refactor(server): extract SpeechService and defer init until after listen

Speech runtime was leaking 8+ individual resolvers into bootstrap and
blocking server startup for ~3s with synchronous Sherpa native model
loading. Refactor into a self-contained SpeechService that owns its
full lifecycle (init, download, monitor, cleanup) and defer start()
until after httpServer.listen() so native loading doesn't block
accepting connections. Server startup drops from ~3.2s to ~0.5s.

Also removes unused client-triggered speech model download/list path
(CLI commands, session handlers, message types) — the service manages
its own models.

packages/cli/src/commands/speech/download.ts

@@ -1,63 +0,0 @@
-import type { Command } from "commander";
-import type { CommandError, CommandOptions, ListResult, OutputSchema } from "../../output/index.js";
-import { connectToDaemon } from "../../utils/client.js";
-
-interface SpeechDownloadRow {
-  modelId: string;
-  status: "downloaded";
-}
-
-const speechDownloadSchema: OutputSchema<SpeechDownloadRow> = {
-  idField: "modelId",
-  columns: [
-    { header: "MODEL", field: "modelId", width: 36 },
-    { header: "STATUS", field: "status", width: 12, color: () => "green" },
-  ],
-};
-
-export type SpeechDownloadResult = ListResult<SpeechDownloadRow>;
-
-export interface SpeechDownloadOptions extends CommandOptions {
-  host?: string;
-  model?: string[];
-}
-
-export async function runSpeechDownloadCommand(
-  options: SpeechDownloadOptions,
-  _command: Command,
-): Promise<SpeechDownloadResult> {
-  const client = await connectToDaemon({ host: options.host });
-  try {
-    const response = await client.downloadSpeechModels({
-      modelIds: options.model && options.model.length > 0 ? options.model : undefined,
-    });
-    if (response.error) {
-      const commandError: CommandError = {
-        code: "SPEECH_MODELS_DOWNLOAD_FAILED",
-        message: response.error,
-      };
-      throw commandError;
-    }
-
-    return {
-      type: "list",
-      data: response.downloadedModelIds.map((modelId) => ({
-        modelId,
-        status: "downloaded" as const,
-      })),
-      schema: speechDownloadSchema,
-    };
-  } catch (error) {
-    if (typeof error === "object" && error && "code" in error && "message" in error) {
-      throw error;
-    }
-    const message = error instanceof Error ? error.message : String(error);
-    const commandError: CommandError = {
-      code: "SPEECH_MODELS_DOWNLOAD_FAILED",
-      message: `Failed to download speech models: ${message}`,
-    };
-    throw commandError;
-  } finally {
-    await client.close().catch(() => {});
-  }
-}

packages/cli/src/commands/speech/index.ts

@@ -1,22 +1,5 @@
 import { Command } from "commander";
-import { withOutput } from "../../output/index.js";
-import { runSpeechModelsCommand } from "./models.js";
-import { runSpeechDownloadCommand } from "./download.js";
-import { addJsonAndDaemonHostOptions, collectMultiple } from "../../utils/command-options.js";
 
 export function createSpeechCommand(): Command {
-  const speech = new Command("speech").description("Manage local speech models");
-
-  addJsonAndDaemonHostOptions(
-    speech.command("models").description("List local speech model download status"),
-  ).action(withOutput(runSpeechModelsCommand));
-
-  addJsonAndDaemonHostOptions(
-    speech
-      .command("download")
-      .description("Download local speech models")
-      .option("--model <id>", "Model ID to download (repeatable)", collectMultiple, []),
-  ).action(withOutput(runSpeechDownloadCommand));
-
-  return speech;
+  return new Command("speech").description("Speech commands");
 }

packages/cli/src/commands/speech/models.ts

@@ -1,67 +0,0 @@
-import type { Command } from "commander";
-import type { CommandError, CommandOptions, ListResult, OutputSchema } from "../../output/index.js";
-import { connectToDaemon } from "../../utils/client.js";
-
-interface SpeechModelListItem {
-  id: string;
-  kind: string;
-  status: "downloaded" | "missing";
-  modelDir: string;
-  missingFiles: string;
-}
-
-const speechModelsSchema: OutputSchema<SpeechModelListItem> = {
-  idField: "id",
-  columns: [
-    { header: "MODEL", field: "id", width: 36 },
-    { header: "KIND", field: "kind", width: 12 },
-    {
-      header: "STATUS",
-      field: "status",
-      width: 12,
-      color: (value) => (value === "downloaded" ? "green" : "yellow"),
-    },
-    { header: "MODEL DIR", field: "modelDir", width: 44 },
-    { header: "MISSING FILES", field: "missingFiles", width: 40 },
-  ],
-};
-
-export type SpeechModelsResult = ListResult<SpeechModelListItem>;
-
-export interface SpeechModelsOptions extends CommandOptions {
-  host?: string;
-}
-
-export async function runSpeechModelsCommand(
-  options: SpeechModelsOptions,
-  _command: Command,
-): Promise<SpeechModelsResult> {
-  const client = await connectToDaemon({ host: options.host });
-  try {
-    const response = await client.listSpeechModels();
-    const rows: SpeechModelListItem[] = response.models
-      .slice()
-      .sort((a, b) => a.kind.localeCompare(b.kind) || a.id.localeCompare(b.id))
-      .map((model) => ({
-        id: model.id,
-        kind: model.kind,
-        status: model.isDownloaded ? "downloaded" : "missing",
-        modelDir: model.modelDir,
-        missingFiles: model.missingFiles?.join(", ") ?? "",
-      }));
-    return {
-      type: "list",
-      data: rows,
-      schema: speechModelsSchema,
-    };
-  } catch (error) {
-    const message = error instanceof Error ? error.message : String(error);
-    const commandError: CommandError = {
-      code: "SPEECH_MODELS_LIST_FAILED",
-      message: `Failed to list speech models: ${message}`,
-    };
-    throw commandError;
-  } finally {
-    await client.close().catch(() => {});
-  }
-}