egaki 0.4.0 → 0.5.0

package/CHANGELOG.md

@@ -2,6 +2,33 @@
 
 All notable changes to this project will be documented in this file.
 
+## 0.5.0
+
+1. **Interactive model picker** — omitting `--model` now shows a select prompt with popular models instead of silently using a default:
+
+   ```bash
+   # Shows a model picker in the terminal
+   egaki image "a dreamy watercolor landscape" -o landscape.png
+
+   # Explicit model skips the picker
+   egaki image "a dreamy watercolor landscape" -m imagen-4.0-generate-001 -o landscape.png
+   ```
+
+   Works for both `egaki image` and `egaki video`. In non-interactive (piped/scripted) mode, the default model is used automatically so existing scripts are unaffected.
+
+## 0.4.1
+
+1. **Added `gpt-image-2` model** — OpenAI's new flagship image generation model (released 2026-04-21) is now available in the model catalog:
+
+   ```bash
+   egaki image "a dreamy watercolor landscape" -m gpt-image-2 -o landscape.png
+   egaki models --json | jq '.[] | select(.id == "gpt-image-2")'
+   ```
+
+   Supports editing, inpainting, and multiple output images. Token-based pricing (~$0.053 for medium-quality 1024×1024).
+
+2. **`chatgpt-image-latest` now tracks `gpt-image-2`** — the rolling alias has been updated from `gpt-image-1.5` to `gpt-image-2`, with pricing adjusted accordingly.
+
 ## 0.4.0
 
 1. **ChatGPT login for OpenAI image generation** — use your ChatGPT subscription in `egaki` without a separate OpenAI platform API key:

package/dist/cli.js

@@ -11,6 +11,7 @@ import { goke } from 'goke';
 import { z } from 'zod';
 import dedent from 'string-dedent';
 import { generateImage as aiGenerateImage, generateText, experimental_generateVideo as aiGenerateVideo, } from 'ai';
+import { select, isCancel, cancel } from '@clack/prompts';
 import fs from 'node:fs';
 import path from 'node:path';
 import pc from 'picocolors';
@@ -18,6 +19,7 @@ import { createParser } from 'eventsource-parser';
 import pkg from '../package.json' with { type: 'json' };
 import { injectCredentialsToEnv, PROVIDERS, getChatGptAuth, saveChatGptAuth, getKeyStatus, } from './credentials.js';
 import { IMAGE_MODELS, VIDEO_MODELS, DEFAULT_MODEL, DEFAULT_VIDEO_MODEL, getModelConfig, createImageModel, createTextModel, createVideoModel, shouldUseResponsesApi, } from './models.js';
+import { CATALOG } from './model-catalog.js';
 import { VIDEO_CATALOG } from './video-model-catalog.js';
 import { loginInteractive, loginNonInteractive, showLoginStatus, removeLogin, readKeyFromStdin, } from './login.js';
 import { subscribeInteractive, subscribeNonInteractive, unsubscribe, showUsage, } from './subscription.js';
@@ -144,7 +146,7 @@ cli
       from the model ID: imagen-* uses the image API, everything else uses
       the text API with image output enabled.
     `)
-    .option('-m, --model [model]', z.enum(IMAGE_MODELS).default(DEFAULT_MODEL).describe('Model ID for generation'))
+    .option('-m, --model [model]', z.enum(IMAGE_MODELS).describe('Model ID for generation. If omitted, shows an interactive picker (or uses default in non-TTY mode)'))
     .option('-o, --output [path]', z
     .string()
     .default('egaki-output.png')
@@ -187,10 +189,7 @@ cli
     .action(async (prompt, options) => {
     // Inject stored API keys as env vars before calling the AI SDK.
     injectCredentialsToEnv();
-    // goke infers schema .default() values via Zod's input type, which leaves
-    // options.model / options.output / options.count as `T | undefined` even
-    // though the runtime always resolves them. Apply the same defaults here.
-    const model = options.model ?? DEFAULT_MODEL;
+    const model = options.model ?? await resolveImageModel();
     const outputPath = options.output ?? 'egaki-output.png';
     const count = options.count ?? 1;
     const config = getModelConfig(model);
@@ -261,7 +260,7 @@ cli
       Agent note: video generation can be slow. When invoking this command from
       automation, use a command timeout of at least 5 minutes.
     `)
-    .option('-m, --model [model]', z.enum(VIDEO_MODELS).default(DEFAULT_VIDEO_MODEL).describe('Video model ID for generation'))
+    .option('-m, --model [model]', z.enum(VIDEO_MODELS).describe('Video model ID for generation. If omitted, shows an interactive picker (or uses default in non-TTY mode)'))
     .option('-o, --output [path]', z
     .string()
     .default('egaki-output.mp4')
@@ -291,8 +290,7 @@ cli
     .example('egaki video "waves crashing on cliffs" -n 2 -o waves.mp4')
     .action(async (prompt, options) => {
     injectCredentialsToEnv();
-    // Apply the same .default(...) values goke's input-type inference drops.
-    const model = options.model ?? DEFAULT_VIDEO_MODEL;
+    const model = options.model ?? await resolveVideoModel();
     const outputPath = options.output ?? 'egaki-output.mp4';
     const count = options.count ?? 1;
     const config = getModelConfig(model);
@@ -336,7 +334,6 @@ cli
     .describe('Filter by model type: image (image+text-image), video, or all'))
     .option('--json', 'Output as JSON instead of YAML')
     .action(async (options) => {
-    const { CATALOG } = await import('./model-catalog.js');
     const yaml = await import('js-yaml');
     const providerStatuses = Object.fromEntries(Object.keys(PROVIDERS).map((provider) => [provider, getKeyStatus(provider)]));
     let models = options.type === 'video'
@@ -392,6 +389,84 @@ cli
 cli.help();
 cli.version(pkg.version);
 cli.parse();
+// ─── interactive model pickers ──────────────────────────────────────────────
+/** Curated image models shown in the interactive picker, ordered by recommendation. */
+const FEATURED_IMAGE_MODELS = [
+    'imagen-4.0-generate-001',
+    'imagen-4.0-ultra-generate-001',
+    'imagen-4.0-fast-generate-001',
+    'gemini-2.5-flash-image',
+    'gemini-3.1-flash-image-preview',
+    'nano-banana-pro-preview',
+    'gpt-image-2',
+    'gpt-image-1.5',
+    'chatgpt-image-latest',
+    'flux-kontext-pro',
+    'flux-kontext-max',
+    'flux-pro-1.1-ultra',
+    'recraft-v4',
+    'dall-e-3',
+    'grok-imagine-image-pro',
+];
+/**
+ * Resolve the image model: show interactive picker in TTY, use default otherwise.
+ */
+async function resolveImageModel() {
+    const isTTY = process.stdin.isTTY && process.stderr.isTTY;
+    if (!isTTY)
+        return DEFAULT_MODEL;
+    const options = FEATURED_IMAGE_MODELS
+        .map((id) => CATALOG.find((m) => m.id === id))
+        .filter(Boolean)
+        .map((m) => {
+        const cost = m.cost.type === 'per-image'
+            ? pc.dim(`$${m.cost.perImage}/img`)
+            : m.cost.type === 'per-token'
+                ? pc.dim(`$${m.cost.outputPerM}/M out`)
+                : '';
+        return {
+            value: m.id,
+            label: `${m.name} ${cost}`,
+            hint: m.id,
+        };
+    });
+    const selected = await select({
+        message: 'Select an image model',
+        options,
+    });
+    if (isCancel(selected)) {
+        cancel('Cancelled.');
+        process.exit(0);
+    }
+    return selected;
+}
+/**
+ * Resolve the video model: show interactive picker in TTY, use default otherwise.
+ */
+async function resolveVideoModel() {
+    const isTTY = process.stdin.isTTY && process.stderr.isTTY;
+    if (!isTTY)
+        return DEFAULT_VIDEO_MODEL;
+    const options = VIDEO_CATALOG.map((m) => {
+        const dur = m.features.durationRangeSec
+            ? pc.dim(`${m.features.durationRangeSec.min}-${m.features.durationRangeSec.max}s`)
+            : '';
+        return {
+            value: m.id,
+            label: `${m.name} ${dur}`,
+            hint: m.id,
+        };
+    });
+    const selected = await select({
+        message: 'Select a video model',
+        options,
+    });
+    if (isCancel(selected)) {
+        cancel('Cancelled.');
+        process.exit(0);
+    }
+    return selected;
+}
 function isUrl(input) {
     return /^https?:\/\//i.test(input);
 }

package/dist/model-catalog.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"model-catalog.d.ts","sourceRoot":"","sources":["../src/model-catalog.ts"],"names":[],"mappings":"AAUA,MAAM,MAAM,YAAY,GAAG;IACzB,IAAI,EAAE,WAAW,CAAA;IACjB,kDAAkD;IAClD,QAAQ,EAAE,MAAM,CAAA;CACjB,CAAA;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,IAAI,EAAE,WAAW,CAAA;IACjB,mCAAmC;IACnC,SAAS,EAAE,MAAM,CAAA;IACjB,oCAAoC;IACpC,UAAU,EAAE,MAAM,CAAA;CACnB,CAAA;AAED,MAAM,MAAM,SAAS,GAAG,YAAY,GAAG,YAAY,CAAA;AAEnD,MAAM,MAAM,aAAa,GAAG;IAC1B,wCAAwC;IACxC,OAAO,EAAE,OAAO,CAAA;IAChB,mCAAmC;IACnC,UAAU,EAAE,OAAO,CAAA;IACnB,8BAA8B;IAC9B,YAAY,EAAE,MAAM,EAAE,CAAA;IACtB,mDAAmD;IACnD,KAAK,CAAC,EAAE,MAAM,EAAE,CAAA;IAChB,kCAAkC;IAClC,IAAI,EAAE,OAAO,CAAA;IACb,uCAAuC;IACvC,cAAc,EAAE,OAAO,CAAA;CACxB,CAAA;AAED,MAAM,MAAM,UAAU,GAAG;IACvB,EAAE,EAAE,MAAM,CAAA;IACV,IAAI,EAAE,MAAM,CAAA;IACZ,iEAAiE;IACjE,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,QAAQ,EAAE,MAAM,CAAA;IAChB,QAAQ,EAAE,OAAO,GAAG,MAAM,CAAA;IAC1B,mDAAmD;IACnD,QAAQ,EAAE,MAAM,CAAA;IAChB,IAAI,EAAE,SAAS,CAAA;IACf,QAAQ,EAAE,aAAa,CAAA;CACxB,CAAA;AA+ED,eAAO,MAAM,OAAO,EAAE,UAAU,EAq5B/B,CAAA;AAMD,wBAAgB,SAAS,CAAC,EAAE,EAAE,MAAM,GAAG,UAAU,GAAG,SAAS,CAE5D"}
+{"version":3,"file":"model-catalog.d.ts","sourceRoot":"","sources":["../src/model-catalog.ts"],"names":[],"mappings":"AAUA,MAAM,MAAM,YAAY,GAAG;IACzB,IAAI,EAAE,WAAW,CAAA;IACjB,kDAAkD;IAClD,QAAQ,EAAE,MAAM,CAAA;CACjB,CAAA;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,IAAI,EAAE,WAAW,CAAA;IACjB,mCAAmC;IACnC,SAAS,EAAE,MAAM,CAAA;IACjB,oCAAoC;IACpC,UAAU,EAAE,MAAM,CAAA;CACnB,CAAA;AAED,MAAM,MAAM,SAAS,GAAG,YAAY,GAAG,YAAY,CAAA;AAEnD,MAAM,MAAM,aAAa,GAAG;IAC1B,wCAAwC;IACxC,OAAO,EAAE,OAAO,CAAA;IAChB,mCAAmC;IACnC,UAAU,EAAE,OAAO,CAAA;IACnB,8BAA8B;IAC9B,YAAY,EAAE,MAAM,EAAE,CAAA;IACtB,mDAAmD;IACnD,KAAK,CAAC,EAAE,MAAM,EAAE,CAAA;IAChB,kCAAkC;IAClC,IAAI,EAAE,OAAO,CAAA;IACb,uCAAuC;IACvC,cAAc,EAAE,OAAO,CAAA;CACxB,CAAA;AAED,MAAM,MAAM,UAAU,GAAG;IACvB,EAAE,EAAE,MAAM,CAAA;IACV,IAAI,EAAE,MAAM,CAAA;IACZ,iEAAiE;IACjE,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,QAAQ,EAAE,MAAM,CAAA;IAChB,QAAQ,EAAE,OAAO,GAAG,MAAM,CAAA;IAC1B,mDAAmD;IACnD,QAAQ,EAAE,MAAM,CAAA;IAChB,IAAI,EAAE,SAAS,CAAA;IACf,QAAQ,EAAE,aAAa,CAAA;CACxB,CAAA;AA+ED,eAAO,MAAM,OAAO,EAAE,UAAU,EAy6B/B,CAAA;AAMD,wBAAgB,SAAS,CAAC,EAAE,EAAE,MAAM,GAAG,UAAU,GAAG,SAAS,CAE5D"}

package/dist/model-catalog.js

@@ -437,14 +437,33 @@ export const CATALOG = [
             multipleImages: true,
         },
     },
+    {
+        id: 'gpt-image-2',
+        name: 'GPT Image 2',
+        description: 'State-of-the-art image generation model with token-based pricing. ' +
+            'Cost varies by quality and size; $0.053 is the medium-quality 1024×1024 estimate.',
+        released: '2026-04-21',
+        ...openaiImage,
+        // Token-based pricing: $30/1M output image tokens. Medium quality 1024×1024 ≈ $0.053.
+        cost: { type: 'per-image', perImage: 0.053 },
+        features: {
+            editing: true,
+            inpainting: true,
+            aspectRatios: [],
+            sizes: ['1024x1024', '1536x1024', '1024x1536'],
+            seed: false,
+            multipleImages: true,
+        },
+    },
     {
         id: 'chatgpt-image-latest',
         name: 'ChatGPT Image',
         description: 'Rolling-latest alias that tracks the newest ChatGPT image model. Currently points ' +
-            'to gpt-image-1.5. Pricing may shift when the alias retargets to a newer model.',
+            'to gpt-image-2. Pricing may shift when the alias retargets to a newer model.',
         released: '2026-01',
         ...openaiImage,
-        cost: { type: 'per-image', perImage: 0.034 },
+        // Matches gpt-image-2 medium quality 1024×1024 estimate.
+        cost: { type: 'per-image', perImage: 0.053 },
         features: {
             editing: true,
             inpainting: true,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "egaki",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "type": "module",
   "main": "./dist/cli.js",
   "types": "./dist/cli.d.ts",

package/skills/egaki/SKILL.md

@@ -45,6 +45,15 @@ the key with `egaki login --provider egaki --key egaki_...`.
 When login requires a URL flow, run login in the background and send the login URL
 to the user so they can complete auth interactively.
 
+## Model selection
+
+The `--model` / `-m` flag is **optional** on both `egaki image` and `egaki video`.
+
+- **Interactive (TTY):** omitting `--model` shows a picker with popular models
+- **Non-interactive (piped/scripted):** omitting `--model` uses a sensible default
+
+Agents should always pass `-m` explicitly to avoid the interactive picker.
+
 ## Example commands
 
 ```bash
@@ -60,7 +69,7 @@ egaki subscribe
 # check subscription usage
 egaki usage
 
-# generate an image
+# generate an image (interactive model picker if TTY)
 egaki image "a watercolor fox reading a map" -o fox.png
 
 # select a model explicitly

package/src/cli.ts

@@ -15,6 +15,7 @@ import {
   generateText,
   experimental_generateVideo as aiGenerateVideo,
 } from 'ai'
+import { select, isCancel, cancel } from '@clack/prompts'
 import fs from 'node:fs'
 import path from 'node:path'
 import pc from 'picocolors'
@@ -38,6 +39,7 @@ import {
   createVideoModel,
   shouldUseResponsesApi,
 } from './models.js'
+import { CATALOG } from './model-catalog.js'
 import { VIDEO_CATALOG } from './video-model-catalog.js'
 import {
   loginInteractive,
@@ -225,7 +227,7 @@ cli
   )
   .option(
     '-m, --model [model]',
-    z.enum(IMAGE_MODELS).default(DEFAULT_MODEL).describe('Model ID for generation'),
+    z.enum(IMAGE_MODELS).describe('Model ID for generation. If omitted, shows an interactive picker (or uses default in non-TTY mode)'),
   )
   .option(
     '-o, --output [path]',
@@ -308,10 +310,7 @@ cli
     // Inject stored API keys as env vars before calling the AI SDK.
     injectCredentialsToEnv()
 
-    // goke infers schema .default() values via Zod's input type, which leaves
-    // options.model / options.output / options.count as `T | undefined` even
-    // though the runtime always resolves them. Apply the same defaults here.
-    const model = options.model ?? DEFAULT_MODEL
+    const model = options.model ?? await resolveImageModel()
     const outputPath = options.output ?? 'egaki-output.png'
     const count = options.count ?? 1
     const config = getModelConfig(model)
@@ -391,7 +390,7 @@ cli
   )
   .option(
     '-m, --model [model]',
-    z.enum(VIDEO_MODELS).default(DEFAULT_VIDEO_MODEL).describe('Video model ID for generation'),
+    z.enum(VIDEO_MODELS).describe('Video model ID for generation. If omitted, shows an interactive picker (or uses default in non-TTY mode)'),
   )
   .option(
     '-o, --output [path]',
@@ -455,8 +454,7 @@ cli
   .action(async (prompt, options) => {
     injectCredentialsToEnv()
 
-    // Apply the same .default(...) values goke's input-type inference drops.
-    const model = options.model ?? DEFAULT_VIDEO_MODEL
+    const model = options.model ?? await resolveVideoModel()
     const outputPath = options.output ?? 'egaki-output.mp4'
     const count = options.count ?? 1
     const config = getModelConfig(model)
@@ -515,7 +513,6 @@ cli
   )
   .option('--json', 'Output as JSON instead of YAML')
   .action(async (options) => {
-    const { CATALOG } = await import('./model-catalog.js')
     const yaml = await import('js-yaml')
     const providerStatuses = Object.fromEntries(
       Object.keys(PROVIDERS).map((provider) => [provider, getKeyStatus(provider)]),
@@ -579,6 +576,94 @@ cli.help()
 cli.version(pkg.version)
 cli.parse()
 
+// ─── interactive model pickers ──────────────────────────────────────────────
+
+/** Curated image models shown in the interactive picker, ordered by recommendation. */
+const FEATURED_IMAGE_MODELS = [
+  'imagen-4.0-generate-001',
+  'imagen-4.0-ultra-generate-001',
+  'imagen-4.0-fast-generate-001',
+  'gemini-2.5-flash-image',
+  'gemini-3.1-flash-image-preview',
+  'nano-banana-pro-preview',
+  'gpt-image-2',
+  'gpt-image-1.5',
+  'chatgpt-image-latest',
+  'flux-kontext-pro',
+  'flux-kontext-max',
+  'flux-pro-1.1-ultra',
+  'recraft-v4',
+  'dall-e-3',
+  'grok-imagine-image-pro',
+]
+
+/**
+ * Resolve the image model: show interactive picker in TTY, use default otherwise.
+ */
+async function resolveImageModel(): Promise<string> {
+  const isTTY = process.stdin.isTTY && process.stderr.isTTY
+  if (!isTTY) return DEFAULT_MODEL
+
+  const options = FEATURED_IMAGE_MODELS
+    .map((id) => CATALOG.find((m) => m.id === id))
+    .filter(Boolean)
+    .map((m) => {
+      const cost = m!.cost.type === 'per-image'
+        ? pc.dim(`$${m!.cost.perImage}/img`)
+        : m!.cost.type === 'per-token'
+          ? pc.dim(`$${m!.cost.outputPerM}/M out`)
+          : ''
+      return {
+        value: m!.id,
+        label: `${m!.name} ${cost}`,
+        hint: m!.id,
+      }
+    })
+
+  const selected = await select({
+    message: 'Select an image model',
+    options,
+  })
+
+  if (isCancel(selected)) {
+    cancel('Cancelled.')
+    process.exit(0)
+  }
+
+  return selected
+}
+
+/**
+ * Resolve the video model: show interactive picker in TTY, use default otherwise.
+ */
+async function resolveVideoModel(): Promise<string> {
+  const isTTY = process.stdin.isTTY && process.stderr.isTTY
+  if (!isTTY) return DEFAULT_VIDEO_MODEL
+
+  const options = VIDEO_CATALOG.map((m) => {
+    const dur = m.features.durationRangeSec
+      ? pc.dim(`${m.features.durationRangeSec.min}-${m.features.durationRangeSec.max}s`)
+      : ''
+    return {
+      value: m.id,
+      label: `${m.name} ${dur}`,
+      hint: m.id,
+    }
+  })
+
+  const selected = await select({
+    message: 'Select a video model',
+    options,
+  })
+
+  if (isCancel(selected)) {
+    cancel('Cancelled.')
+    process.exit(0)
+  }
+
+  return selected
+}
+
 function isUrl(input: string): boolean {
   return /^https?:\/\//i.test(input)
 }

package/src/model-catalog.ts

@@ -506,15 +506,35 @@ export const CATALOG: ModelEntry[] = [
       multipleImages: true,
     },
   },
+  {
+    id: 'gpt-image-2',
+    name: 'GPT Image 2',
+    description:
+      'State-of-the-art image generation model with token-based pricing. ' +
+      'Cost varies by quality and size; $0.053 is the medium-quality 1024×1024 estimate.',
+    released: '2026-04-21',
+    ...openaiImage,
+    // Token-based pricing: $30/1M output image tokens. Medium quality 1024×1024 ≈ $0.053.
+    cost: { type: 'per-image', perImage: 0.053 },
+    features: {
+      editing: true,
+      inpainting: true,
+      aspectRatios: [],
+      sizes: ['1024x1024', '1536x1024', '1024x1536'],
+      seed: false,
+      multipleImages: true,
+    },
+  },
   {
     id: 'chatgpt-image-latest',
     name: 'ChatGPT Image',
     description:
       'Rolling-latest alias that tracks the newest ChatGPT image model. Currently points ' +
-      'to gpt-image-1.5. Pricing may shift when the alias retargets to a newer model.',
+      'to gpt-image-2. Pricing may shift when the alias retargets to a newer model.',
     released: '2026-01',
     ...openaiImage,
-    cost: { type: 'per-image', perImage: 0.034 },
+    // Matches gpt-image-2 medium quality 1024×1024 estimate.
+    cost: { type: 'per-image', perImage: 0.053 },
     features: {
       editing: true,
       inpainting: true,