feat(audio): add audio.transcribe tool with Whisper-compatible API support

- Add createAudioTranscribeTool with OpenAI/Groq/Ollama/llama.cpp provider support
- Refactor audio config schema to nested audio.enabled + audio.provider structure
- Move audio tool registration to initTools() for conditional enablement
- Fix duplication bug in audio-transcribe.ts URL download handler
- Support base64 data and URL-based audio input with format detection

src/daemon/index.ts

@@ -5,7 +5,6 @@ import { mkdirSync } from 'fs';
 
 // ── Config & Types ──
 import type { Config } from '../config/index.js';
-import type { AudioTranscriptionConfig } from '../models/media.js';
 import type { ToolRegistry, ToolExecutor, BrowserManager } from '../tools/index.js';
 import type { AgentConfigRegistry, AgentRouter } from '../agents/index.js';
 import type { SandboxManager } from '../sandbox/index.js';
@@ -100,12 +99,6 @@ export async function startDaemon(config: Config): Promise<DaemonContext> {
   const { skillRegistry, skillInstaller } = initSkills(config);
   const { agentConfigRegistry, agentRouter, sandboxManager } = await initAgents({ config, lifecycle });
 
-  // ── Model & Prompt ──
-  const audioConfig: AudioTranscriptionConfig = {
-    endpoint: config.audio.transcription_endpoint,
-    apiKey: config.audio.transcription_api_key,
-    model: config.audio.transcription_model,
-  };
   const modelRouter = createModelRouter(config);
 
   // Restore persisted model tier
@@ -133,7 +126,7 @@ export async function startDaemon(config: Config): Promise<DaemonContext> {
 
   const messageRouter = createMessageRouter({
     sessionManager, modelRouter, systemPrompt, toolRegistry, toolExecutor,
-    config, memoryStore, agentConfigRegistry, agentRouter, sandboxManager, audioConfig,
+    config, memoryStore, agentConfigRegistry, agentRouter, sandboxManager,
   });
   channelRegistry.setMessageHandler(messageRouter.handler);
   channelAgents = messageRouter.agents;