@tasopen/mcp-alphabanana 1.2.1 → 1.3.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.ja.md CHANGED
@@ -2,18 +2,19 @@
2
2
 
3
3
  English | [日本語](README.ja.md)
4
4
 
5
- Google Gemini AI を使って画像アセットを生成する Model Context Protocol (MCP) サーバー。
5
+ Google Gemini AI(Gemini 3.1 Flash/Nano Banana 2対応)で画像アセットを生成する Model Context Protocol (MCP) サーバー。
6
6
 
7
7
  [FastMCP 3](https://www.npmjs.com/package/fastmcp) で構築され、シンプルなコードベースと柔軟な出力オプションを提供します。
8
8
 
9
9
  ## 機能
10
10
 
11
- - **Google Gemini AI による汎用画像生成**
12
- - **カラーキー後処理による透過 PNG 出力**
11
+ - **超高速画像生成**(Gemini 3.1 Flash、0.5K/1K/2K/4K)
12
+ - **高度なマルチ画像推論**(最大14枚の参照画像)
13
+ - **Thinking/Grounding対応**(Flash3.1のみ)
14
+ - **透過PNG/WebP出力**(カラーキー+デスピル)
13
15
  - **複数の出力形式**: ファイル、base64、または両方
14
- - **スタイル参照のための参照画像サポート**
15
16
  - **柔軟なリサイズモード**: crop、stretch、letterbox、contain
16
- - **複数のモデルティア**: flash (Gemini 2.5 Flash) または pro (Gemini 3 Pro)
17
+ - **複数のモデルティア**: Flash3.1、Flash2.5、Pro3、レガシーエイリアス
17
18
 
18
19
  ## インストール
19
20
 
@@ -21,7 +22,7 @@ Google Gemini AI を使って画像アセットを生成する Model Context Pro
21
22
 
22
23
  ## 設定
23
24
 
24
- `GEMINI_API_KEY` を MCP 設定(例: `mcp.json`)で設定します。一部のエージェント環境では OS 環境変数にアクセスできないため、`mcp.json` で OS 環境変数を参照するか、直接キーを記載できます。
25
+ `GEMINI_API_KEY` を MCP 設定(例: `mcp.json`)で設定します。
25
26
 
26
27
  例:
27
28
 
@@ -35,169 +36,50 @@ Google Gemini AI を使って画像アセットを生成する Model Context Pro
35
36
  }
36
37
  ```
37
38
 
38
- - `mcp.json` に直接キーを記載(エージェントが OS 環境変数を参照できない場合に有用):
39
-
40
- ```json
41
- {
42
- "env": {
43
- "GEMINI_API_KEY": "your_api_key_here"
44
- }
45
- }
46
- ```
47
-
48
- 運用に合わせて選択してください(可能であれば環境参照を推奨)。
49
-
50
- ## VS Code 統合
51
-
52
- VS Code 設定(`.vscode/settings.json` またはユーザー設定)に追加し、サーバー `env` を `mcp.json` または VS Code MCP 設定で設定します。OS 環境変数参照または直接キー記載の両方に対応:
39
+ ```bash
40
+ # 開発モード(MCP CLI)
41
+ npm run dev
53
42
 
54
- ```json
55
- {
56
- "mcp": {
57
- "servers": {
58
- "mcp-alphabanana": {
59
- "command": "npx",
60
- "args":["-y", "@tasopen/mcp-alphabanana"],
61
- "env": {
62
- "GEMINI_API_KEY": "${env:GEMINI_API_KEY}" // or "your_api_key_here"
63
- }
64
- }
43
+ # MCP Inspector (Web UI)
65
44
  }
66
- }
67
- }
68
- ```
69
-
70
- **オプション:** 書き込み失敗時のカスタムフォールバックディレクトリを指定する場合は `MCP_FALLBACK_OUTPUT` を `env` に追加してください。
71
-
72
- ## Antigravity (mcp_config.json)
73
45
 
74
- Antigravity ではグローバルな `mcp_config.json` で MCP サーバーを登録します。例:
75
-
76
- ```json
77
- {
78
- "mcpServers": {
79
- "mcp-alphabanana": {
80
- "command": "npx",
81
- "args": ["-y", "@tasopen/mcp-alphabanana"],
82
- "env": {
83
- "GEMINI_API_KEY": "your_api_key_here"
84
- }
85
- }
86
- }
87
- }
46
+ # 本番用ビルド
47
+ npm run build
88
48
  ```
89
49
 
90
- 注: このリポジトリでは `mcp_config.json` に `mcp-alphabanana` エントリ(Antigravity)を追加し、サーバー起動と画像生成の動作を確認しています。
91
-
92
- ## Claude Desktop
93
-
94
- Claude Desktop で MCP サーバーを動かす場合は `claude_desktop_config.json` にエントリを追加:
50
+ ## ライセンス
95
51
 
96
- ```json
97
- {
98
- "mcpServers": {
99
- "mcp-alphabanana": {
100
- "command": "npx",
101
- "args": ["-y", "@tasopen/mcp-alphabanana"],
102
- "env": {
103
- "GEMINI_API_KEY": "your_api_key_here"
104
- }
105
- }
106
- }
52
+ MIT
53
+ ]
107
54
  }
108
55
  ```
109
56
 
110
- テスト済み: 上記エントリを Claude Desktop に追加しサーバーを起動すると、MCP サーバーが起動し画像生成が動作しました。
111
-
112
- ### 環境変数
113
-
114
- | 変数 | 必須 | 説明 |
115
- |----------|----------|-------------|
116
- | `GEMINI_API_KEY` | Yes | Google AI Studio APIキー。`mcp.json` で OS 環境変数参照(`${env:GEMINI_API_KEY}`)または直接キー記載(エージェントが OS 環境変数にアクセス不可な場合) |
117
- | `MCP_FALLBACK_OUTPUT` | No | 書き込み失敗時のフォールバックディレクトリ(デフォルト: `<install-dir>/fallback-output`) |
118
-
119
- ### 出力パスのベストプラクティス
57
+ ---
120
58
 
121
- **常に `outputPath` には絶対パスを使ってください:**
59
+ ## 透過・出力形式
122
60
 
123
- 良い例: "C:/Users/you/project/assets", "/home/user/images"
124
- 悪い例: `"./"`, `"output/"`, `"../images"`
61
+ - **PNG**: 完全アルファ、カラーキー+デスピル
62
+ - **WebP**: 完全アルファ、高圧縮(Flash3.1+)
63
+ - **JPEG**: 透過なし(不透明背景にフォールバック)
125
64
 
126
- 相対パスは MCP サーバーの作業ディレクトリから解決されるため、サービス実行時に予期しない場所にファイルが作成される可能性があります。
127
-
128
- **フォールバック動作:**
129
- - 指定した `outputPath` が書き込み可能 → 通常通り画像を保存
130
- - 書き込み不可(パーミッション拒否など) → フォールバックディレクトリに保存しレスポンスに `warning` を含める
131
- - フォールバックも失敗した場合 → エラーを返す
65
+ ---
132
66
 
133
67
  ## 開発
134
68
 
135
69
  ```bash
136
70
  # 開発モード(MCP CLI)
137
- npm run dev
71
+ #### 透過 + フリンジ制御
138
72
 
139
73
  # MCP Inspector (Web UI)
140
- npm run inspect
74
+
141
75
 
142
76
  # 本番用ビルド
143
77
  npm run build
144
78
  ```
145
79
 
146
- ## ツール: generate_image
147
-
148
- Gemini AI で画像アセットを生成(オプションで透過・参照画像対応)。
149
-
150
- ### パラメータ
151
-
152
- | パラメータ | 型 | デフォルト | 説明 |
153
- |-----------|------|---------|-------------|
154
- | `prompt` | string | *必須* | 生成する画像の説明 |
155
- | `outputFileName` | string | *必須* | 出力ファイル名(拡張子がない場合は自動追加) |
156
- | `outputType` | enum | `combine` | 出力形式: `file`、`base64`、または`combine` |
157
- | `modelTier` | enum | *必須* | モデル: `flash` (Gemini 2.5 Flash, 最大3参照) または `pro` (Gemini 3 Pro, 最大14参照) |
158
- | `sourceResolution` | enum | `1K` | Gemini生成解像度: `1K`、`2K`、または`4K` (2K/4Kはproのみ) |
159
- | `outputWidth` | number | `1024` | 出力幅(ピクセル、8-4096) |
160
- | `outputHeight` | number | `1024` | 出力高さ(ピクセル、8-4096) |
161
- | `outputFormat` | enum | `png` | 出力形式: `png`または`jpg` |
162
- | `outputPath` | string | *任意* | 絶対パスの出力ディレクトリ(ファイル保存時は必須) |
163
- | `transparent` | boolean | `false` | 透過背景をリクエスト(PNGのみ) |
164
- | `transparentColor` | string | `null` | 透過にする色(例: `#FF00FF`、未指定時は #FF00FF) |
165
- | `colorTolerance` | number | `30` | 透過色マッチングの許容範囲(0-255) |
166
- | `fringeMode` | enum | `auto` | フリンジ処理: `auto`、`crisp`、`hd`(autoは128px以下で`crisp`、それ以外は`hd`) |
167
- | `resizeMode` | enum | `crop` | リサイズモード: `crop`、`stretch`、`letterbox`、または`contain` |
168
- | `referenceImages` | array | `[]` | スタイルガイダンス用の参照画像(ファイルパス) |
169
- | `debug` | boolean | `false` | デバッグモード: 中間画像を出力 |
170
-
171
- ### 使用例
172
-
173
- #### 基本的な生成
174
-
175
- ```json
176
- {
177
- "prompt": "ピクセルアートの宝箱、金の装飾、木の質感",
178
- "modelTier": "flash",
179
- "outputFileName": "chest",
180
- "outputWidth": 64,
181
- "outputHeight": 64,
182
- "transparent": true
183
- }
184
- ```
185
-
186
- #### 透過スプライト
187
-
188
- ```json
189
- {
190
- "prompt": "レトロゲーム風の宇宙戦闘機スプライト、青みがかった銀色、輪郭がはっきり、背景なし、文字・数字・ロゴを含めない",
191
- "modelTier": "flash",
192
- "outputFileName": "space_fighter",
193
- "outputWidth": 64,
194
- "outputHeight": 64,
195
- "transparent": true
196
- }
197
- ```
198
-
199
- #### 透過 + フリンジ制御
80
+ ## ライセンス
200
81
 
82
+ MIT
201
83
  ```json
202
84
  {
203
85
  "prompt": "アニメ調 自転車に乗った少女",
package/README.md CHANGED
@@ -2,26 +2,27 @@
2
2
 
3
3
  English | [日本語](README.ja.md)
4
4
 
5
- A Model Context Protocol (MCP) server for generating image assets using Google Gemini AI.
5
+ A Model Context Protocol (MCP) server for generating image assets using Google Gemini AI (Gemini 3.1 Flash/Nano Banana 2 supported).
6
6
 
7
7
  Built with [FastMCP 3](https://www.npmjs.com/package/fastmcp) for a simplified codebase and flexible output options.
8
8
 
9
9
  ## Features
10
10
 
11
- - **General-purpose image generation** using Google Gemini AI
12
- - **Transparent PNG output** via color-key post-processing
11
+ - **Ultra-fast image generation** (Gemini 3.1 Flash, 0.5K/1K/2K/4K)
12
+ - **Advanced multi-image reasoning** (up to 14 reference images)
13
+ - **Thinking/Grounding support** (Flash3.1 only)
14
+ - **Transparent PNG/WebP output** (color-key post-processing, despill)
13
15
  - **Multiple output formats**: file, base64, or both
14
- - **Reference image support** for style guidance
15
16
  - **Flexible resize modes**: crop, stretch, letterbox, contain
16
- - **Multiple model tiers**: flash (Gemini 2.5 Flash) or pro (Gemini 3 Pro)
17
+ - **Multiple model tiers**: Flash3.1, Flash2.5, Pro3, legacy aliases
17
18
 
18
19
  ## Installation
19
20
 
20
- add `@tasopen/mcp-alphabanana` to your MCP Servers configuration.
21
+ Add `@tasopen/mcp-alphabanana` to your MCP Servers configuration.
21
22
 
22
23
  ## Configuration
23
24
 
24
- Configure the `GEMINI_API_KEY` in your MCP configuration (e.g. `mcp.json`). Some agent environments cannot access OS environment variables, so you can either reference an OS environment variable from `mcp.json` or provide the key directly in `mcp.json`.
25
+ Configure the `GEMINI_API_KEY` in your MCP configuration (e.g. `mcp.json`).
25
26
 
26
27
  Examples:
27
28
 
@@ -35,7 +36,7 @@ Examples:
35
36
  }
36
37
  ```
37
38
 
38
- - Provide the key directly in `mcp.json` (useful when the agent cannot access OS env vars):
39
+ - Provide the key directly in `mcp.json`:
39
40
 
40
41
  ```json
41
42
  {
@@ -45,11 +46,9 @@ Examples:
45
46
  }
46
47
  ```
47
48
 
48
- Use whichever method suits your deployment — prefer environment references when possible, but include the key directly when agent constraints require it.
49
-
50
49
  ## VS Code Integration
51
50
 
52
- Add to your VS Code settings (`.vscode/settings.json` or user settings), configuring the server `env` in `mcp.json` or via the VS Code MCP settings. You can reference an OS environment variable or put the key directly:
51
+ Add to your VS Code settings (`.vscode/settings.json` or user settings), configuring the server `env` in `mcp.json` or via the VS Code MCP settings.
53
52
 
54
53
  ```json
55
54
  {
@@ -69,114 +68,40 @@ Add to your VS Code settings (`.vscode/settings.json` or user settings), configu
69
68
 
70
69
  **Optional:** Set a custom fallback directory for write failures by adding `MCP_FALLBACK_OUTPUT` to the `env` object.
71
70
 
72
- ## Antigravity (mcp_config.json)
73
-
74
- Antigravity uses a global `mcp_config.json` to register MCP servers. Example `mcp_config.json`:
75
-
76
- ```json
77
- {
78
- "mcpServers": {
79
- "mcp-alphabanana": {
80
- "command": "npx",
81
- "args": ["-y", "@tasopen/mcp-alphabanana"],
82
- "env": {
83
- "GEMINI_API_KEY": "your_api_key_here"
84
- }
85
- }
86
- }
87
- }
88
- ```
89
-
90
- Note: In this repository we used an `mcp_config.json` with the `mcp-alphabanana` entry (Antigravity) and confirmed the server started and generated images successfully.
91
-
92
- ## Claude Desktop
93
-
94
- If you run MCP servers via Claude Desktop, add an entry in `claude_desktop_config.json`:
95
-
96
- ```json
97
- {
98
- "mcpServers": {
99
- "mcp-alphabanana": {
100
- "command": "npx",
101
- "args": ["-y", "@tasopen/mcp-alphabanana"],
102
- "env": {
103
- "GEMINI_API_KEY": "your_api_key_here"
104
- }
105
- }
106
- }
107
- }
108
- ```
109
-
110
- Tested: Adding the above entry in Claude Desktop and starting the server launched the MCP server and image generation worked.
111
-
112
- ### Environment Variables
113
-
114
- | Variable | Required | Description |
115
- |----------|----------|-------------|
116
- | `GEMINI_API_KEY` | Yes | Google AI Studio API Key. Configure in `mcp.json` either by referencing an OS environment variable (`${env:GEMINI_API_KEY}`) or by including the key directly (some agents cannot access OS env vars). |
117
- | `MCP_FALLBACK_OUTPUT` | No | Fallback directory for write failures (default: `<install-dir>/fallback-output`) |
118
-
119
- ### Output Path Best Practices
120
-
121
- **Always use absolute paths for `outputPath`:**
122
-
123
- ✅ **Good:** "C:/Users/you/project/assets", "/home/user/images"
124
- ❌ **Bad:** `"./"`, `"output/"`, `"../images"`
125
-
126
- Relative paths are resolved from the MCP server's working directory (unpredictable when running as a service), leading to unexpected file locations.
127
-
128
- **Fallback behavior:**
129
- - If the requested `outputPath` is writable → image saved there as usual
130
- - If not writable (permission denied, etc.) → saves to fallback directory with `warning` in response
131
- - Fallback directory: `MCP_FALLBACK_OUTPUT` env var, or `<install-dir>/fallback-output` by default
132
- - If fallback also fails → returns error
133
-
134
- ## Development
135
-
136
- ```bash
137
- # Development mode with MCP CLI
138
- npm run dev
139
-
140
- # MCP Inspector (Web UI)
141
- npm run inspect
142
-
143
- # Build for production
144
- npm run build
145
- ```
146
-
147
- ## Tool: generate_image
71
+ ## Model Selection & Parameters
148
72
 
149
- Generate image assets using Gemini AI with optional transparency and reference images.
73
+ | Input Model ID | Internal Model ID | Description |
74
+ | --- | --- | --- |
75
+ | `Flash3.1` | `gemini-3.1-flash-image-preview` | Ultra-fast, supports Thinking/Grounding. |
76
+ | `Flash2.5` | `gemini-2.5-flash-image` | Legacy Flash. High stability. Low cost. |
77
+ | `Pro3` | `gemini-3.0-pro-image-preview` | High-fidelity Pro model. |
78
+ | `flash` | `gemini-3.1-flash-image-preview` | Alias for backward compatibility. |
79
+ | `pro` | `gemini-3.0-pro-image-preview` | Alias for backward compatibility. |
150
80
 
151
- ### Parameters
81
+ ### Parameters (v2.0)
152
82
 
153
83
  | Parameter | Type | Default | Description |
154
84
  |-----------|------|---------|-------------|
155
85
  | `prompt` | string | *required* | Description of the image to generate |
156
- | `outputFileName` | string | *required* | Output filename (extension auto-added if missing) |
157
- | `outputType` | enum | `combine` | Output format: `file`, `base64`, or `combine` |
158
- | `modelTier` | enum | *required* | Model: `flash` (Gemini 2.5 Flash, max 3 refs) or `pro` (Gemini 3 Pro, max 14 refs) |
159
- | `sourceResolution` | enum | `1K` | Gemini source resolution: `1K`, `2K`, or `4K` (2K/4K pro-only) |
160
- | `outputWidth` | number | `1024` | Output width in pixels (8-4096) |
161
- | `outputHeight` | number | `1024` | Output height in pixels (8-4096) |
162
- | `outputFormat` | enum | `png` | Output format: `png` or `jpg` |
163
- | `outputPath` | string | *optional* | Absolute output directory path (required when saving files) |
164
- | `transparent` | boolean | `false` | Request transparent background (PNG only) |
165
- | `transparentColor` | string | `null` | Color to make transparent (e.g., `#FF00FF`), defaults to `#FF00FF` when null |
166
- | `colorTolerance` | number | `30` | Tolerance for transparent color matching (0-255) |
167
- | `fringeMode` | enum | `auto` | Fringe reduction mode: `auto`, `crisp`, `hd` (auto uses `crisp` <= 128px, otherwise `hd`) |
168
- | `resizeMode` | enum | `crop` | Resize mode: `crop`, `stretch`, `letterbox`, or `contain` |
169
- | `referenceImages` | array | `[]` | Reference images for style guidance (file paths) |
170
- | `debug` | boolean | `false` | Debug mode: output intermediate images |
171
-
172
- ### Usage Examples
86
+ | `model` | enum | `Flash3.1` | Model: `Flash3.1`, `Flash2.5`, `Pro3`, `flash`, `pro` |
87
+ | `output_resolution` | enum | `1K` | `0.5K`, `1K`, `2K`, `4K` (0.5K/2K/4K: Flash3.1 only) |
88
+ | `output_format` | enum | `png` | `png`, `jpg`, `webp` (WebP: alpha supported) |
89
+ | `transparent` | boolean | `false` | Transparent background (PNG/WebP only) |
90
+ | `grounding_type` | enum | `none` | `none`, `text`, `image`, `both` (Flash3.1 only) |
91
+ | `thinking_mode` | enum | `minimal` | `minimal`, `high` (Flash3.1 only) |
92
+ | `include_thoughts` | boolean | `false` | Return model's "thoughts" (Flash3.1 only) |
93
+ | `reference_images` | array | `[]` | Up to 14 (Flash3.1/Pro3), 3 (Flash2.5) |
94
+
95
+ ---
96
+
97
+ ## Usage Examples
173
98
 
174
99
  #### Basic Generation
175
100
 
176
101
  ```json
177
102
  {
178
103
  "prompt": "A pixel art treasure chest, golden trim, wooden texture",
179
- "modelTier": "flash",
104
+ "model": "Flash3.1",
180
105
  "outputFileName": "chest",
181
106
  "outputWidth": 64,
182
107
  "outputHeight": 64,
@@ -184,113 +109,66 @@ Generate image assets using Gemini AI with optional transparency and reference i
184
109
  }
185
110
  ```
186
111
 
187
-
188
- #### Transparent Sprite
189
-
190
- ```json
191
- {
192
- "prompt": "Retro game-style space fighter sprite, blue-tinted silver, clean silhouette, no background, no text, no letters, no numbers, no logos",
193
- "modelTier": "flash",
194
- "outputFileName": "space_fighter",
195
- "outputWidth": 64,
196
- "outputHeight": 64,
197
- "transparent": true
198
- }
199
- ```
200
-
201
-
202
- #### Transparency + Fringe Control
203
-
204
- ```json
205
- {
206
- "prompt": "Anime-style girl riding a bicycle",
207
- "modelTier": "flash",
208
- "outputFileName": "bicycle_girl",
209
- "outputWidth": 1024,
210
- "outputHeight": 576,
211
- "transparent": true,
212
- "colorTolerance": 30,
213
- "fringeMode": "crisp"
214
- }
215
- ```
216
-
217
-
218
- #### High-Resolution Background
112
+ #### Advanced (WebP, thinking, grounding)
219
113
 
220
114
  ```json
221
115
  {
222
- "prompt": "Fantasy forest at twilight with glowing mushrooms",
223
- "outputFileName": "forest_bg",
224
- "modelTier": "pro",
225
- "sourceResolution": "4K",
226
- "outputWidth": 3840,
227
- "outputHeight": 2160,
228
- "outputFormat": "jpg",
229
- "outputPath": "C:/Users/you/project/assets/backgrounds"
116
+ "prompt": "A photorealistic girl with wings flying over a European countryside",
117
+ "model": "Flash3.1",
118
+ "outputFileName": "girl_wings",
119
+ "outputWidth": 632,
120
+ "outputHeight": 424,
121
+ "output_format": "webp",
122
+ "thinking_mode": "high",
123
+ "grounding_type": "both",
124
+ "include_thoughts": true
230
125
  }
231
126
  ```
232
127
 
233
- #### With Reference Image
128
+ #### With Reference Images
234
129
 
235
130
  ```json
236
131
  {
237
- "prompt": "A matching treasure chest, open state, same pixel art style",
238
- "modelTier": "flash",
132
+ "prompt": "A matching treasure chest, open state, same pixel art style as references",
133
+ "model": "Pro3",
239
134
  "outputFileName": "chest_open",
240
135
  "outputWidth": 64,
241
136
  "outputHeight": 64,
242
137
  "transparent": true,
243
- "referenceImages": [
138
+ "reference_images": [
244
139
  {
245
140
  "description": "Closed chest for style reference",
246
- "filePath": "C:/Users/you/project/assets/references/chest_closed.png"
141
+ "data": "...base64..."
247
142
  }
248
143
  ]
249
144
  }
250
145
  ```
251
146
 
252
- ## Output Types
253
-
254
- | Type | Saves File | Returns Base64 | MCP Image Content |
255
- |------|-----------|----------------|-------------------|
256
- | `file` | ✓ | ✗ | ✗ |
257
- | `base64` | ✗ | ✓ | ✓ |
258
- | `combine` | ✓ | ✓ | ✓ |
147
+ ---
259
148
 
260
- ## Transparency Processing
149
+ ## Transparency & Output Formats
261
150
 
262
- The server selects the background color by histogram analysis and hue proximity to the requested key color, then applies RGB-distance keying and despill. If no histogram candidate qualifies, it falls back to the closest-hue corner color.
151
+ - **PNG**: Full alpha, color-key + despill
152
+ - **WebP**: Full alpha, better compression (Flash3.1+)
153
+ - **JPEG**: No transparency (falls back to solid background)
263
154
 
264
- ### Model Notes
155
+ ---
265
156
 
266
- - Flash is sufficient for transparent PNG output in typical use.
267
- - `colorTolerance` around 30 performed best in tests; higher values can increase false positives.
157
+ ## Development
268
158
 
269
- ### Recommended Background Colors
159
+ ```bash
160
+ # Development mode with MCP CLI
161
+ npm run dev
270
162
 
271
- | Color | Hex | Best For |
272
- |-------|-----|----------|
273
- | Magenta | `#FF00FF` | Most sprites (default, works with both models) |
274
- | Green | `#00FF00` | Purple/pink objects |
275
- | Cyan | `#00FFFF` | Red/orange objects |
276
- | Blue | `#0000FF` | Yellow/green objects |
277
163
 
278
- ### Examples
164
+ # MCP Inspector (Web UI)
165
+ npm run inspect
279
166
 
280
- ```json
281
- {
282
- "transparent": true,
283
- "transparentColor": "#00FF00",
284
- "colorTolerance": 30
285
- }
167
+ # Build for production
168
+ npm run build
286
169
  ```
287
170
 
288
- ### Fringe Mode Guidance
289
-
290
- - Use `crisp` when thin lines risk being removed (pixel art, bicycle spokes, wire meshes).
291
- - Use `hd` for general high-resolution images where fringe is noticeable.
292
- - Use `auto` for a size-based default (`crisp` for <= 128px, `hd` otherwise).
293
-
294
171
  ## License
295
172
 
296
173
  MIT
174
+