {
+					type: "number",
+					description:
+						"Fade-in ramp at the overlay's audible start, in ms (e.g. 1000 = 1s). Applied by the export mixer. Omit for a hard start.",
+				},
+				fadeOut: {
+					type: "number",
+					description:
+						"Fade-out ramp ending at the overlay's audible end, in ms. Applied for bounded overlays (endMs/maxDurationMs set); ignored on uncapped full-length overlays. Useful for seam-crossfading a looped music bed. Omit for a hard stop.",
+				},
 				maxDurationMs: {
 					type: "number",
 					description:
@@ -2497,6 +2656,20 @@ const TOOLS = [
 		},
 		command: "llm.generate-description",
 	},
+	{
+		name: "llm_generate_caption",
+		description:
+			"Generate an Instagram Reel caption (1-3 punchy sentences + a few hashtags) from the merged transcript, on the local model. Feed the result into export_publish_instagram's caption.",
+		inputSchema: {
+			type: "object",
+			properties: {
+				id: { type: "string" },
+				path: { type: "string" },
+				maxChars: { type: "number", description: "Default 2200" },
+			},
+		},
+		command: "llm.generate-caption",
+	},
 	{
 		name: "llm_generate_timestamps",
 		description: "Generate YouTube-style chapter timestamps from the merged transcript.",
@@ -2760,7 +2933,7 @@ const TOOLS = [
 const server = new Server(
 	{
 		name: "pandastudio",
-		version: "1.70.0",
+		version: "1.71.0",
 	},
 	{
 		capabilities: {

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@writepanda/mcp",
-	"version": "1.87.0",
+	"version": "1.92.0",
 	"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",