@vellumai/assistant 0.7.3 → 0.8.1

package/src/cli/commands/tts.ts

@@ -2,7 +2,7 @@
  * CLI command group: `assistant tts`
  *
  * Text-to-speech operations using the configured TTS provider.
- * Stateless, request-response commands — no daemon dependency.
+ * Thin IPC wrapper — delegates synthesis to the daemon over the IPC socket.
  */
 
 import { randomUUID } from "node:crypto";
@@ -10,28 +10,21 @@ import { existsSync, mkdirSync, readFileSync, writeFileSync } from "node:fs";
 import { tmpdir } from "node:os";
 import { dirname, join } from "node:path";
 
-import { Command } from "commander";
+import type { Command } from "commander";
 
-import { listCatalogProviderIds } from "../../tts/provider-catalog.js";
-import { registerBuiltinTtsProviders } from "../../tts/providers/register-builtins.js";
-import {
-  synthesizeText,
-  TtsSynthesisError,
-} from "../../tts/synthesize-text.js";
-import type { TtsUseCase } from "../../tts/types.js";
+import { cliIpcCall, exitFromIpcResult } from "../../ipc/cli-client.js";
+import { registerCommand } from "../lib/register-command.js";
 import { log } from "../logger.js";
 
 // ---------------------------------------------------------------------------
 // Constants
 // ---------------------------------------------------------------------------
 
-const VALID_USE_CASES: readonly TtsUseCase[] = [
-  "message-playback",
-  "phone-call",
-];
+const VALID_USE_CASES = ["message-playback", "phone-call"] as const;
+type TtsUseCaseCli = (typeof VALID_USE_CASES)[number];
 
 // ---------------------------------------------------------------------------
-// MIME type → file extension mapping
+// MIME type → file extension mapping (presentation-layer concern)
 // ---------------------------------------------------------------------------
 
 function extensionForMime(mimeType: string): string {
@@ -66,15 +59,16 @@ function extensionForMime(mimeType: string): string {
 // ---------------------------------------------------------------------------
 
 export function registerTtsCommand(program: Command): void {
-  const ttsCmd = program
-    .command("tts")
-    .description("Text-to-speech operations");
+  registerCommand(program, {
+    name: "tts",
+    transport: "ipc",
+    description: "Text-to-speech operations",
+    build: (ttsCmd) => {
+      const builtinProviders = "elevenlabs, fish-audio, deepgram, xai";
 
-  const builtinProviders = listCatalogProviderIds().join(", ");
-
-  ttsCmd.addHelpText(
-    "after",
-    `
+      ttsCmd.addHelpText(
+        "after",
+        `
 TTS commands use your configured TTS provider to synthesize text to audio.
 The provider is set via:
 
@@ -86,38 +80,40 @@ Examples:
   $ assistant tts synthesize --text "hello world"
   $ assistant tts synthesize "spoken sentence"
   $ echo "piped input" | assistant tts synthesize`,
-  );
+      );
 
-  // ── synthesize ──────────────────────────────────────────────────────
+      // ── synthesize ──────────────────────────────────────────────────────
 
-  ttsCmd
-    .command("synthesize")
-    .description("Synthesize text to audio using the configured TTS provider")
-    .option(
-      "--text <text>",
-      "Text to synthesize to audio (alternative: pass as positional arg or pipe via stdin)",
-    )
-    .option(
-      "--output <path>",
-      "Path to write the audio file (defaults to system temp dir with auto-generated name)",
-    )
-    .option(
-      "--voice <id>",
-      "Provider-specific voice identifier (ElevenLabs voiceId, Fish Audio referenceId, etc.) — overrides configured default",
-    )
-    .option(
-      "--use-case <case>",
-      "Synthesis use case: 'message-playback' (default, higher quality) or 'phone-call' (lower latency)",
-      "message-playback",
-    )
-    .option("--json", "Output structured JSON instead of plain file path")
-    .argument(
-      "[text...]",
-      "Text to synthesize (joined with spaces; alternative to --text or stdin)",
-    )
-    .addHelpText(
-      "after",
-      `
+      ttsCmd
+        .command("synthesize")
+        .description(
+          "Synthesize text to audio using the configured TTS provider",
+        )
+        .option(
+          "--text <text>",
+          "Text to synthesize to audio (alternative: pass as positional arg or pipe via stdin)",
+        )
+        .option(
+          "--output <path>",
+          "Path to write the audio file (defaults to system temp dir with auto-generated name)",
+        )
+        .option(
+          "--voice <id>",
+          "Provider-specific voice identifier (ElevenLabs voiceId, Fish Audio referenceId, etc.) — overrides configured default",
+        )
+        .option(
+          "--use-case <case>",
+          "Synthesis use case: 'message-playback' (default, higher quality) or 'phone-call' (lower latency)",
+          "message-playback",
+        )
+        .option("--json", "Output structured JSON instead of plain file path")
+        .argument(
+          "[text...]",
+          "Text to synthesize (joined with spaces; alternative to --text or stdin)",
+        )
+        .addHelpText(
+          "after",
+          `
 Input modes (pick one):
   --text <text>         Text to synthesize to audio.
   [text...]             Positional argument(s) joined with spaces.
@@ -145,129 +141,128 @@ Examples:
   $ assistant tts synthesize --text "hi" --voice <voice-id>
   $ assistant tts synthesize --text "hi" --use-case phone-call
   $ assistant tts synthesize --text "hi" --json`,
-    )
-    .action(
-      async (
-        positionalParts: string[],
-        opts: {
-          text?: string;
-          output?: string;
-          voice?: string;
-          useCase: string;
-          json?: boolean;
-        },
-      ) => {
-        const jsonOutput = opts.json ?? false;
+        )
+        .action(
+          async (
+            positionalParts: string[],
+            opts: {
+              text?: string;
+              output?: string;
+              voice?: string;
+              useCase: string;
+              json?: boolean;
+            },
+            cmd: Command,
+          ) => {
+            const jsonOutput = opts.json ?? false;
+
+            const emitError = (msg: string): void => {
+              if (jsonOutput) {
+                process.stdout.write(
+                  JSON.stringify({ ok: false, error: msg }) + "\n",
+                );
+              } else {
+                log.error(msg);
+              }
+            };
 
-        const emitError = (msg: string): void => {
-          if (jsonOutput) {
-            process.stdout.write(
-              JSON.stringify({ ok: false, error: msg }) + "\n",
-            );
-          } else {
-            log.error(msg);
-          }
-        };
+            // Resolve effective text from --text, positional args, or stdin.
+            let messageText =
+              opts.text ??
+              (positionalParts.length > 0 ? positionalParts.join(" ") : "");
+            if (!messageText && !process.stdin.isTTY) {
+              try {
+                messageText = readFileSync("/dev/stdin", "utf-8").trim();
+              } catch {
+                /* stdin unavailable */
+              }
+            }
+            if (!messageText) {
+              emitError(
+                "No text provided. Pass --text, a positional argument, or pipe via stdin.",
+              );
+              process.exitCode = 1;
+              return;
+            }
 
-        // Resolve effective text from --text, positional args, or stdin.
-        let messageText =
-          opts.text ??
-          (positionalParts.length > 0 ? positionalParts.join(" ") : "");
-        if (!messageText && !process.stdin.isTTY) {
-          try {
-            messageText = readFileSync("/dev/stdin", "utf-8").trim();
-          } catch {
-            /* stdin unavailable */
-          }
-        }
-        if (!messageText) {
-          emitError(
-            "No text provided. Pass --text, a positional argument, or pipe via stdin.",
-          );
-          process.exitCode = 1;
-          return;
-        }
+            // Validate --use-case
+            if (!VALID_USE_CASES.includes(opts.useCase as TtsUseCaseCli)) {
+              emitError(
+                `Invalid --use-case: '${opts.useCase}'. Must be one of: ${VALID_USE_CASES.join(", ")}.`,
+              );
+              process.exitCode = 1;
+              return;
+            }
+            const useCase = opts.useCase as TtsUseCaseCli;
 
-        // Validate --use-case
-        if (!VALID_USE_CASES.includes(opts.useCase as TtsUseCase)) {
-          emitError(
-            `Invalid --use-case: '${opts.useCase}'. Must be one of: ${VALID_USE_CASES.join(", ")}.`,
-          );
-          process.exitCode = 1;
-          return;
-        }
-        const useCase = opts.useCase as TtsUseCase;
+            // Call the daemon via IPC.
+            const r = await cliIpcCall<{
+              audioBase64: string;
+              contentType: string;
+            }>("tts_synthesize_cli", {
+              body: {
+                text: messageText,
+                useCase,
+                ...(opts.voice && { voiceId: opts.voice }),
+              },
+            });
 
-        // Providers must be registered in the CLI process since the daemon is
-        // a separate process and each process has its own registry.
-        registerBuiltinTtsProviders();
+            if (!r.ok) {
+              if (jsonOutput) {
+                emitError(r.error ?? "TTS synthesis failed");
+                process.exitCode = 1;
+                return;
+              }
+              return exitFromIpcResult(
+                { ok: false, error: r.error, statusCode: r.statusCode },
+                cmd,
+              );
+            }
 
-        // Synthesis step — any error here is a provider/synthesis failure.
-        // Billing has NOT yet occurred for the post-synthesis write path.
-        let result: Awaited<ReturnType<typeof synthesizeText>>;
-        try {
-          result = await synthesizeText({
-            text: messageText,
-            useCase,
-            voiceId: opts.voice,
-          });
-        } catch (err) {
-          if (
-            err instanceof TtsSynthesisError &&
-            err.code === "TTS_PROVIDER_NOT_CONFIGURED"
-          ) {
-            emitError(
-              `No TTS provider configured or registered. Run 'assistant config set services.tts.provider <provider>' to select one (e.g. ${listCatalogProviderIds().join(", ")}), then 'assistant keys set <provider> <key>' to add the API key.`,
-            );
-            process.exitCode = 1;
-            return;
-          }
+            const { audioBase64, contentType } = r.result!;
 
-          const msg = err instanceof Error ? err.message : String(err);
-          emitError(`TTS synthesis failed: ${msg}`);
-          process.exitCode = 1;
-          return;
-        }
+            // Decode base64 audio.
+            const audioBuffer = Buffer.from(audioBase64, "base64");
 
-        // Write step — synthesis has already succeeded (and been billed) by
-        // this point. Errors from mkdir/writeFile are filesystem failures,
-        // NOT synthesis failures, and must be reported as such so the user
-        // understands their audio was generated but could not be persisted.
-        const filePath =
-          opts.output ??
-          join(
-            tmpdir(),
-            `vellum-tts-${randomUUID()}.${extensionForMime(result.contentType)}`,
-          );
+            // Determine output file path.
+            const filePath =
+              opts.output ??
+              join(
+                tmpdir(),
+                `vellum-tts-${randomUUID()}.${extensionForMime(contentType)}`,
+              );
 
-        try {
-          const dir = dirname(filePath);
-          if (opts.output) {
-            mkdirSync(dir, { recursive: true });
-          } else if (!existsSync(dir)) {
-            mkdirSync(dir, { recursive: true });
-          }
+            // Write audio to disk.
+            try {
+              const dir = dirname(filePath);
+              if (opts.output) {
+                mkdirSync(dir, { recursive: true });
+              } else if (!existsSync(dir)) {
+                mkdirSync(dir, { recursive: true });
+              }
 
-          writeFileSync(filePath, result.audio);
-        } catch (err) {
-          const msg = err instanceof Error ? err.message : String(err);
-          emitError(`Failed to write audio to ${filePath}: ${msg}`);
-          process.exitCode = 1;
-          return;
-        }
+              writeFileSync(filePath, audioBuffer);
+            } catch (err) {
+              const msg = err instanceof Error ? err.message : String(err);
+              emitError(`Failed to write audio to ${filePath}: ${msg}`);
+              process.exitCode = 1;
+              return;
+            }
 
-        if (jsonOutput) {
-          process.stdout.write(
-            JSON.stringify({
-              ok: true,
-              path: filePath,
-              contentType: result.contentType,
-              sizeBytes: result.audio.length,
-            }) + "\n",
-          );
-        } else {
-          process.stdout.write(filePath + "\n");
-        }
-      },
-    );
+            if (jsonOutput) {
+              process.stdout.write(
+                JSON.stringify({
+                  ok: true,
+                  path: filePath,
+                  contentType,
+                  sizeBytes: audioBuffer.length,
+                }) + "\n",
+              );
+            } else {
+              process.stdout.write(filePath + "\n");
+            }
+          },
+        );
+    },
+  });
 }

package/src/cli/commands/ui.ts

@@ -17,12 +17,13 @@ import { readFileSync } from "node:fs";
 import type { Command } from "commander";
 
 import { cliIpcCall } from "../../ipc/cli-client.js";
-import {
-  type InteractiveUiAction,
-  type InteractiveUiResult,
-  RESERVED_ACTION_IDS,
+import type {
+  InteractiveUiAction,
+  InteractiveUiResult,
 } from "../../runtime/interactive-ui-types.js";
+import { registerCommand } from "../lib/register-command.js";
 import { log } from "../logger.js";
+import { resolveConversationId } from "../utils/conversation-id.js";
 
 // ── Constants ─────────────────────────────────────────────────────────
 
@@ -40,41 +41,24 @@ const DEFAULT_REQUEST_TIMEOUT_MS = 5 * 60 * 1000; // 5m
  */
 const IPC_TIMEOUT_BUFFER_MS = 10_000; // 10s
 
-// ── Conversation ID resolution ────────────────────────────────────────
+const CONV_ID_HELP =
+  "No conversation ID available.\n" +
+  "Provide --conversation-id explicitly (run 'assistant conversations list' to find it),\n" +
+  "or run this command from a skill or bash tool context.";
 
 /**
- * Resolve the conversation ID by precedence:
- *   1. Explicit `--conversation-id` flag
- *   2. `__SKILL_CONTEXT_JSON` env var (set by skill sandbox runner)
- *   3. `__CONVERSATION_ID` env var (set by bash tool subprocess)
- *   4. Fail with an actionable error
+ * Action IDs reserved for internal use. Inlined from
+ * `interactive-ui-types.ts` to avoid a runtime import from daemon
+ * internals (the ESLint `cli/no-daemon-internals` rule forbids it for
+ * `ipc`-tagged commands).
  */
-function resolveConversationId(explicit?: string): string {
-  if (explicit) return explicit;
-
-  const contextJson = process.env.__SKILL_CONTEXT_JSON;
-  if (contextJson) {
-    try {
-      const parsed = JSON.parse(contextJson);
-      if (parsed.conversationId && typeof parsed.conversationId === "string") {
-        return parsed.conversationId;
-      }
-    } catch {
-      // Fall through
-    }
-  }
-
-  const envConvId = process.env.__CONVERSATION_ID;
-  if (envConvId && typeof envConvId === "string") {
-    return envConvId;
-  }
-
-  throw new Error(
-    "No conversation ID available.\n" +
-      "Provide --conversation-id explicitly (run 'assistant conversations list' to find it),\n" +
-      "or run this command from a skill or bash tool context.",
-  );
-}
+const RESERVED_ACTION_IDS = new Set([
+  "selection_changed",
+  "content_changed",
+  "state_update",
+  "cancel",
+  "dismiss",
+]);
 
 // ── Payload parsing ───────────────────────────────────────────────────
 
@@ -261,10 +245,11 @@ function parseStrictPositiveInt(value: string): number {
 // ── Registration ──────────────────────────────────────────────────────
 
 export function registerUiCommand(program: Command): void {
-  const ui = program
-    .command("ui")
-    .description("Present interactive UI surfaces to the user");
-
+  registerCommand(program, {
+    name: "ui",
+    transport: "ipc",
+    description: "Present interactive UI surfaces to the user",
+    build: (ui) => {
   ui.addHelpText(
     "after",
     `
@@ -372,7 +357,10 @@ Examples:
         // Resolve conversation ID
         let conversationId: string;
         try {
-          conversationId = resolveConversationId(opts.conversationId);
+          conversationId = resolveConversationId({
+            explicit: opts.conversationId,
+            failureHelp: CONV_ID_HELP,
+          });
         } catch (err) {
           const msg = err instanceof Error ? err.message : String(err);
           if (opts.json) {
@@ -547,7 +535,10 @@ Examples:
         // Resolve conversation ID
         let conversationId: string;
         try {
-          conversationId = resolveConversationId(opts.conversationId);
+          conversationId = resolveConversationId({
+            explicit: opts.conversationId,
+            failureHelp: CONV_ID_HELP,
+          });
         } catch (err) {
           const msg = err instanceof Error ? err.message : String(err);
           if (opts.json) {
@@ -667,4 +658,6 @@ Examples:
         }
       },
     );
+    },
+  });
 }