llmist 2.2.0 → 2.4.0

package/dist/cli.js

@@ -1,5 +1,5 @@
 #!/usr/bin/env node
-import "./chunk-ZDNV7DDO.js";
+import "./chunk-QFRVTS5F.js";
 import {
   AgentBuilder,
   BaseGadget,
@@ -27,7 +27,7 @@ import {
   resolveModel,
   schemaToJSONSchema,
   validateGadgetSchema
-} from "./chunk-GANXNBIZ.js";
+} from "./chunk-6ZDUWO6N.js";
 
 // src/cli/constants.ts
 var CLI_NAME = "llmist";
@@ -36,7 +36,9 @@ var COMMANDS = {
   complete: "complete",
   agent: "agent",
   models: "models",
-  gadget: "gadget"
+  gadget: "gadget",
+  image: "image",
+  speech: "speech"
 };
 var LOG_LEVELS = ["silly", "trace", "debug", "info", "warn", "error", "fatal"];
 var DEFAULT_MODEL = "openai:gpt-5-nano";
@@ -57,7 +59,17 @@ var OPTION_FLAGS = {
   docker: "--docker",
   dockerRo: "--docker-ro",
   noDocker: "--no-docker",
-  dockerDev: "--docker-dev"
+  dockerDev: "--docker-dev",
+  // Image generation options
+  imageSize: "--size <size>",
+  imageQuality: "--quality <quality>",
+  imageCount: "-n, --count <number>",
+  imageOutput: "-o, --output <path>",
+  // Speech generation options
+  voice: "--voice <name>",
+  speechFormat: "--format <format>",
+  speechSpeed: "--speed <value>",
+  speechOutput: "-o, --output <path>"
 };
 var OPTION_DESCRIPTIONS = {
   model: "Model identifier, e.g. openai:gpt-5-nano or anthropic:claude-sonnet-4-5.",
@@ -76,7 +88,17 @@ var OPTION_DESCRIPTIONS = {
   docker: "Run agent in a Docker sandbox container for security isolation.",
   dockerRo: "Run in Docker with current directory mounted read-only.",
   noDocker: "Disable Docker sandboxing (override config).",
-  dockerDev: "Run in Docker dev mode (mount local source instead of npm install)."
+  dockerDev: "Run in Docker dev mode (mount local source instead of npm install).",
+  // Image generation descriptions
+  imageSize: "Image size/aspect ratio, e.g. '1024x1024', '1:1', '16:9'.",
+  imageQuality: "Image quality: 'standard', 'hd', 'low', 'medium', 'high'.",
+  imageCount: "Number of images to generate (model dependent, usually 1-4).",
+  imageOutput: "Output path for the generated image. Defaults to stdout if not specified.",
+  // Speech generation descriptions
+  voice: "Voice name for speech generation, e.g. 'nova', 'alloy', 'Zephyr'.",
+  speechFormat: "Audio format: 'mp3', 'opus', 'aac', 'flac', 'wav', 'pcm'.",
+  speechSpeed: "Speech speed multiplier (0.25 to 4.0, default 1.0).",
+  speechOutput: "Output path for audio file. Defaults to stdout if not specified."
 };
 var SUMMARY_PREFIX = "[llmist]";
 
@@ -86,8 +108,8 @@ import { Command, InvalidArgumentError as InvalidArgumentError2 } from "commande
 // package.json
 var package_default = {
   name: "llmist",
-  version: "2.1.0",
-  description: "Universal TypeScript LLM client with streaming-first agent framework. Works with any model - no structured outputs or native tool calling required. Implements its own flexible grammar for function calling.",
+  version: "2.4.0",
+  description: "TypeScript LLM client with streaming tool execution. Tools fire mid-stream. Built-in function calling works with any model\u2014no structured outputs or native tool support required.",
   type: "module",
   main: "dist/index.cjs",
   module: "dist/index.js",
@@ -161,9 +183,16 @@ var package_default = {
     "universal-client",
     "multi-provider",
     "hooks",
-    "gadgets"
+    "gadgets",
+    "chatbot",
+    "chatgpt",
+    "agentic",
+    "language-model",
+    "generative-ai",
+    "bun",
+    "nodejs"
   ],
-  author: "",
+  author: "Zbigniew Sobiecki <zbigniew@sobiecki.name>",
   license: "MIT",
   dependencies: {
     "@anthropic-ai/sdk": "^0.69.0",
@@ -2101,6 +2130,22 @@ var AGENT_CONFIG_KEYS = /* @__PURE__ */ new Set([
   "docker-cwd-permission"
   // Override CWD mount permission for this profile
 ]);
+var IMAGE_CONFIG_KEYS = /* @__PURE__ */ new Set([
+  "model",
+  "size",
+  "quality",
+  "count",
+  "output",
+  "quiet"
+]);
+var SPEECH_CONFIG_KEYS = /* @__PURE__ */ new Set([
+  "model",
+  "voice",
+  "format",
+  "speed",
+  "output",
+  "quiet"
+]);
 var CUSTOM_CONFIG_KEYS = /* @__PURE__ */ new Set([
   ...COMPLETE_CONFIG_KEYS,
   ...AGENT_CONFIG_KEYS,
@@ -2361,6 +2406,75 @@ function validateAgentConfig(raw, section) {
   }
   return result;
 }
+function validateImageConfig(raw, section) {
+  if (typeof raw !== "object" || raw === null) {
+    throw new ConfigError(`[${section}] must be a table`);
+  }
+  const rawObj = raw;
+  for (const key of Object.keys(rawObj)) {
+    if (!IMAGE_CONFIG_KEYS.has(key)) {
+      throw new ConfigError(`[${section}].${key} is not a valid option`);
+    }
+  }
+  const result = {};
+  if ("model" in rawObj) {
+    result.model = validateString(rawObj.model, "model", section);
+  }
+  if ("size" in rawObj) {
+    result.size = validateString(rawObj.size, "size", section);
+  }
+  if ("quality" in rawObj) {
+    result.quality = validateString(rawObj.quality, "quality", section);
+  }
+  if ("count" in rawObj) {
+    result.count = validateNumber(rawObj.count, "count", section, {
+      integer: true,
+      min: 1,
+      max: 10
+    });
+  }
+  if ("output" in rawObj) {
+    result.output = validateString(rawObj.output, "output", section);
+  }
+  if ("quiet" in rawObj) {
+    result.quiet = validateBoolean(rawObj.quiet, "quiet", section);
+  }
+  return result;
+}
+function validateSpeechConfig(raw, section) {
+  if (typeof raw !== "object" || raw === null) {
+    throw new ConfigError(`[${section}] must be a table`);
+  }
+  const rawObj = raw;
+  for (const key of Object.keys(rawObj)) {
+    if (!SPEECH_CONFIG_KEYS.has(key)) {
+      throw new ConfigError(`[${section}].${key} is not a valid option`);
+    }
+  }
+  const result = {};
+  if ("model" in rawObj) {
+    result.model = validateString(rawObj.model, "model", section);
+  }
+  if ("voice" in rawObj) {
+    result.voice = validateString(rawObj.voice, "voice", section);
+  }
+  if ("format" in rawObj) {
+    result.format = validateString(rawObj.format, "format", section);
+  }
+  if ("speed" in rawObj) {
+    result.speed = validateNumber(rawObj.speed, "speed", section, {
+      min: 0.25,
+      max: 4
+    });
+  }
+  if ("output" in rawObj) {
+    result.output = validateString(rawObj.output, "output", section);
+  }
+  if ("quiet" in rawObj) {
+    result.quiet = validateBoolean(rawObj.quiet, "quiet", section);
+  }
+  return result;
+}
 function validateStringOrBoolean(value, field, section) {
   if (typeof value === "string" || typeof value === "boolean") {
     return value;
@@ -2483,6 +2597,10 @@ function validateConfig(raw, configPath) {
         result.complete = validateCompleteConfig(value, key);
       } else if (key === "agent") {
         result.agent = validateAgentConfig(value, key);
+      } else if (key === "image") {
+        result.image = validateImageConfig(value, key);
+      } else if (key === "speech") {
+        result.speech = validateSpeechConfig(value, key);
       } else if (key === "prompts") {
         result.prompts = validatePromptsConfig(value, key);
       } else if (key === "docker") {
@@ -2527,7 +2645,7 @@ function loadConfig() {
   return resolveTemplatesInConfig(inherited, configPath);
 }
 function getCustomCommandNames(config) {
-  const reserved = /* @__PURE__ */ new Set(["global", "complete", "agent", "prompts", "docker"]);
+  const reserved = /* @__PURE__ */ new Set(["global", "complete", "agent", "image", "speech", "prompts", "docker"]);
   return Object.keys(config).filter((key) => !reserved.has(key));
 }
 function resolveTemplatesInConfig(config, configPath) {
@@ -4122,19 +4240,118 @@ function registerGadgetCommand(program, env) {
   );
 }
 
+// src/cli/image-command.ts
+import { writeFileSync as writeFileSync2 } from "node:fs";
+var DEFAULT_IMAGE_MODEL = "dall-e-3";
+async function executeImage(promptArg, options, env) {
+  const prompt = await resolvePrompt(promptArg, env);
+  const client = env.createClient();
+  const model = options.model;
+  const n = options.count ? Number.parseInt(options.count, 10) : 1;
+  const stderrTTY = env.stderr.isTTY === true;
+  if (!options.quiet && stderrTTY) {
+    env.stderr.write(`${SUMMARY_PREFIX} Generating image with ${model}...
+`);
+  }
+  const result = await client.image.generate({
+    model,
+    prompt,
+    size: options.size,
+    quality: options.quality,
+    n,
+    responseFormat: options.output ? "b64_json" : "url"
+  });
+  if (options.output) {
+    const imageData = result.images[0];
+    if (imageData.b64Json) {
+      const buffer = Buffer.from(imageData.b64Json, "base64");
+      writeFileSync2(options.output, buffer);
+      if (!options.quiet) {
+        env.stderr.write(`${SUMMARY_PREFIX} Image saved to ${options.output}
+`);
+      }
+    } else if (imageData.url) {
+      env.stdout.write(`${imageData.url}
+`);
+    }
+  } else {
+    for (const image of result.images) {
+      if (image.url) {
+        env.stdout.write(`${image.url}
+`);
+      } else if (image.b64Json) {
+        env.stdout.write(image.b64Json);
+      }
+    }
+  }
+  if (!options.quiet && stderrTTY) {
+    const parts = [
+      `${result.images.length} image(s)`,
+      `size: ${result.usage.size}`,
+      `quality: ${result.usage.quality}`
+    ];
+    if (result.cost !== void 0) {
+      parts.push(`cost: ${formatCost(result.cost)}`);
+    }
+    env.stderr.write(`${SUMMARY_PREFIX} ${parts.join(" | ")}
+`);
+  }
+}
+function registerImageCommand(program, env, config) {
+  program.command(COMMANDS.image).description("Generate images from a text prompt.").argument("[prompt]", "Image generation prompt. If omitted, stdin is used when available.").option(
+    OPTION_FLAGS.model,
+    OPTION_DESCRIPTIONS.model,
+    config?.model ?? DEFAULT_IMAGE_MODEL
+  ).option(OPTION_FLAGS.imageSize, OPTION_DESCRIPTIONS.imageSize, config?.size).option(OPTION_FLAGS.imageQuality, OPTION_DESCRIPTIONS.imageQuality, config?.quality).option(OPTION_FLAGS.imageCount, OPTION_DESCRIPTIONS.imageCount, config?.count?.toString()).option(OPTION_FLAGS.imageOutput, OPTION_DESCRIPTIONS.imageOutput, config?.output).option(OPTION_FLAGS.quiet, OPTION_DESCRIPTIONS.quiet, config?.quiet ?? false).action(
+    (prompt, options) => executeAction(() => executeImage(prompt, options, env), env)
+  );
+}
+
 // src/cli/models-command.ts
 import chalk8 from "chalk";
 init_model_shortcuts();
 async function handleModelsCommand(options, env) {
   const client = env.createClient();
-  const models = client.modelRegistry.listModels(options.provider);
+  const showText = options.all || options.text || !options.image && !options.speech;
+  const showImage = options.all || options.image;
+  const showSpeech = options.all || options.speech;
+  const textModels = showText ? client.modelRegistry.listModels(options.provider) : [];
+  const imageModels = showImage ? client.image.listModels().filter((m) => !options.provider || m.provider === options.provider) : [];
+  const speechModels = showSpeech ? client.speech.listModels().filter((m) => !options.provider || m.provider === options.provider) : [];
   if (options.format === "json") {
-    renderJSON(models, env.stdout);
+    renderJSON(textModels, imageModels, speechModels, env.stdout);
   } else {
-    renderTable(models, options.verbose || false, env.stdout);
+    renderAllTables(textModels, imageModels, speechModels, options.verbose || false, env.stdout);
   }
 }
-function renderTable(models, verbose, stream) {
+function renderAllTables(textModels, imageModels, speechModels, verbose, stream) {
+  const hasAnyModels = textModels.length > 0 || imageModels.length > 0 || speechModels.length > 0;
+  if (!hasAnyModels) {
+    stream.write(chalk8.yellow("\nNo models found matching the specified criteria.\n\n"));
+    return;
+  }
+  stream.write(chalk8.bold.cyan("\nAvailable Models\n"));
+  stream.write(chalk8.cyan("=".repeat(80)) + "\n\n");
+  if (textModels.length > 0) {
+    renderTextTable(textModels, verbose, stream);
+  }
+  if (imageModels.length > 0) {
+    renderImageTable(imageModels, verbose, stream);
+  }
+  if (speechModels.length > 0) {
+    renderSpeechTable(speechModels, verbose, stream);
+  }
+  if (textModels.length > 0) {
+    stream.write(chalk8.bold.magenta("Model Shortcuts\n"));
+    stream.write(chalk8.dim("\u2500".repeat(80)) + "\n");
+    const shortcuts = Object.entries(MODEL_ALIASES).sort((a, b) => a[0].localeCompare(b[0]));
+    for (const [shortcut, fullName] of shortcuts) {
+      stream.write(chalk8.cyan(`  ${shortcut.padEnd(15)}`) + chalk8.dim(" \u2192 ") + chalk8.white(fullName) + "\n");
+    }
+    stream.write("\n");
+  }
+}
+function renderTextTable(models, verbose, stream) {
   const grouped = /* @__PURE__ */ new Map();
   for (const model of models) {
     const provider = model.provider;
@@ -4143,13 +4360,13 @@ function renderTable(models, verbose, stream) {
     }
     grouped.get(provider).push(model);
   }
-  stream.write(chalk8.bold.cyan("\nAvailable Models\n"));
-  stream.write(chalk8.cyan("=".repeat(80)) + "\n\n");
+  stream.write(chalk8.bold.blue("\u{1F4DD} Text/LLM Models\n"));
+  stream.write(chalk8.dim("\u2500".repeat(80)) + "\n\n");
   const providers = Array.from(grouped.keys()).sort();
   for (const provider of providers) {
     const providerModels = grouped.get(provider);
     const providerName = provider.charAt(0).toUpperCase() + provider.slice(1);
-    stream.write(chalk8.bold.yellow(`${providerName} Models
+    stream.write(chalk8.bold.yellow(`${providerName}
 `));
     if (verbose) {
       renderVerboseTable(providerModels, stream);
@@ -4158,13 +4375,6 @@ function renderTable(models, verbose, stream) {
     }
     stream.write("\n");
   }
-  stream.write(chalk8.bold.magenta("Model Shortcuts\n"));
-  stream.write(chalk8.dim("\u2500".repeat(80)) + "\n");
-  const shortcuts = Object.entries(MODEL_ALIASES).sort((a, b) => a[0].localeCompare(b[0]));
-  for (const [shortcut, fullName] of shortcuts) {
-    stream.write(chalk8.cyan(`  ${shortcut.padEnd(15)}`) + chalk8.dim(" \u2192 ") + chalk8.white(fullName) + "\n");
-  }
-  stream.write("\n");
 }
 function renderCompactTable(models, stream) {
   const idWidth = 25;
@@ -4241,9 +4451,171 @@ function renderVerboseTable(models, stream) {
   }
   stream.write("\n");
 }
-function renderJSON(models, stream) {
-  const output = {
-    models: models.map((model) => ({
+function renderImageTable(models, verbose, stream) {
+  stream.write(chalk8.bold.green("\u{1F3A8} Image Generation Models\n"));
+  stream.write(chalk8.dim("\u2500".repeat(80)) + "\n\n");
+  const grouped = /* @__PURE__ */ new Map();
+  for (const model of models) {
+    if (!grouped.has(model.provider)) {
+      grouped.set(model.provider, []);
+    }
+    grouped.get(model.provider).push(model);
+  }
+  for (const [provider, providerModels] of Array.from(grouped.entries()).sort()) {
+    const providerName = provider.charAt(0).toUpperCase() + provider.slice(1);
+    stream.write(chalk8.bold.yellow(`${providerName}
+`));
+    if (verbose) {
+      for (const model of providerModels) {
+        stream.write(chalk8.bold.green(`
+  ${model.modelId}
+`));
+        stream.write(chalk8.dim("  " + "\u2500".repeat(60)) + "\n");
+        stream.write(`  ${chalk8.dim("Name:")}      ${chalk8.white(model.displayName)}
+`);
+        stream.write(`  ${chalk8.dim("Sizes:")}     ${chalk8.yellow(model.supportedSizes.join(", "))}
+`);
+        if (model.supportedQualities) {
+          stream.write(`  ${chalk8.dim("Qualities:")} ${chalk8.yellow(model.supportedQualities.join(", "))}
+`);
+        }
+        stream.write(`  ${chalk8.dim("Max Images:")} ${chalk8.yellow(model.maxImages.toString())}
+`);
+        stream.write(`  ${chalk8.dim("Pricing:")}   ${chalk8.cyan(formatImagePrice(model))}
+`);
+        if (model.features) {
+          const features = [];
+          if (model.features.textRendering) features.push("text-rendering");
+          if (model.features.transparency) features.push("transparency");
+          if (model.features.conversational) features.push("conversational");
+          if (features.length > 0) {
+            stream.write(`  ${chalk8.dim("Features:")}  ${chalk8.blue(features.join(", "))}
+`);
+          }
+        }
+      }
+    } else {
+      const idWidth = 32;
+      const nameWidth = 25;
+      const sizesWidth = 20;
+      const priceWidth = 15;
+      stream.write(chalk8.dim("\u2500".repeat(idWidth + nameWidth + sizesWidth + priceWidth + 6)) + "\n");
+      stream.write(
+        chalk8.bold(
+          "Model ID".padEnd(idWidth) + "  " + "Display Name".padEnd(nameWidth) + "  " + "Sizes".padEnd(sizesWidth) + "  " + "Price".padEnd(priceWidth)
+        ) + "\n"
+      );
+      stream.write(chalk8.dim("\u2500".repeat(idWidth + nameWidth + sizesWidth + priceWidth + 6)) + "\n");
+      for (const model of providerModels) {
+        const sizes = model.supportedSizes.length > 2 ? model.supportedSizes.slice(0, 2).join(", ") + "..." : model.supportedSizes.join(", ");
+        stream.write(
+          chalk8.green(model.modelId.padEnd(idWidth)) + "  " + chalk8.white(model.displayName.substring(0, nameWidth - 1).padEnd(nameWidth)) + "  " + chalk8.yellow(sizes.padEnd(sizesWidth)) + "  " + chalk8.cyan(formatImagePrice(model).padEnd(priceWidth)) + "\n"
+        );
+      }
+      stream.write(chalk8.dim("\u2500".repeat(idWidth + nameWidth + sizesWidth + priceWidth + 6)) + "\n");
+    }
+    stream.write("\n");
+  }
+}
+function renderSpeechTable(models, verbose, stream) {
+  stream.write(chalk8.bold.magenta("\u{1F3A4} Speech (TTS) Models\n"));
+  stream.write(chalk8.dim("\u2500".repeat(80)) + "\n\n");
+  const grouped = /* @__PURE__ */ new Map();
+  for (const model of models) {
+    if (!grouped.has(model.provider)) {
+      grouped.set(model.provider, []);
+    }
+    grouped.get(model.provider).push(model);
+  }
+  for (const [provider, providerModels] of Array.from(grouped.entries()).sort()) {
+    const providerName = provider.charAt(0).toUpperCase() + provider.slice(1);
+    stream.write(chalk8.bold.yellow(`${providerName}
+`));
+    if (verbose) {
+      for (const model of providerModels) {
+        stream.write(chalk8.bold.green(`
+  ${model.modelId}
+`));
+        stream.write(chalk8.dim("  " + "\u2500".repeat(60)) + "\n");
+        stream.write(`  ${chalk8.dim("Name:")}    ${chalk8.white(model.displayName)}
+`);
+        stream.write(`  ${chalk8.dim("Voices:")}  ${chalk8.yellow(model.voices.length.toString())} voices
+`);
+        if (model.voices.length <= 6) {
+          stream.write(`            ${chalk8.dim(model.voices.join(", "))}
+`);
+        } else {
+          stream.write(`            ${chalk8.dim(model.voices.slice(0, 6).join(", ") + "...")}
+`);
+        }
+        stream.write(`  ${chalk8.dim("Formats:")} ${chalk8.yellow(model.formats.join(", "))}
+`);
+        stream.write(`  ${chalk8.dim("Max Input:")} ${chalk8.yellow(model.maxInputLength.toString())} chars
+`);
+        stream.write(`  ${chalk8.dim("Pricing:")} ${chalk8.cyan(formatSpeechPrice(model))}
+`);
+        if (model.features) {
+          const features = [];
+          if (model.features.multiSpeaker) features.push("multi-speaker");
+          if (model.features.voiceInstructions) features.push("voice-instructions");
+          if (model.features.languages) features.push(`${model.features.languages} languages`);
+          if (features.length > 0) {
+            stream.write(`  ${chalk8.dim("Features:")} ${chalk8.blue(features.join(", "))}
+`);
+          }
+        }
+      }
+    } else {
+      const idWidth = 30;
+      const nameWidth = 28;
+      const voicesWidth = 12;
+      const priceWidth = 18;
+      stream.write(chalk8.dim("\u2500".repeat(idWidth + nameWidth + voicesWidth + priceWidth + 6)) + "\n");
+      stream.write(
+        chalk8.bold(
+          "Model ID".padEnd(idWidth) + "  " + "Display Name".padEnd(nameWidth) + "  " + "Voices".padEnd(voicesWidth) + "  " + "Price".padEnd(priceWidth)
+        ) + "\n"
+      );
+      stream.write(chalk8.dim("\u2500".repeat(idWidth + nameWidth + voicesWidth + priceWidth + 6)) + "\n");
+      for (const model of providerModels) {
+        stream.write(
+          chalk8.green(model.modelId.padEnd(idWidth)) + "  " + chalk8.white(model.displayName.substring(0, nameWidth - 1).padEnd(nameWidth)) + "  " + chalk8.yellow(`${model.voices.length} voices`.padEnd(voicesWidth)) + "  " + chalk8.cyan(formatSpeechPrice(model).padEnd(priceWidth)) + "\n"
+        );
+      }
+      stream.write(chalk8.dim("\u2500".repeat(idWidth + nameWidth + voicesWidth + priceWidth + 6)) + "\n");
+    }
+    stream.write("\n");
+  }
+}
+function formatImagePrice(model) {
+  if (model.pricing.perImage !== void 0) {
+    return `$${model.pricing.perImage.toFixed(2)}/img`;
+  }
+  if (model.pricing.bySize) {
+    const prices = Object.values(model.pricing.bySize);
+    const minPrice = Math.min(...prices.flatMap((p) => typeof p === "number" ? [p] : Object.values(p)));
+    const maxPrice = Math.max(...prices.flatMap((p) => typeof p === "number" ? [p] : Object.values(p)));
+    if (minPrice === maxPrice) {
+      return `$${minPrice.toFixed(2)}/img`;
+    }
+    return `$${minPrice.toFixed(2)}-${maxPrice.toFixed(2)}`;
+  }
+  return "varies";
+}
+function formatSpeechPrice(model) {
+  if (model.pricing.perCharacter !== void 0) {
+    const perMillion = model.pricing.perCharacter * 1e6;
+    return `$${perMillion.toFixed(0)}/1M chars`;
+  }
+  if (model.pricing.perMinute !== void 0) {
+    return `~$${model.pricing.perMinute.toFixed(2)}/min`;
+  }
+  return "varies";
+}
+function renderJSON(textModels, imageModels, speechModels, stream) {
+  const output = {};
+  if (textModels.length > 0) {
+    output.textModels = textModels.map((model) => ({
       provider: model.provider,
       modelId: model.modelId,
       displayName: model.displayName,
@@ -4259,9 +4631,33 @@ function renderJSON(models, stream) {
       knowledgeCutoff: model.knowledgeCutoff,
       features: model.features,
       metadata: model.metadata
-    })),
-    shortcuts: MODEL_ALIASES
-  };
+    }));
+    output.shortcuts = MODEL_ALIASES;
+  }
+  if (imageModels.length > 0) {
+    output.imageModels = imageModels.map((model) => ({
+      provider: model.provider,
+      modelId: model.modelId,
+      displayName: model.displayName,
+      supportedSizes: model.supportedSizes,
+      supportedQualities: model.supportedQualities,
+      maxImages: model.maxImages,
+      pricing: model.pricing,
+      features: model.features
+    }));
+  }
+  if (speechModels.length > 0) {
+    output.speechModels = speechModels.map((model) => ({
+      provider: model.provider,
+      modelId: model.modelId,
+      displayName: model.displayName,
+      voices: model.voices,
+      formats: model.formats,
+      maxInputLength: model.maxInputLength,
+      pricing: model.pricing,
+      features: model.features
+    }));
+  }
   stream.write(JSON.stringify(output, null, 2) + "\n");
 }
 function formatTokens2(count) {
@@ -4274,7 +4670,7 @@ function formatTokens2(count) {
   }
 }
 function registerModelsCommand(program, env) {
-  program.command(COMMANDS.models).description("List all available LLM models with pricing and capabilities.").option("--provider <name>", "Filter by provider (openai, anthropic, gemini)").option("--format <format>", "Output format: table or json", "table").option("--verbose", "Show detailed model information", false).action(
+  program.command(COMMANDS.models).description("List available models with pricing and capabilities.").option("--provider <name>", "Filter by provider (openai, anthropic, gemini)").option("--format <format>", "Output format: table or json", "table").option("--verbose", "Show detailed model information", false).option("--text", "Show text/LLM models (default if no type specified)").option("--image", "Show image generation models").option("--speech", "Show speech/TTS models").option("--all", "Show all model types (text, image, speech)").action(
     (options) => executeAction(
       () => handleModelsCommand(options, env),
       env
@@ -4282,6 +4678,60 @@ function registerModelsCommand(program, env) {
   );
 }
 
+// src/cli/speech-command.ts
+import { writeFileSync as writeFileSync3 } from "node:fs";
+var DEFAULT_SPEECH_MODEL = "tts-1";
+var DEFAULT_VOICE = "nova";
+async function executeSpeech(textArg, options, env) {
+  const text = await resolvePrompt(textArg, env);
+  const client = env.createClient();
+  const model = options.model;
+  const voice = options.voice ?? DEFAULT_VOICE;
+  const speed = options.speed ? Number.parseFloat(options.speed) : void 0;
+  const stderrTTY = env.stderr.isTTY === true;
+  if (!options.quiet && stderrTTY) {
+    env.stderr.write(`${SUMMARY_PREFIX} Generating speech with ${model} (voice: ${voice})...
+`);
+  }
+  const result = await client.speech.generate({
+    model,
+    input: text,
+    voice,
+    responseFormat: options.format,
+    speed
+  });
+  const audioBuffer = Buffer.from(result.audio);
+  if (options.output) {
+    writeFileSync3(options.output, audioBuffer);
+    if (!options.quiet) {
+      env.stderr.write(`${SUMMARY_PREFIX} Audio saved to ${options.output}
+`);
+    }
+  } else {
+    env.stdout.write(audioBuffer);
+  }
+  if (!options.quiet && stderrTTY) {
+    const parts = [
+      `${result.usage.characterCount} characters`,
+      `format: ${result.format}`
+    ];
+    if (result.cost !== void 0) {
+      parts.push(`cost: ${formatCost(result.cost)}`);
+    }
+    env.stderr.write(`${SUMMARY_PREFIX} ${parts.join(" | ")}
+`);
+  }
+}
+function registerSpeechCommand(program, env, config) {
+  program.command(COMMANDS.speech).description("Generate speech audio from text.").argument("[text]", "Text to convert to speech. If omitted, stdin is used when available.").option(
+    OPTION_FLAGS.model,
+    OPTION_DESCRIPTIONS.model,
+    config?.model ?? DEFAULT_SPEECH_MODEL
+  ).option(OPTION_FLAGS.voice, OPTION_DESCRIPTIONS.voice, config?.voice ?? DEFAULT_VOICE).option(OPTION_FLAGS.speechFormat, OPTION_DESCRIPTIONS.speechFormat, config?.format).option(OPTION_FLAGS.speechSpeed, OPTION_DESCRIPTIONS.speechSpeed, config?.speed?.toString()).option(OPTION_FLAGS.speechOutput, OPTION_DESCRIPTIONS.speechOutput, config?.output).option(OPTION_FLAGS.quiet, OPTION_DESCRIPTIONS.quiet, config?.quiet ?? false).action(
+    (text, options) => executeAction(() => executeSpeech(text, options, env), env)
+  );
+}
+
 // src/cli/environment.ts
 init_client();
 init_logger();
@@ -4433,6 +4883,8 @@ function createProgram(env, config) {
   });
   registerCompleteCommand(program, env, config?.complete);
   registerAgentCommand(program, env, config?.agent);
+  registerImageCommand(program, env, config?.image);
+  registerSpeechCommand(program, env, config?.speech);
   registerModelsCommand(program, env);
   registerGadgetCommand(program, env);
   if (config) {