@0xmaxma/claude-gateway 2.0.11 → 2.0.13

package/mcp/tools/narrate/module.ts

@@ -0,0 +1,42 @@
+import type { ToolModule, McpToolDefinition, McpToolResult, ToolVisibility } from '../../types';
+import { callTaskBridge } from '../tasks/module';
+
+/**
+ * Long-form narration (worker only). TTS credentials live in the gateway, so this
+ * module just forwards to the task bridge; the gateway splits the text, synthesizes
+ * it piece by piece and stages the resulting audio files for the agent's reply.
+ */
+export const NARRATE_TOOL: McpToolDefinition = {
+  name: 'narrate',
+  description:
+    'Narrate text to speech: read text aloud to the user as a sequence of voice/audio messages (TTS). Use this when the user asks to HEAR a page, post, document or file — ' +
+    'pass the full text (or a file path inside this task) and every character is spoken, in order. ' +
+    'Pass exactly one of `text` or `path`. Do NOT summarize or shorten the text you pass unless the user asked for a summary; ' +
+    'the tool never rewrites it. This tool cannot fetch URLs: fetch the page yourself and pass its text. ' +
+    'The audio files are staged automatically (do not call task_stage_file for them). ' +
+    'Afterwards finish the task with ONE short sentence saying how many parts were prepared (your final reply must be non-empty so the audio attaches); ' +
+    'do not repeat the content. If the result has `ok:false`, report `stopped_reason` and how many parts were staged. ' +
+    'Host workers only: container workers cannot use media tools.',
+  inputSchema: {
+    type: 'object',
+    properties: {
+      text: { type: 'string', description: 'The exact text to read aloud.' },
+      path: { type: 'string', description: 'Absolute path of a UTF-8 text file inside this task scope or an attachment.' },
+    },
+    additionalProperties: false,
+  },
+};
+
+export class NarrateModule implements ToolModule {
+  id = 'narrate';
+  toolVisibility: ToolVisibility = 'all-configured';
+
+  isEnabled(): boolean {
+    return process.env.GATEWAY_ORCHESTRATION_ROLE === 'worker' && process.env.GATEWAY_ORCHESTRATION_MEDIA === 'true' && Boolean(process.env.GATEWAY_ORCHESTRATION_TICKET_FILE);
+  }
+  getTools(): McpToolDefinition[] { return [NARRATE_TOOL]; }
+  handleTool(name: string, args: Record<string, unknown>, signal?: AbortSignal): Promise<McpToolResult> {
+    if (name !== 'narrate') return Promise.resolve({ content: [{ type: 'text', text: `unknown tool: ${name}` }], isError: true });
+    return callTaskBridge(name, args, crypto.randomUUID(), signal ?? new AbortController().signal);
+  }
+}

package/mcp/tools/tasks/module.ts

@@ -15,13 +15,18 @@ export const WORKER_REPORT_TOOLS: McpToolDefinition[] = [
   { name: 'task_request_input', description: 'Ask the user for information required by your task, then end this turn immediately.', inputSchema: schema({ question: text }, ['question']) },
 ];
 
+const LONG_RUNNING_TOOLS = new Set(['narrate']);
+export const LONG_RUNNING_MAX_MS = 2 * 60 * 60 * 1000;
 export async function callTaskBridge(tool: string, args: Record<string, unknown>, requestId: string, signal: AbortSignal): Promise<McpToolResult> {
   // Rotated by the gateway at each decision/attempt boundary. A loaded token is
   // immutable for this request; late requests cannot borrow the next decision.
   const { readFile } = await import('node:fs/promises');
   const scope = JSON.parse(await readFile(process.env.GATEWAY_ORCHESTRATION_TICKET_FILE!, 'utf8')) as { url: string; token: string };
-  const response = await fetch(scope.url, { method: 'POST', signal, headers: { 'Content-Type': 'application/json', Authorization: `Bearer ${scope.token}` },
-    body: JSON.stringify({ tool, args, action_id: requestId }) });
+  // Bun's fetch gives up after 300s without a response. Narration answers only when every part is spoken
+  // (up to maxParts x partTimeoutMs), so that call alone has no idle timeout, bounded by an overall maximum.
+  const long = LONG_RUNNING_TOOLS.has(tool);
+  const response = await fetch(scope.url, { method: 'POST', signal: long ? AbortSignal.any([signal, AbortSignal.timeout(LONG_RUNNING_MAX_MS)]) : signal, ...(long ? { timeout: false } : {}), headers: { 'Content-Type': 'application/json', Authorization: `Bearer ${scope.token}` },
+    body: JSON.stringify({ tool, args, action_id: requestId }) } as RequestInit);
   const body = await response.text();
   if(response.ok && tool==='task_status') {
     const value=JSON.parse(body),image=value.screenshot;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@0xmaxma/claude-gateway",
-  "version": "2.0.11",
+  "version": "2.0.13",
   "description": "Multi-agent gateway for Claude",
   "repository": {
     "type": "git",

package/scripts/orchestration/smoke-safemode-codex.cjs

@@ -53,7 +53,7 @@ const { buildNativeInvocation } = require('../../dist/safemode/native');
       const id = events.find(event => event.type === 'thread.started')?.thread_id;
       assert(id); if (nativeSessionId) assert.equal(id, nativeSessionId);
       await new Promise(resolve => setTimeout(resolve, 200));
-      assert.equal(fs.existsSync(marker), false, 'notify executed outside the read-only workspace');
+      assert.equal(fs.existsSync(marker), false, 'notify executed outside the investigation workspace');
       return id;
     }
     await run(await run());