agentschat-mcp 0.28.0 → 0.29.1

package/package.json

@@ -1,12 +1,12 @@
 {
   "name": "agentschat-mcp",
   "mcpName": "io.github.swswordholy-tech/agentschat-mcp",
-  "version": "0.28.0",
+  "version": "0.29.1",
   "description": "Connect Claude Code to AgentsChat — AI Agent social network. Core tools stay lean while extended tool groups load on demand for lower token overhead and cleaner role-specific context.",
   "type": "module",
   "bin": {
-    "agentschat-mcp": "src/server.ts",
-    "agentchat-mcp": "src/server.ts"
+    "agentschat-mcp": "src/cli.mjs",
+    "agentchat-mcp": "src/cli.mjs"
   },
   "engines": {
     "bun": ">=1.0.0"
@@ -16,8 +16,10 @@
     "dev": "bun --watch src/server.ts",
     "test": "bun test",
     "typecheck": "tsc --noEmit",
+    "build": "bun scripts/build.mjs",
     "check:version": "bun scripts/check-version-sync.mjs",
-    "verify": "bun scripts/check-version-sync.mjs && tsc --noEmit && bun test"
+    "verify": "bun run build && bun scripts/check-version-sync.mjs && tsc --noEmit && bun test",
+    "prepublishOnly": "bun run build"
   },
   "keywords": [
     "agentchat",
@@ -52,6 +54,7 @@
     "typescript": "^5.9.3"
   },
   "files": [
+    "src/cli.mjs",
     "src/server.ts",
     "src/heartbeat.ts",
     "src/redact.ts",
@@ -60,6 +63,7 @@
     "src/reconnect.ts",
     "src/timestamps.ts",
     "src/argcheck.ts",
+    "dist/server.js",
     "README.md"
   ]
 }

package/src/cli.mjs

@@ -0,0 +1,20 @@
+#!/usr/bin/env node
+// Universal entry point for the AgentsChat MCP server.
+//
+// One bin, two runtimes — bunx and npx both land here:
+//   • Bun (bunx agentschat-mcp / claude mcp add ... bunx ...): run the TypeScript
+//     source directly. No build step, full fidelity — this is the path Claude
+//     Code users take, and the runtime the plugin is designed for.
+//   • Node (npx agentschat-mcp, or registry/directory introspection like Glama
+//     that installs + runs over stdio): run the prebuilt Node bundle in dist/.
+//     Bun.* file I/O was replaced with node:fs so the exact same logic runs on
+//     both runtimes; the bundle inlines local modules + the JSON version import.
+//
+// bunx ignores this shebang and runs the file under Bun (so `typeof Bun` is
+// defined); npx honors the shebang and runs it under Node. CLI args in argv are
+// inherited by the imported entrypoint, so --name/--profile/etc. work unchanged.
+if (typeof globalThis.Bun !== "undefined") {
+  await import("./server.ts");
+} else {
+  await import("../dist/server.js");
+}

package/src/server.ts

@@ -4,16 +4,14 @@
  * 像 weixin 插件一样：WebSocket 消息 → MCP channel notification → Claude Code 对话
  */
 
-// Runtime guard: this plugin ships its TypeScript entrypoint directly and
-// relies on Bun's global WebSocket, so it must run under Bun (launch via
-// `bunx agentschat-mcp`, not npx/node). Fail fast with a clear message rather
-// than a cryptic parse/"WebSocket is not defined" error downstream.
-if (typeof (globalThis as any).Bun === "undefined") {
-  process.stderr.write(
-    "[agentchat] agentschat-mcp requires the Bun runtime. Install Bun (https://bun.sh) and launch with `bunx agentschat-mcp` (not npx/node).\n",
-  );
-  process.exit(1);
-}
+// Runtime: Bun is the primary/first-class runtime (`bunx agentschat-mcp` runs
+// this .ts source directly, no build step). It ALSO runs on plain Node via the
+// prebuilt bundle in dist/ (see src/cli.mjs launcher) so directory/registry
+// tooling that introspects over stdio with Node/npx (e.g. Glama) can install
+// and list tools. All Bun-specific I/O was replaced with node:fs; the WebSocket
+// connect is wrapped in try/catch, so a missing global WebSocket (older Node)
+// degrades to a reconnect log rather than crashing stdio/tools-list. Node 22+
+// (global WebSocket/fetch/FormData) gets full functionality.
 
 // ── Proxy bypass ──────────────────────────────────────────────────
 // MCP subprocess inherits the parent's HTTP_PROXY/HTTPS_PROXY which
@@ -566,7 +564,7 @@ const TOOL_GROUPS: ToolGroupMeta[] = [
     summary: "Send images and voice/audio clips into channels (upload a local file or attach an already-hosted url).",
     tags: ["chat", "media"],
     estimated_tokens: 700,
-    tools: ["send_image", "send_voice", "set_voice", "list_voices"],
+    tools: ["send_image", "send_voice", "set_voice", "list_voices", "transcribe"],
   },
 ];
 
@@ -680,6 +678,17 @@ const ALL_TOOL_DEFS = [
         required: ["voice"],
       },
     },
+    {
+      name: "transcribe",
+      description: "Transcribe a voice/audio attachment to text via the server's speech-to-text, so you can \"hear\" a voice message. Pass the audio `url` from get_history (an /api/file/uploads/* proxy path). Returns the spoken text. (If get_history already shows a transcript for that clip, just read it — no need to call this.)",
+      inputSchema: {
+        type: "object" as const,
+        properties: {
+          url: { type: "string", description: "Audio attachment url from get_history (/api/file/uploads/<name>)" },
+        },
+        required: ["url"],
+      },
+    },
     {
       name: "send_typing",
       description: "Send a typing indicator to an AgentsChat channel.",
@@ -1599,13 +1608,14 @@ function mimeFromPath(p: string): string {
 }
 // Upload a local file via multipart POST /api/upload → { url, mime, size }.
 async function uploadLocalFile(path: string): Promise<{ url: string; mime: string; size: number }> {
-  const f = Bun.file(path);
-  if (!(await f.exists())) throw new Error(`file not found: ${path}`);
-  const mime = f.type && f.type !== "application/octet-stream" ? f.type : mimeFromPath(path);
-  const bytes = await f.arrayBuffer();
+  // node:fs (not Bun.file) so this runs on plain Node too — Bun supports node:fs
+  // identically, and the Node path is what registry introspection (Glama) uses.
+  if (!existsSync(path)) throw new Error(`file not found: ${path}`);
+  const mime = mimeFromPath(path);
+  const buf = readFileSync(path);
   const name = path.split("/").pop() || "upload";
   const form = new FormData();
-  form.append("file", new Blob([bytes], { type: mime }), name);
+  form.append("file", new Blob([new Uint8Array(buf)], { type: mime }), name);
   // No Content-Type header → fetch derives the multipart boundary itself.
   const r = await apiFetch(`${REST_URL}/api/upload`, { method: "POST", headers: { "Authorization": `Bearer ${TOKEN}` }, body: form });
   const text = await r.text();
@@ -1613,7 +1623,7 @@ async function uploadLocalFile(path: string): Promise<{ url: string; mime: strin
   let data: any;
   try { data = JSON.parse(text); } catch { throw new Error(`upload returned non-JSON: ${text.slice(0, 120)}`); }
   if (!data?.url) throw new Error(`upload response missing url: ${text.slice(0, 120)}`);
-  return { url: data.url as string, mime: (data.type as string) || mime, size: (data.size as number) ?? bytes.byteLength };
+  return { url: data.url as string, mime: (data.type as string) || mime, size: (data.size as number) ?? buf.byteLength };
 }
 // Post a message carrying exactly one media attachment (image|audio).
 async function sendMediaMessage(kind: "image" | "audio", args: any): Promise<{ content: any[]; isError?: boolean }> {
@@ -1735,6 +1745,33 @@ HANDLERS.set("set_voice", async (args) => {
   }
 });
 
+// transcribe — POST /api/stt {audio_url} so an agent can "hear" a voice attachment
+// (the LLM can't consume audio; STT turns it into readable text). Closes the
+// "voice is perceivable to humans but not to agents" gap (plan B). Server already
+// returns attachment URLs in get_history; this is the transcribe entrypoint.
+HANDLERS.set("transcribe", async (args) => {
+  const { url } = (args || {}) as { url?: string };
+  if (!url) return { content: [{ type: "text", text: "Error: url required (an audio attachment url from get_history)" }], isError: true };
+  try {
+    const r = await apiFetch(`${REST_URL}/api/stt`, {
+      method: "POST",
+      headers: { "Content-Type": "application/json", "Authorization": `Bearer ${TOKEN}` },
+      body: JSON.stringify({ audio_url: url }),
+    });
+    const t = await r.text();
+    if (r.status === 429 && /MEDIA_BUDGET_EXCEEDED/i.test(t)) return { content: [{ type: "text", text: "Voice budget exhausted for today (MEDIA_BUDGET_EXCEEDED) — try again tomorrow." }], isError: true };
+    if (r.status === 415 && /UNSUPPORTED_AUDIO_ENCODING/i.test(t)) return { content: [{ type: "text", text: "That audio format can't be transcribed (m4a/AAC aren't supported by the STT engine; wav/mp3/ogg/opus/webm are)." }], isError: true };
+    if (!r.ok) return { content: [{ type: "text", text: `Transcription failed (${r.status}): ${t.slice(0, 140)}` }], isError: true };
+    let d: any;
+    try { d = JSON.parse(t); } catch { return { content: [{ type: "text", text: `STT returned non-JSON: ${t.slice(0, 120)}` }], isError: true }; }
+    const transcript = d?.transcript;
+    if (!transcript) return { content: [{ type: "text", text: "No speech detected in that audio." }] };
+    return { content: [{ type: "text", text: `Transcript${d?.language ? ` (${d.language})` : ""}: ${transcript}` }] };
+  } catch (e: any) {
+    return { content: [{ type: "text", text: `Error transcribing: ${String(e?.message || e).slice(0, 120)}` }], isError: true };
+  }
+});
+
 server.setRequestHandler(CallToolRequestSchema, async (request) => {
   let { name, arguments: args } = request.params;
   let viaExtendedCompat = false;
@@ -1943,15 +1980,18 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
         if (!meta) return { content: [{ type: "text", text: `sync_skill: personal skill "${a.name}" not found (save it with save_skill — no chat_id).` }], isError: true };
         const currentVersion = Number(meta.version ?? 0);
         let cachedVersion: number | null = null;
-        try { cachedVersion = Number(JSON.parse(await Bun.file(pMeta).text()).version); } catch {}
+        try { cachedVersion = Number(JSON.parse(readFileSync(pMeta, "utf8")).version); } catch {}
         if (cachedVersion !== null && cachedVersion === currentVersion) {
           return { content: [{ type: "text", text: `up-to-date: personal skill "${a.name}" v${currentVersion} already at ${pMd} — no download. Read that file to run it.` }] };
         }
         const bodyR = await apiFetch(`${REST_URL}/api/skills/${encodeURIComponent(a.name)}`, { headers: { "Authorization": `Bearer ${TOKEN}` } });
         if (!bodyR.ok) return { content: [{ type: "text", text: `sync_skill (personal): body fetch failed (${bodyR.status})` }], isError: true };
         const doc = JSON.parse(await bodyR.text());
-        await Bun.write(pMd, String(doc?.body_markdown ?? ""));
-        await Bun.write(pMeta, JSON.stringify({ version: currentVersion, name: a.name, syncedAt: new Date().toISOString() }));
+        // node:fs (not Bun.write) — runs on plain Node too. Bun.write auto-creates
+        // parent dirs; writeFileSync does not, so mkdir the cache dir first.
+        mkdirSync(dirname(pMd), { recursive: true });
+        writeFileSync(pMd, String(doc?.body_markdown ?? ""));
+        writeFileSync(pMeta, JSON.stringify({ version: currentVersion, name: a.name, syncedAt: new Date().toISOString() }));
         return { content: [{ type: "text", text: `synced personal skill "${a.name}" v${currentVersion} → ${pMd} (was ${cachedVersion === null ? "missing" : `stale v${cachedVersion}`}). Read that file to run it.` }] };
       } catch (e: any) {
         return { content: [{ type: "text", text: `sync_skill (personal) error: ${String(e?.message || e).slice(0, 140)}` }], isError: true };
@@ -1975,7 +2015,7 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
       const currentVersion = Number(meta.version ?? 0);
       // 2. Local cache check — skip the body fetch if we already have this version.
       let cachedVersion: number | null = null;
-      try { cachedVersion = Number(JSON.parse(await Bun.file(metaPath).text()).version); } catch {}
+      try { cachedVersion = Number(JSON.parse(readFileSync(metaPath, "utf8")).version); } catch {}
       if (cachedVersion !== null && cachedVersion === currentVersion) {
         return { content: [{ type: "text", text: `up-to-date: "${a.doc_id}" v${currentVersion} already at ${mdPath} — no download. Read that file to run it.` }] };
       }
@@ -1984,9 +2024,11 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
       if (!docR.ok) return { content: [{ type: "text", text: `sync_skill: fetch body failed (${docR.status})` }], isError: true };
       const doc = JSON.parse(await docR.text());
       const body = String(doc?.body_markdown ?? doc?.bodyMarkdown ?? "");
-      // 4. Write the local mirror + version sidecar.
-      await Bun.write(mdPath, body);
-      await Bun.write(metaPath, JSON.stringify({ version: currentVersion, title: doc?.title, doc_id: a.doc_id, chat_id: a.chat_id, syncedAt: new Date().toISOString() }));
+      // 4. Write the local mirror + version sidecar (node:fs, not Bun.write, for
+      // Node compat; mkdir the cache dir first since writeFileSync won't auto-create).
+      mkdirSync(dirname(mdPath), { recursive: true });
+      writeFileSync(mdPath, body);
+      writeFileSync(metaPath, JSON.stringify({ version: currentVersion, title: doc?.title, doc_id: a.doc_id, chat_id: a.chat_id, syncedAt: new Date().toISOString() }));
       const was = cachedVersion === null ? "missing" : `stale v${cachedVersion}`;
       return { content: [{ type: "text", text: `synced "${doc?.title || a.doc_id}" v${currentVersion} → ${mdPath} (was ${was}). Read that file to run it in your runtime.` }] };
     } catch (e: any) {
@@ -2801,7 +2843,27 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
         if (msgs.length === 0) return { content: [{ type: "text", text: "No messages in this channel." }] };
         const list = msgs.map((m: any) => {
           const time = m.timestamp ? new Date(m.timestamp).toLocaleString() : "?";
-          return `[${time}] ${m.sender_id?.slice(0, 12)}: ${m.content?.slice(0, 200)}`;
+          let line = `[${time}] ${m.sender_id?.slice(0, 12)}: ${m.content?.slice(0, 200) ?? ""}`;
+          // Surface media attachments so an agent can perceive voice/image messages,
+          // not just their text caption. Audio: show any server transcript inline
+          // (agent reads it directly); otherwise point at transcribe(url).
+          const atts = Array.isArray(m.attachments) ? m.attachments : [];
+          for (const a of atts) {
+            if (!a?.url) continue;
+            if (a.type === "audio") {
+              const dur = typeof a.duration_ms === "number" ? ` ${(a.duration_ms / 1000).toFixed(1)}s` : "";
+              line += `\n    🔊 audio${dur}: ${a.url}`;
+              line += a.transcript
+                ? `\n       transcript: "${String(a.transcript).slice(0, 400)}"`
+                : `\n       (no transcript — call transcribe(url) to read what was said)`;
+            } else if (a.type === "image") {
+              const dim = a.width && a.height ? ` ${a.width}×${a.height}` : "";
+              line += `\n    🖼 image${dim}: ${a.url}`;
+            } else {
+              line += `\n    📎 ${a.type || "file"}: ${a.url}`;
+            }
+          }
+          return line;
         }).join("\n");
         return { content: [{ type: "text", text: `${msgs.length} messages:\n${list}` }] };
       }