@llmist/cli 15.19.0 → 16.0.1

package/dist/cli.js

@@ -110,7 +110,7 @@ import { Command, InvalidArgumentError as InvalidArgumentError2 } from "commande
 // package.json
 var package_default = {
   name: "@llmist/cli",
-  version: "15.19.0",
+  version: "16.0.1",
   description: "CLI for llmist - run LLM agents from the command line",
   type: "module",
   main: "dist/cli.js",
@@ -136,7 +136,8 @@ var package_default = {
     build: "tsup",
     typecheck: "tsc --noEmit",
     clean: "rimraf dist",
-    postinstall: "node scripts/postinstall.js"
+    postinstall: "node scripts/postinstall.js",
+    dev: "npx --silent tsx src/cli.ts"
   },
   homepage: "https://llmist.dev",
   repository: {
@@ -166,7 +167,7 @@ var package_default = {
     node: ">=22.0.0"
   },
   dependencies: {
-    llmist: "^15.19.0",
+    llmist: "^16.0.1",
     "@unblessed/node": "^1.0.0-alpha.23",
     chalk: "^5.6.2",
     commander: "^12.1.0",
@@ -180,7 +181,7 @@ var package_default = {
     zod: "^4.1.12"
   },
   devDependencies: {
-    "@llmist/testing": "^15.19.0",
+    "@llmist/testing": "^16.0.1",
     "@types/diff": "^8.0.0",
     "@types/js-yaml": "^4.0.9",
     "@types/marked-terminal": "^6.1.1",
@@ -195,13 +196,194 @@ var package_default = {
 import { AgentBuilder, GadgetRegistry, HookPresets, isAbortError, text } from "llmist";
 
 // src/builtin-gadgets.ts
-import { createGadget, HumanInputRequiredException, TaskCompletionSignal } from "llmist";
+import { createGadget as createGadget2, HumanInputRequiredException, TaskCompletionSignal } from "llmist";
+import { z as z2 } from "zod";
+
+// src/builtins/text-to-speech.ts
+import { createGadget, getErrorMessage, resultWithAudio } from "llmist";
 import { z } from "zod";
-var askUser = createGadget({
+
+// src/ffmpeg.ts
+import { spawn } from "child_process";
+var CONVERSION_TIMEOUT_MS = 3e4;
+var ffmpegCheckPromise = null;
+async function isFFmpegAvailable() {
+  if (ffmpegCheckPromise !== null) return ffmpegCheckPromise;
+  ffmpegCheckPromise = new Promise((resolve3) => {
+    const proc = spawn("ffmpeg", ["-version"], { stdio: "ignore" });
+    proc.on("error", () => resolve3(false));
+    proc.on("close", (code) => resolve3(code === 0));
+  });
+  return ffmpegCheckPromise;
+}
+async function convertToMp3(input, inputFormat, timeout = CONVERSION_TIMEOUT_MS) {
+  return new Promise((resolve3) => {
+    let timeoutId;
+    const inputArgs = inputFormat === "pcm16" ? ["-f", "s16le", "-ar", "24000", "-ac", "1"] : ["-f", inputFormat];
+    const proc = spawn(
+      "ffmpeg",
+      [
+        ...inputArgs,
+        "-i",
+        "pipe:0",
+        // Read from stdin
+        "-f",
+        "mp3",
+        // Output format
+        "-ab",
+        "128k",
+        // Bitrate
+        "pipe:1"
+        // Write to stdout
+      ],
+      { stdio: ["pipe", "pipe", "ignore"] }
+    );
+    const chunks = [];
+    proc.stdout.on("data", (chunk) => chunks.push(chunk));
+    proc.on("error", () => {
+      clearTimeout(timeoutId);
+      resolve3(null);
+    });
+    proc.on("close", (code) => {
+      clearTimeout(timeoutId);
+      resolve3(code === 0 ? Buffer.concat(chunks) : null);
+    });
+    timeoutId = setTimeout(() => {
+      proc.kill();
+      resolve3(null);
+    }, timeout);
+    proc.stdin.on("error", () => {
+    });
+    proc.stdin.write(input);
+    proc.stdin.end();
+  });
+}
+
+// src/builtins/text-to-speech.ts
+var TTS_VOICES = ["alloy", "echo", "fable", "onyx", "nova", "shimmer"];
+var TTS_FORMATS = ["mp3", "opus", "aac", "flac", "wav", "pcm16"];
+function createTextToSpeech(config) {
+  if (config?.voice && !TTS_VOICES.includes(config.voice)) {
+    throw new Error(`Invalid TTS voice "${config.voice}". Valid voices: ${TTS_VOICES.join(", ")}`);
+  }
+  if (config?.format && !TTS_FORMATS.includes(config.format)) {
+    throw new Error(
+      `Invalid TTS format "${config.format}". Valid formats: ${TTS_FORMATS.join(", ")}`
+    );
+  }
+  if (config?.speed !== void 0 && (config.speed < 0.25 || config.speed > 4)) {
+    throw new Error(`Invalid TTS speed "${config.speed}". Must be between 0.25 and 4.0`);
+  }
+  const defaultModel = config?.model ?? "tts-1";
+  const defaultVoice = config?.voice ?? "nova";
+  const defaultFormat = config?.format ?? "mp3";
+  const defaultSpeed = config?.speed ?? 1;
+  return createGadget({
+    name: "TextToSpeech",
+    description: `Convert text to speech audio. Uses configured TTS model. Defaults: ${defaultVoice} voice, ${defaultFormat} format.`,
+    schema: z.object({
+      text: z.string().min(1).describe("Text to convert to speech"),
+      voice: z.enum(TTS_VOICES).optional().describe(`Voice to use (default: ${defaultVoice})`),
+      format: z.enum(TTS_FORMATS).optional().describe(`Output format (default: ${defaultFormat})`),
+      speed: z.number().min(0.25).max(4).optional().describe(`Speech speed 0.25-4.0 (default: ${defaultSpeed})`)
+    }),
+    examples: [
+      {
+        comment: "Generate speech with default settings",
+        params: { text: "Hello, welcome to our application!" },
+        output: "Generated audio (mp3, 35 chars, $0.000525)"
+      },
+      {
+        comment: "Use a specific voice",
+        params: {
+          text: "This is an important announcement.",
+          voice: "onyx"
+        },
+        output: "Generated audio (mp3, 35 chars, $0.001050)"
+      },
+      {
+        comment: "Generate slower speech in WAV format",
+        params: {
+          text: "Please listen carefully to these instructions.",
+          format: "wav",
+          speed: 0.8
+        },
+        output: "Generated audio (wav, 46 chars, $0.000690)"
+      },
+      {
+        comment: "Result when cost is unavailable (some providers)",
+        params: { text: "Hello there!" },
+        output: "Generated audio (mp3, 12 chars, $N/A)"
+      },
+      {
+        comment: "Auto-converted to MP3 (when ffmpeg is available and format isn't MP3)",
+        params: { text: "Converted audio.", format: "pcm16" },
+        output: "Generated audio (mp3, 16 chars, $0.000240) [converted from pcm16]"
+      }
+    ],
+    execute: async ({ text: text3, voice, format, speed }, ctx) => {
+      if (!ctx?.llmist?.speech?.generate) {
+        return "status=1\n\nerror: Speech generation requires LLMist client with speech capability.";
+      }
+      const selectedModel = defaultModel;
+      const selectedVoice = voice ?? defaultVoice;
+      const selectedFormat = format ?? defaultFormat;
+      const selectedSpeed = speed ?? (defaultSpeed !== 1 ? defaultSpeed : void 0);
+      const maxRetries = 2;
+      let lastError;
+      for (let attempt = 0; attempt <= maxRetries; attempt++) {
+        try {
+          const result = await ctx.llmist.speech.generate({
+            model: selectedModel,
+            input: text3,
+            voice: selectedVoice,
+            // Cast is safe: Zod validates LLM input, config values validated at factory time
+            responseFormat: selectedFormat,
+            ...selectedSpeed !== void 0 ? { speed: selectedSpeed } : {}
+          });
+          let audioBuffer = Buffer.from(result.audio);
+          let finalFormat = result.format;
+          if (finalFormat !== "mp3" && await isFFmpegAvailable()) {
+            const mp3Buffer = await convertToMp3(audioBuffer, finalFormat);
+            if (mp3Buffer) {
+              audioBuffer = mp3Buffer;
+              finalFormat = "mp3";
+            }
+          }
+          const conversionNote = finalFormat !== result.format ? ` [converted from ${result.format}]` : "";
+          return resultWithAudio(
+            `Generated audio (${finalFormat}, ${result.usage.characterCount} chars, $${result.cost?.toFixed(6) ?? "N/A"})${conversionNote}`,
+            audioBuffer,
+            {
+              mimeType: `audio/${finalFormat}`,
+              cost: result.cost,
+              description: `TTS: "${text3.slice(0, 50)}${text3.length > 50 ? "..." : ""}"`
+            }
+          );
+        } catch (error) {
+          lastError = error;
+          const errorMsg = getErrorMessage(error);
+          const isRetryable = errorMsg.includes("HTTP 4") || errorMsg.includes("HTTP 5");
+          if (!isRetryable || attempt === maxRetries) {
+            break;
+          }
+          await new Promise((r) => setTimeout(r, 500 * (attempt + 1)));
+        }
+      }
+      return `status=1
+
+error: ${getErrorMessage(lastError)}`;
+    }
+  });
+}
+var textToSpeech = createTextToSpeech();
+
+// src/builtin-gadgets.ts
+var askUser = createGadget2({
   name: "AskUser",
   description: "Ask the user a question when you need more information or clarification. The user's response will be provided back to you.",
-  schema: z.object({
-    question: z.string().describe("The question to ask the user in plain-text or Markdown")
+  schema: z2.object({
+    question: z2.string().describe("The question to ask the user in plain-text or Markdown")
   }),
   examples: [
     {
@@ -219,12 +401,12 @@ var askUser = createGadget({
     throw new HumanInputRequiredException(question);
   }
 });
-var tellUser = createGadget({
+var tellUser = createGadget2({
   name: "TellUser",
   description: "Tell the user something important.",
-  schema: z.object({
-    message: z.string().optional().describe("The message to display to the user in Markdown"),
-    type: z.enum(["info", "success", "warning", "error"]).default("info").describe("Message type: info, success, warning, or error")
+  schema: z2.object({
+    message: z2.string().optional().describe("The message to display to the user in Markdown"),
+    type: z2.enum(["info", "success", "warning", "error"]).default("info").describe("Message type: info, success, warning, or error")
   }),
   examples: [
     {
@@ -255,10 +437,10 @@ var tellUser = createGadget({
     return prefixes[type] + message;
   }
 });
-var finish = createGadget({
+var finish = createGadget2({
   name: "Finish",
   description: "Signal that you have completed your task. Call this when your work is done.",
-  schema: z.object({}),
+  schema: z2.object({}),
   examples: [
     {
       comment: "Signal task completion",
@@ -269,7 +451,10 @@ var finish = createGadget({
     throw new TaskCompletionSignal("Task completed");
   }
 });
-var builtinGadgets = [askUser, tellUser, finish];
+var builtinGadgets = [askUser, tellUser, finish, createTextToSpeech()];
+function getBuiltinGadgets(speechConfig) {
+  return [askUser, tellUser, finish, createTextToSpeech(speechConfig)];
+}
 
 // src/config.ts
 import { existsSync as existsSync2, readFileSync as readFileSync2 } from "fs";
@@ -1530,8 +1715,8 @@ import { AbstractGadget } from "llmist";
 
 // src/builtins/filesystem/edit-file.ts
 import { readFileSync as readFileSync3, writeFileSync } from "fs";
-import { createGadget as createGadget2 } from "llmist";
-import { z as z2 } from "zod";
+import { createGadget as createGadget3 } from "llmist";
+import { z as z3 } from "zod";
 
 // src/builtins/filesystem/editfile/matcher.ts
 var DEFAULT_OPTIONS = {
@@ -1844,7 +2029,7 @@ function formatFailure(filePath, search, failure, fileContent) {
   lines.push("", "CURRENT FILE CONTENT:", "```", fileContent, "```");
   return lines.join("\n");
 }
-var editFile = createGadget2({
+var editFile = createGadget3({
   name: "EditFile",
   description: `Edit a file by searching for content and replacing it.
 
@@ -1858,10 +2043,10 @@ For multiple edits to the same file, call this gadget multiple times.
 Each call provides immediate feedback, allowing you to adjust subsequent edits.`,
   maxConcurrent: 1,
   // Sequential execution to prevent race conditions
-  schema: z2.object({
-    filePath: z2.string().describe("Path to the file to edit (relative or absolute)"),
-    search: z2.string().describe("The content to search for in the file"),
-    replace: z2.string().describe("The content to replace it with (empty string to delete)")
+  schema: z3.object({
+    filePath: z3.string().describe("Path to the file to edit (relative or absolute)"),
+    search: z3.string().describe("The content to search for in the file"),
+    replace: z3.string().describe("The content to replace it with (empty string to delete)")
   }),
   examples: [
     {
@@ -1955,8 +2140,8 @@ ${newContent}
 // src/builtins/filesystem/list-directory.ts
 import fs2 from "fs";
 import path2 from "path";
-import { createGadget as createGadget3 } from "llmist";
-import { z as z3 } from "zod";
+import { createGadget as createGadget4 } from "llmist";
+import { z as z4 } from "zod";
 function listFiles(dirPath, basePath = dirPath, maxDepth = 1, currentDepth = 1) {
   const entries = [];
   try {
@@ -2040,12 +2225,12 @@ function formatEntriesAsString(entries) {
   );
   return [header, ...rows].join("\n");
 }
-var listDirectory = createGadget3({
+var listDirectory = createGadget4({
   name: "ListDirectory",
   description: "List files and directories in a directory with full details (names, types, sizes, modification dates). Use maxDepth to explore subdirectories recursively. The directory path must be within the current working directory or its subdirectories.",
-  schema: z3.object({
-    directoryPath: z3.string().default(".").describe("Path to the directory to list"),
-    maxDepth: z3.number().int().min(1).max(10).default(3).describe(
+  schema: z4.object({
+    directoryPath: z4.string().default(".").describe("Path to the directory to list"),
+    maxDepth: z4.number().int().min(1).max(10).default(3).describe(
       "Maximum depth to recurse (1 = immediate children only, 2 = include grandchildren, etc.)"
     )
   }),
@@ -2077,13 +2262,13 @@ ${formattedList}`;
 
 // src/builtins/filesystem/read-file.ts
 import fs3 from "fs";
-import { createGadget as createGadget4 } from "llmist";
-import { z as z4 } from "zod";
-var readFile2 = createGadget4({
+import { createGadget as createGadget5 } from "llmist";
+import { z as z5 } from "zod";
+var readFile2 = createGadget5({
   name: "ReadFile",
   description: "Read the entire content of a file and return it as text. The file path must be within the current working directory or its subdirectories.",
-  schema: z4.object({
-    filePath: z4.string().describe("Path to the file to read (relative or absolute)")
+  schema: z5.object({
+    filePath: z5.string().describe("Path to the file to read (relative or absolute)")
   }),
   examples: [
     {
@@ -2109,16 +2294,16 @@ ${content}`;
 // src/builtins/filesystem/write-file.ts
 import fs4 from "fs";
 import path3 from "path";
-import { createGadget as createGadget5 } from "llmist";
-import { z as z5 } from "zod";
-var writeFile = createGadget5({
+import { createGadget as createGadget6 } from "llmist";
+import { z as z6 } from "zod";
+var writeFile = createGadget6({
   name: "WriteFile",
   description: "Write content to a file. Creates parent directories if needed. Overwrites existing files. The file path must be within the current working directory or its subdirectories.",
   maxConcurrent: 1,
   // Sequential execution to prevent race conditions
-  schema: z5.object({
-    filePath: z5.string().describe("Path to the file to write (relative or absolute)"),
-    content: z5.string().describe("Content to write to the file")
+  schema: z6.object({
+    filePath: z6.string().describe("Path to the file to write (relative or absolute)"),
+    content: z6.string().describe("Content to write to the file")
   }),
   examples: [
     {
@@ -2163,8 +2348,8 @@ Wrote ${bytesWritten} bytes${dirNote}`;
 });
 
 // src/builtins/run-command.ts
-import { createGadget as createGadget6 } from "llmist";
-import { z as z6 } from "zod";
+import { createGadget as createGadget7 } from "llmist";
+import { z as z7 } from "zod";
 
 // src/spawn.ts
 import { spawn as nodeSpawn } from "child_process";
@@ -2187,7 +2372,7 @@ function nodeStreamToReadableStream(nodeStream) {
     }
   });
 }
-function spawn(argv, options = {}) {
+function spawn2(argv, options = {}) {
   const [command, ...args] = argv;
   const proc = nodeSpawn(command, args, {
     cwd: options.cwd,
@@ -2225,13 +2410,13 @@ function spawn(argv, options = {}) {
 }
 
 // src/builtins/run-command.ts
-var runCommand = createGadget6({
+var runCommand = createGadget7({
   name: "RunCommand",
   description: "Execute a command with arguments and return its output. Uses argv array to bypass shell - arguments are passed directly without interpretation. Returns stdout/stderr combined with exit status.",
-  schema: z6.object({
-    argv: z6.array(z6.string()).describe("Command and arguments as array (e.g., ['git', 'commit', '-m', 'message'])"),
-    cwd: z6.string().optional().describe("Working directory for the command (default: current directory)"),
-    timeout: z6.number().default(3e4).describe("Timeout in milliseconds (default: 30000)")
+  schema: z7.object({
+    argv: z7.array(z7.string()).describe("Command and arguments as array (e.g., ['git', 'commit', '-m', 'message'])"),
+    cwd: z7.string().optional().describe("Working directory for the command (default: current directory)"),
+    timeout: z7.number().default(3e4).describe("Timeout in milliseconds (default: 30000)")
   }),
   examples: [
     {
@@ -2294,7 +2479,7 @@ var runCommand = createGadget6({
     }
     let timeoutId;
     try {
-      const proc = spawn(argv, {
+      const proc = spawn2(argv, {
         cwd: workingDir,
         stdout: "pipe",
         stderr: "pipe"
@@ -2338,7 +2523,8 @@ var builtinGadgetRegistry = {
   ReadFile: readFile2,
   WriteFile: writeFile,
   EditFile: editFile,
-  RunCommand: runCommand
+  RunCommand: runCommand,
+  TextToSpeech: textToSpeech
 };
 function getBuiltinGadget(name) {
   return builtinGadgetRegistry[name];
@@ -4679,6 +4865,17 @@ function formatGadgetExpanded(node) {
     }
     lines.push(`${indent}${chalk3.dim(BOX.bottomLeft + BOX.horizontal.repeat(width - 1))}`);
   }
+  if (node.mediaOutputs && node.mediaOutputs.length > 0) {
+    const headerLine = `${BOX.topLeft}${BOX.horizontal} Media ${BOX.horizontal.repeat(width - 9)}`;
+    lines.push(`${indent}${chalk3.dim(headerLine)}`);
+    for (const media of node.mediaOutputs) {
+      const kindEmoji = media.kind === "audio" ? "\u{1F50A}" : media.kind === "image" ? "\u{1F5BC}\uFE0F" : media.kind === "video" ? "\u{1F3AC}" : "\u{1F4C4}";
+      const maxPathLen = width - 8;
+      const displayPath = media.path.length > maxPathLen ? "..." + media.path.slice(-(maxPathLen - 3)) : media.path;
+      lines.push(`${indent}${chalk3.dim(BOX.vertical)} ${kindEmoji} ${chalk3.cyan(displayPath)}`);
+    }
+    lines.push(`${indent}${chalk3.dim(BOX.bottomLeft + BOX.horizontal.repeat(width - 1))}`);
+  }
   if (node.children.length > 0) {
     const headerLine = `${BOX.topLeft}${BOX.horizontal} Subagent Activity ${BOX.horizontal.repeat(width - 21)}`;
     lines.push(`${indent}${chalk3.dim(headerLine)}`);
@@ -4906,14 +5103,16 @@ var BlockRenderer = class _BlockRenderer {
   /**
    * Complete a gadget with result.
    */
-  completeGadget(invocationId, result, error, executionTimeMs, cost) {
+  completeGadget(invocationId, options = {}) {
     const node = this.findGadgetByInvocationId(invocationId);
     if (!node) return;
+    const { result, error, executionTimeMs, cost, mediaOutputs } = options;
     node.isComplete = true;
     node.result = result;
     node.error = error;
     node.executionTimeMs = executionTimeMs;
     node.cost = cost;
+    node.mediaOutputs = mediaOutputs;
     if (result) {
       node.resultTokens = Math.ceil(result.length / 4);
     }
@@ -5939,17 +6138,25 @@ ${indicator}`;
         break;
       }
       case "gadget_complete": {
-        this.completeGadget(
-          event.invocationId,
-          event.result,
-          void 0,
-          event.executionTimeMs,
-          event.cost
-        );
+        const mediaOutputs = event.storedMedia?.map((m) => ({
+          kind: m.kind,
+          path: m.path,
+          mimeType: m.mimeType,
+          description: m.description
+        }));
+        this.completeGadget(event.invocationId, {
+          result: event.result,
+          executionTimeMs: event.executionTimeMs,
+          cost: event.cost,
+          mediaOutputs
+        });
         break;
       }
       case "gadget_error": {
-        this.completeGadget(event.invocationId, void 0, event.error, event.executionTimeMs);
+        this.completeGadget(event.invocationId, {
+          error: event.error,
+          executionTimeMs: event.executionTimeMs
+        });
         break;
       }
       case "gadget_skipped": {
@@ -6776,11 +6983,8 @@ var InputHandler = class {
   handlePaste(content) {
     if (!content) return;
     if (content.includes("\n")) {
-      const currentValue = this.inputBar.getValue();
-      this.openEditorForInput(currentValue + content);
+      this.openEditorForInput(this.inputBar.getValue());
     } else {
-      const currentValue = this.inputBar.getValue();
-      this.inputBar.setValue(currentValue + content);
       this.inputBar.readInput();
     }
   }
@@ -8683,8 +8887,15 @@ async function executeAgent(promptArg, options, env, commandName) {
     prompt = await resolvePrompt(promptArg, env);
   }
   const registry = new GadgetRegistry();
+  let speechConfig;
+  try {
+    const fullConfig = loadConfig();
+    speechConfig = fullConfig.speech;
+  } catch {
+  }
   if (options.builtins !== false) {
-    for (const gadget of builtinGadgets) {
+    const builtins = getBuiltinGadgets(speechConfig);
+    for (const gadget of builtins) {
       if (gadget.name === "AskUser" && (options.builtinInteraction === false || !stdinIsInteractive || !stdoutTTY)) {
         continue;
       }
@@ -9005,6 +9216,7 @@ ${ctx.gadgetName} requires interactive approval. Run in a terminal to approve.`
     if (tui) {
       tui.resetAbort();
       tui.startNewSession();
+      tui.showUserMessage(userPrompt);
       builder.withSignal(tui.getAbortSignal());
     }
     if (currentAgent) {
@@ -9070,7 +9282,6 @@ ${ctx.gadgetName} requires interactive approval. Run in a terminal to approve.`
     if (!currentPrompt) {
       tui.setFocusMode("input");
       currentPrompt = await tui.waitForPrompt();
-      tui.showUserMessage(currentPrompt);
     }
     while (true) {
       try {
@@ -9081,7 +9292,6 @@ ${ctx.gadgetName} requires interactive approval. Run in a terminal to approve.`
         }
       }
       currentPrompt = await tui.waitForPrompt();
-      tui.showUserMessage(currentPrompt);
     }
   } else {
     try {