imsg-mcp 1.6.2 → 1.7.0

package/dist/index.js

@@ -1,21 +1,21 @@
 #!/usr/bin/env node
 import { execFileSync, execFile } from "node:child_process";
 import { existsSync, mkdirSync, readdirSync, readFileSync, writeFileSync, renameSync, chmodSync, statSync, mkdtempSync, rmSync } from "node:fs";
-import { join, dirname, isAbsolute } from "node:path";
+import { join, dirname, basename, isAbsolute } from "node:path";
 import { promisify } from "node:util";
 import { Server } from "@modelcontextprotocol/sdk/server/index.js";
 import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
 import { ListToolsRequestSchema, CallToolRequestSchema, ListResourcesRequestSchema, ListResourceTemplatesRequestSchema, ReadResourceRequestSchema } from "@modelcontextprotocol/sdk/types.js";
-import { g as getHumansDirPath, s as sanitizeSlugPart, c as checkLocalAccess, f as formatAccessReport, a as shutdown, b as getImsgDbPath, d as getContactsDbPaths, e as getSlugsDbPath, h as appendLog, i as getFileLogLines, j as getLogDirectory, p as perf, k as checkMessagesAvailable, l as sendToChat, m as sendToChatId, n as sendMessageReliable, o as sendMessageAlt, q as sendAttachment, r as checkImessageAvailability, t as computeRelationshipLeaderboard, u as dispatchAnalytic, v as installShutdownHandlers, w as registerCleanup, x as enableStdinEofDetection, y as enableOrphanWatchdog, z as logStartup, A as info, B as startHeapMonitor, C as getLogs, D as setLastSendError, E as logShutdown, F as stopHeapMonitor, G as getLogFilePath, H as getLastSendError } from "./shutdown-Vh0gjb6m.js";
+import { g as getHumansDirPath, s as sanitizeSlugPart, a as getTranscribeCloudConfig, c as checkLocalAccess, f as formatAccessReport, b as shutdown, d as getImsgDbPath, e as getContactsDbPaths, h as getSlugsDbPath, i as appendLog, j as getFileLogLines, k as getLogDirectory, p as perf, l as checkMessagesAvailable, m as sendToChat, n as sendToChatId, o as sendMessageReliable, q as sendMessageAlt, r as sendAttachment, t as checkImessageAvailability, u as computeRelationshipLeaderboard, v as dispatchAnalytic, w as installShutdownHandlers, x as registerCleanup, y as enableStdinEofDetection, z as enableOrphanWatchdog, A as logStartup, B as info, C as startHeapMonitor, D as getLogs, E as setLastSendError, F as logShutdown, G as stopHeapMonitor, H as getLogFilePath, I as getLastSendError } from "./shutdown-CQ9wzrxA.js";
 import { homedir, tmpdir } from "node:os";
 import Database from "better-sqlite3";
-import { A as APP_VERSION, a as APP_NAME, r as renderAnalyticText } from "./meta-GGXPVODc.js";
-import { n as normalizedPhoneVariants, I as IMessageDB, r as rankFuzzy, h as hasNativeModule } from "./imessage-db-Dpxn-q2z.js";
+import { A as APP_VERSION, a as APP_NAME, r as renderAnalyticText } from "./meta-Lpz_PHJ5.js";
+import { n as normalizedPhoneVariants, I as IMessageDB, r as rankFuzzy, h as hasNativeModule } from "./imessage-db-Cid38QGT.js";
 import { parseUserDate } from "./date-parse-DJXMfq3a.js";
 import { streamExport } from "./exportStream-DgPVRZTZ.js";
 import { z } from "zod";
 import { randomUUID } from "node:crypto";
-import { m as minMessageId, r as resolveRecipient, d as defaultCountryFromEnv, i as installWatchdog, n as noteActivity, a as readWatchdogState } from "./watchdog-DZ92dXs_.js";
+import { m as minMessageId, r as resolveRecipient, d as defaultCountryFromEnv, i as installWatchdog, n as noteActivity, a as readWatchdogState } from "./watchdog-Cl5PoHhW.js";
 const DEFAULT_TTL_MS = 24 * 60 * 60 * 1e3;
 let dbPath = join(homedir(), ".imsg-mcp", "analytics-cache.db");
 let db = null;
@@ -1948,7 +1948,10 @@ const GetAttachmentOutputSchema = z.object({
   base64: z.string().optional(),
   converted: z.string().optional().describe("Set when source was HEIC and we converted to PNG."),
   mediaInfo: z.string().optional().describe("Compact duration/dimensions/codec summary for audio/video (via mdls)."),
-  transcript: z.string().optional().describe("Audio transcript when an optional transcriber (hear/yap/whisper-cli) is installed."),
+  transcript: z.string().optional().describe(
+    "Audio transcript. Local on-device transcriber (hear/yap/whisper-cli) when installed; otherwise the opt-in cloud fallback (IMSG_TRANSCRIBE_PROVIDER + IMSG_TRANSCRIBE_API_KEY) if configured."
+  ),
+  transcriptSource: z.enum(["local", "cloud"]).optional().describe("Where the transcript came from: on-device (local) or the opt-in cloud provider."),
   imageBlockIncluded: z.boolean().optional().describe("True when the tool result carries a real MCP image content block.")
 });
 const noArgsSchema = { type: "object", properties: {} };
@@ -2336,7 +2339,7 @@ const TOOLS = [
   },
   {
     name: "get_attachment",
-    description: "Fetch an attachment by ROWID (from search_attachments or a message's attachments[].rowId). IMAGES: returned as a real MCP image content block (downscaled ≤1536px, HEIC auto-converted) so the model can SEE it — plus full-size base64 in structuredContent when ≤ inlineMaxBytes (default 5MB). VIDEO: QuickLook poster frame as an image block + duration/resolution metadata + file path. AUDIO (voice memos): metadata + file path, and a transcript when a transcriber (hear, yap, or whisper-cli) is installed on PATH.",
+    description: "Fetch an attachment by ROWID (from search_attachments or a message's attachments[].rowId). IMAGES: returned as a real MCP image content block (downscaled ≤1536px, HEIC auto-converted) so the model can SEE it — plus full-size base64 in structuredContent when ≤ inlineMaxBytes (default 5MB). VIDEO: QuickLook poster frame as an image block + duration/resolution metadata + file path. AUDIO (voice memos): metadata + file path, and a transcript when a local transcriber (hear, yap, or whisper-cli) is installed on PATH, or via an opt-in OpenAI-compatible cloud endpoint (set IMSG_TRANSCRIBE_PROVIDER + IMSG_TRANSCRIBE_API_KEY — audio leaves the device; local always preferred).",
     annotations: annotations.read,
     inputSchema: {
       type: "object",
@@ -2568,6 +2571,42 @@ function transcribeAudio(path, timeoutMs = 6e4) {
     return null;
   }
 }
+async function transcribeAudioCloud(path, config, opts = {}) {
+  if (!existsSync(path)) return null;
+  const doFetch = opts.fetchImpl ?? globalThis.fetch;
+  const controller = new AbortController();
+  const timer = setTimeout(() => controller.abort(), opts.timeoutMs ?? 6e4);
+  try {
+    const form = new FormData();
+    form.append("file", new Blob([readFileSync(path)]), basename(path));
+    form.append("model", config.model);
+    form.append("response_format", "text");
+    const res = await doFetch(`${config.baseUrl}/audio/transcriptions`, {
+      method: "POST",
+      headers: { Authorization: `Bearer ${config.apiKey}` },
+      body: form,
+      signal: controller.signal
+    });
+    if (!res.ok) return null;
+    const text = (await res.text()).trim();
+    return text.length > 0 ? text : null;
+  } catch {
+    return null;
+  } finally {
+    clearTimeout(timer);
+  }
+}
+async function transcribeAudioBest(path, opts = {}) {
+  const timeoutMs = opts.timeoutMs ?? 6e4;
+  const local = transcribeAudio(path, timeoutMs);
+  if (local) return { transcript: local, source: "local" };
+  const cloud = opts.cloudConfig !== void 0 ? opts.cloudConfig : getTranscribeCloudConfig();
+  if (cloud) {
+    const t = await transcribeAudioCloud(path, cloud, { timeoutMs, fetchImpl: opts.fetchImpl });
+    if (t) return { transcript: t, source: "cloud" };
+  }
+  return null;
+}
 const INSTRUCTIONS_UUID = randomUUID();
 function wrapUntrusted(text) {
   if (text == null || text === "") return "";
@@ -3919,6 +3958,7 @@ ${formatted}`, {
     let finalPath = resolvedPath;
     let mediaInfo;
     let transcript;
+    let transcriptSource;
     let imageBlock = null;
     if (isImage) {
       imageBlock = imageBlockFromFile(resolvedPath);
@@ -3950,7 +3990,11 @@ ${formatted}`, {
     } else if (isAudio) {
       mediaInfo = mediaMetadata(resolvedPath) ?? void 0;
       if (sizeBytes <= TRANSCRIBE_MAX_BYTES) {
-        transcript = transcribeAudio(resolvedPath) ?? void 0;
+        const t = await transcribeAudioBest(resolvedPath);
+        if (t) {
+          transcript = t.transcript;
+          transcriptSource = t.source;
+        }
       }
     }
     const lines = [];
@@ -3963,10 +4007,18 @@ ${formatted}`, {
     } else if (isAudio) {
       lines.push(`Audio attachment ${rowId} (${sizeBytes}B${mediaInfo ? `, ${mediaInfo}` : ""}).`);
       if (transcript) {
-        lines.push("", "Transcript:", wrapUntrusted(sanitizeUserText(transcript) ?? ""));
+        lines.push(
+          "",
+          transcriptSource === "cloud" ? "Transcript (via cloud provider):" : "Transcript:",
+          wrapUntrusted(sanitizeUserText(transcript) ?? "")
+        );
+      } else if (detectTranscriber()) {
+        lines.push("Transcription produced no text.");
+      } else if (getTranscribeCloudConfig()) {
+        lines.push("Cloud transcription produced no text.");
       } else {
         lines.push(
-          detectTranscriber() ? "Transcription produced no text." : "No transcriber installed — `brew install yap` (macOS 26+) or `brew install sveinbjornt/hear/hear` enables on-device transcription."
+          "No transcriber installed — `brew install yap` (macOS 26+) or `brew install sveinbjornt/hear/hear` enables on-device transcription; or set IMSG_TRANSCRIBE_PROVIDER + IMSG_TRANSCRIBE_API_KEY for an OpenAI-compatible cloud fallback (audio leaves this device)."
         );
       }
       lines.push(`Path: ${resolvedPath}`);
@@ -3998,6 +4050,7 @@ ${formatted}`, {
         converted: convertedNote,
         mediaInfo,
         transcript,
+        transcriptSource,
         imageBlockIncluded: imageBlock != null
       }
     };