@haposoft/cafekit 0.13.4 → 0.14.1

package/src/claude/skills/ai-multimodal/references/audio-processing.md

@@ -42,7 +42,7 @@ myfile = client.files.upload(file='meeting.mp3')
 
 # Transcribe
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Generate a transcript of the speech.', myfile]
 )
 print(response.text)
@@ -52,7 +52,7 @@ print(response.text)
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Generate transcript with timestamps in MM:SS format.', myfile]
 )
 ```
@@ -61,7 +61,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Transcribe with speaker labels. Format: [Speaker 1], [Speaker 2], etc.', myfile]
 )
 ```
@@ -70,7 +70,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Transcribe only the segment from 02:30 to 05:15.', myfile]
 )
 ```
@@ -81,7 +81,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Summarize key points in 5 bullets with timestamps.', myfile]
 )
 ```
@@ -91,19 +91,19 @@ response = client.models.generate_content(
 ```python
 # Music analysis
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Identify the musical instruments and genre.', myfile]
 )
 
 # Environmental sounds
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Identify all sounds: voices, music, ambient noise.', myfile]
 )
 
 # Birdsong identification
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Identify bird species based on their calls.', myfile]
 )
 ```
@@ -112,7 +112,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['What is discussed from 10:30 to 15:45? Provide key points.', myfile]
 )
 ```
@@ -127,13 +127,13 @@ myfile = client.files.upload(file='large-audio.mp3')
 
 # First query
 response1 = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Transcribe this', myfile]
 )
 
 # Second query (reuses same file)
 response2 = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Summarize this', myfile]
 )
 ```
@@ -147,7 +147,7 @@ with open('small-audio.mp3', 'rb') as f:
     audio_bytes = f.read()
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Describe this audio',
         types.Part.from_bytes(data=audio_bytes, mime_type='audio/mp3')
@@ -161,14 +161,14 @@ response = client.models.generate_content(
 
 | Model | Quality | Speed | Cost/1M tokens |
 |-------|---------|-------|----------------|
-| `gemini-2.5-flash-native-audio-preview-09-2025` | High | Fast | $10 |
-| `gemini-2.5-pro` TTS mode | Premium | Slower | $20 |
+| `gemma-4-31b-it-native-audio-preview-09-2025` | High | Fast | $10 |
+| `gemma-4-31b-it` TTS mode | Premium | Slower | $20 |
 
 ### Basic TTS
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash-native-audio-preview-09-2025',
+    model='gemma-4-31b-it-native-audio-preview-09-2025',
     contents='Generate audio: Welcome to today\'s episode.'
 )
 
@@ -182,19 +182,19 @@ with open('output.wav', 'wb') as f:
 ```python
 # Professional tone
 response = client.models.generate_content(
-    model='gemini-2.5-flash-native-audio-preview-09-2025',
+    model='gemma-4-31b-it-native-audio-preview-09-2025',
     contents='Generate audio in a professional, clear tone: Welcome to our quarterly earnings call.'
 )
 
 # Casual and friendly
 response = client.models.generate_content(
-    model='gemini-2.5-flash-native-audio-preview-09-2025',
+    model='gemma-4-31b-it-native-audio-preview-09-2025',
     contents='Generate audio in a friendly, conversational tone: Hey there! Let\'s dive into today\'s topic.'
 )
 
 # Narrative style
 response = client.models.generate_content(
-    model='gemini-2.5-flash-native-audio-preview-09-2025',
+    model='gemma-4-31b-it-native-audio-preview-09-2025',
     contents='Generate audio in a narrative, storytelling tone: Once upon a time, in a land far away...'
 )
 ```
@@ -244,9 +244,7 @@ response = client.models.generate_content(
 - 9.5 hours = 1,094,400 tokens
 
 **Model selection**:
-- Use `gemini-2.5-flash` ($1/1M tokens) for most tasks
-- Upgrade to `gemini-2.5-pro` ($3/1M tokens) for complex analysis
-- For high-volume: `gemini-1.5-flash` ($0.70/1M tokens)
+- Use `gemma-4-31b-it` for transcription and analysis (read model from `.claude/runtime.json`: `gemini.model`)
 
 **Reduce costs**:
 - Process only relevant segments using timestamps
@@ -265,7 +263,7 @@ def transcribe_with_retry(file_path, max_retries=3):
         try:
             myfile = client.files.upload(file=file_path)
             response = client.models.generate_content(
-                model='gemini-2.5-flash',
+                model='gemma-4-31b-it',
                 contents=['Transcribe with timestamps', myfile]
             )
             return response.text
@@ -283,7 +281,7 @@ def transcribe_with_retry(file_path, max_retries=3):
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Transcribe this meeting with:
         1. Speaker labels
@@ -299,7 +297,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Create podcast summary with:
         1. Main topics with timestamps
@@ -315,7 +313,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Analyze interview:
         1. Questions asked with timestamps
@@ -331,7 +329,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Verify audio content:
         1. Check for specific keywords or phrases
@@ -348,7 +346,7 @@ response = client.models.generate_content(
 ```python
 # Gemini auto-detects language
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Transcribe this audio and translate to English if needed.', myfile]
 )
 ```

package/src/claude/skills/ai-multimodal/references/video-analysis.md

@@ -22,12 +22,8 @@ Comprehensive guide for video understanding, temporal analysis, and YouTube proc
 
 ## Model Selection
 
-### Gemini 3 Series (Latest)
-- **gemini-3-pro-preview**: Latest, agentic workflows, 1M context, dynamic thinking
-
-### Gemini 2.5 Series (Recommended)
-- **gemini-2.5-pro**: Best quality, 1M-2M context
-- **gemini-2.5-flash**: Balanced, 1M-2M context (recommended)
+### Recommended
+- **gemma-4-31b-it**: 1M-2M context, balanced quality/speed (read model from `.claude/runtime.json`: `gemini.model`)
 
 ### Context Windows
 - **2M token models**: ~2 hours (default) or ~6 hours (low-res)
@@ -57,7 +53,7 @@ if myfile.state.name == 'FAILED':
 
 # Analyze
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Summarize this video in 3 key points', myfile]
 )
 print(response.text)
@@ -69,7 +65,7 @@ print(response.text)
 from google.genai import types
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Summarize the main topics discussed',
         types.Part.from_uri(
@@ -87,7 +83,7 @@ with open('short-clip.mp4', 'rb') as f:
     video_bytes = f.read()
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'What happens in this video?',
         types.Part.from_bytes(data=video_bytes, mime_type='video/mp4')
@@ -102,7 +98,7 @@ response = client.models.generate_content(
 ```python
 # Analyze specific time range
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Summarize this segment',
         types.Part.from_video_metadata(
@@ -119,7 +115,7 @@ response = client.models.generate_content(
 ```python
 # Lower FPS for static content (saves tokens)
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Analyze this presentation',
         types.Part.from_video_metadata(
@@ -131,7 +127,7 @@ response = client.models.generate_content(
 
 # Higher FPS for fast-moving content
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Analyze rapid movements in this sports video',
         types.Part.from_video_metadata(
@@ -155,7 +151,7 @@ for video in [video1, video2]:
         video = client.files.get(name=video.name)
 
 response = client.models.generate_content(
-    model='gemini-2.5-pro',
+    model='gemma-4-31b-it',
     contents=[
         'Compare these two product demos. Which explains features better?',
         video1,
@@ -170,7 +166,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'What happens at 01:15 and how does it relate to 02:30?',
         myfile
@@ -182,7 +178,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Create a timeline with timestamps:
         - Key events
@@ -199,7 +195,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Identify all scene changes with timestamps and describe each scene',
         myfile
@@ -213,7 +209,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Transcribe the audio from this video',
         myfile
@@ -225,7 +221,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Transcribe with visual context:
         - Audio transcription
@@ -241,7 +237,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Transcribe with speaker labels and timestamps',
         myfile
@@ -255,7 +251,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Summarize this video:
         1. Main topic and purpose
@@ -271,7 +267,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Create educational materials:
         1. List key concepts taught
@@ -287,7 +283,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'List all actions performed in this tutorial with timestamps',
         myfile
@@ -299,7 +295,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Review video content:
         1. Identify any problematic content
@@ -315,7 +311,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Analyze interview:
         1. Questions asked (timestamps)
@@ -332,7 +328,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Analyze sports video:
         1. Key plays with timestamps
@@ -362,7 +358,7 @@ response = client.models.generate_content(
 youtube_uri = 'https://www.youtube.com/watch?v=dQw4w9WgXcQ'
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Create chapter markers with timestamps',
         types.Part.from_uri(uri=youtube_uri, mime_type='video/mp4')
@@ -444,7 +440,7 @@ class VideoAnalysis(BaseModel):
     duration: str
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Analyze this video', myfile],
     config=genai.types.GenerateContentConfig(
         response_mime_type='application/json',

package/src/claude/skills/ai-multimodal/references/vision-understanding.md

@@ -25,9 +25,9 @@ Comprehensive guide for image analysis, object detection, and visual understandi
 ## Model Selection
 
 ### Gemini 2.5 Series
-- **gemini-2.5-pro**: Best quality, segmentation + detection
-- **gemini-2.5-flash**: Fast, efficient, all features
-- **gemini-2.5-flash-lite**: Lightweight, all features
+- **gemma-4-31b-it**: Best quality, segmentation + detection
+- **gemma-4-31b-it**: Fast, efficient, all features
+- **gemma-4-31b-it-lite**: Lightweight, all features
 
 ### Feature Requirements
 - **Segmentation**: Requires 2.5+ models
@@ -49,7 +49,7 @@ with open('image.jpg', 'rb') as f:
     img_bytes = f.read()
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Describe this image in detail',
         genai.types.Part.from_bytes(data=img_bytes, mime_type='image/jpeg')
@@ -62,7 +62,7 @@ print(response.text)
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Classify this image. Provide category and confidence level.',
         img_part
@@ -74,7 +74,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'How many people are in this image and what are they doing?',
         img_part
@@ -88,7 +88,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Detect all objects in this image and provide bounding boxes',
         img_part
@@ -103,7 +103,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Create a segmentation mask for all people in this image',
         img_part
@@ -122,7 +122,7 @@ img1 = PIL.Image.open('photo1.jpg')
 img2 = PIL.Image.open('photo2.jpg')
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Compare these two images. What are the differences?',
         img1,
@@ -135,7 +135,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Extract all visible text from this image',
         img_part
@@ -155,7 +155,7 @@ with open('image.jpg', 'rb') as f:
     img_bytes = f.read()
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Analyze this image',
         types.Part.from_bytes(data=img_bytes, mime_type='image/jpeg')
@@ -171,7 +171,7 @@ import PIL.Image
 img = PIL.Image.open('photo.jpg')
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['What is in this image?', img]
 )
 ```
@@ -184,12 +184,12 @@ myfile = client.files.upload(file='large-image.jpg')
 
 # Use multiple times
 response1 = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Describe this image', myfile]
 )
 
 response2 = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['What colors dominate this image?', myfile]
 )
 ```
@@ -198,7 +198,7 @@ response2 = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Analyze this image',
         types.Part.from_uri(
@@ -250,7 +250,7 @@ class ImageAnalysis(BaseModel):
     scene_type: str
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Analyze this image', img_part],
     config=genai.types.GenerateContentConfig(
         response_mime_type='application/json',
@@ -272,7 +272,7 @@ images = [
 ]
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=['Analyze these images and find common themes'] + images
 )
 ```
@@ -284,7 +284,7 @@ before = PIL.Image.open('before.jpg')
 after = PIL.Image.open('after.jpg')
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Compare before and after. List all visible changes.',
         before,
@@ -300,7 +300,7 @@ reference = PIL.Image.open('target.jpg')
 candidates = [PIL.Image.open(f'option{i}.jpg') for i in range(5)]
 
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Find which candidate images contain objects similar to the reference',
         reference
@@ -333,7 +333,7 @@ response = client.models.generate_content(
 **Few-shot examples**:
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Example: For an image of a cat on a sofa, respond: "Object: cat, Location: sofa"',
         'Now analyze this image:',
@@ -371,7 +371,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Analyze this product image:
         1. Identify the product
@@ -388,7 +388,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Extract all text and UI elements from this screenshot',
         img_part
@@ -400,7 +400,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-pro',
+    model='gemma-4-31b-it',
     contents=[
         'Describe visible features in this medical image. Note: This is for informational purposes only.',
         img_part
@@ -412,7 +412,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         'Extract data from this chart and format as JSON',
         img_part
@@ -424,7 +424,7 @@ response = client.models.generate_content(
 
 ```python
 response = client.models.generate_content(
-    model='gemini-2.5-flash',
+    model='gemma-4-31b-it',
     contents=[
         '''Analyze this scene:
         1. Location type
@@ -451,7 +451,7 @@ def analyze_image_with_retry(image_path, prompt, max_retries=3):
                 img_bytes = f.read()
 
             response = client.models.generate_content(
-                model='gemini-2.5-flash',
+                model='gemma-4-31b-it',
                 contents=[
                     prompt,
                     genai.types.Part.from_bytes(

package/src/claude/skills/ai-multimodal/scripts/check_setup.py

@@ -302,7 +302,7 @@ def main():
         print("\nImage generation models:")
         print("  • gemma-4-31b-it-image    - Nano Banana Flash (DEFAULT - fast)")
         print("  • imagen-4.0-generate-001   - Imagen 4 (alternative - production)")
-        print("  • gemini-3-pro-image-preview - Nano Banana Pro (4K text, reasoning)")
+        print("  • gemma-4-31b-it-image-preview - Nano Banana Pro (4K text, reasoning)")
         print("\nExample (uses default model):")
         print("  python scripts/gemini_batch_process.py --task generate \\")
         print("    --prompt 'A sunset over mountains' --aspect-ratio 16:9 --size 2K")

package/src/claude/skills/ai-multimodal/scripts/gemini_batch_process.py

@@ -8,8 +8,8 @@ Supports all Gemini modalities:
 - Video: Summarization, Q&A, scene detection
 - Document: PDF extraction, structured output
 - Generation: Image creation via Imagen 4 or Nano Banana (Gemini native)
-  - Nano Banana 2 (gemini-3.1-flash-image-preview): Fastest, 95% Pro quality (default)
-  - Nano Banana Pro (gemini-3-pro-image-preview): Quality/4K text/reasoning
+  - Nano Banana 2 (gemma-4-31b-it-image): Fastest, 95% Pro quality (default)
+  - Nano Banana Pro (gemma-4-31b-it-image-preview): Quality/4K text/reasoning
   - Imagen 4 (imagen-4.0-*): Production-grade generation
 """
 
@@ -60,10 +60,10 @@ except ImportError:
 
 
 # Image generation model configuration
-# Default: gemini-3.1-flash-image-preview (Nano Banana 2 - 3-5x faster, 95% Pro quality)
+# Default: gemma-4-31b-it-image (Nano Banana 2 - 3-5x faster, 95% Pro quality)
 # Alternative: imagen-4.0-generate-001 (production quality)
 # All image generation requires billing - no completely free option exists
-IMAGE_MODEL_DEFAULT = 'gemini-3.1-flash-image-preview'  # Nano Banana 2 (fastest, near-Pro quality)
+IMAGE_MODEL_DEFAULT = 'gemma-4-31b-it-image'  # Nano Banana 2 (fastest, near-Pro quality)
 IMAGE_MODEL_FALLBACK = 'gemma-4-31b-it-image'  # Fallback if Nano Banana 2 fails
 IMAGEN_MODELS = {
     'imagen-4.0-generate-001',
@@ -138,7 +138,7 @@ def get_default_model(task: str) -> str:
             return model
         # Default to Nano Banana 2 (fastest, near-Pro quality)
         # Alternative: imagen-4.0-generate-001 for production quality
-        return 'gemini-3.1-flash-image-preview'
+        return 'gemma-4-31b-it-image'
 
     elif task == 'generate-video':
         model = os.getenv('VIDEO_GEN_MODEL')
@@ -180,8 +180,6 @@ def validate_model_task_combination(model: str, task: str) -> None:
             'imagen-4.0-generate-001',
             'imagen-4.0-ultra-generate-001',
             'imagen-4.0-fast-generate-001',
-            'gemini-3.1-flash-image-preview',
-            'gemini-3-pro-image-preview',
             'gemma-4-31b-it-image',
             'gemma-4-31b-it-image-preview',
         ]
@@ -1093,7 +1091,7 @@ Examples:
 
   # Generate images with Nano Banana Pro (4K text, reasoning)
   %(prog)s --task generate --prompt "Travel poster with text 'EXPLORE'" \\
-    --model gemini-3-pro-image-preview --aspect-ratio 3:4 --size 4K
+    --model gemma-4-31b-it-image-preview --aspect-ratio 3:4 --size 4K
 
   # Generate images with Imagen 4 (production quality)
   %(prog)s --task generate --prompt "Product photo of coffee mug" \\
@@ -1119,7 +1117,7 @@ Examples:
                        help='Aspect ratio for image/video generation')
     parser.add_argument('--num-images', type=int, default=1,
                        help='Number of images to generate (1-4, default: 1)')
-    # 4K available for Nano Banana Pro (gemini-3-pro-image-preview)
+    # 4K available for Nano Banana Pro (gemma-4-31b-it-image-preview)
     # Note: Not all models support --size, only use when needed
     parser.add_argument('--size', choices=['1K', '2K', '4K'], default=None,
                        help='Image size - 1K/2K for Imagen 4, 1K/2K/4K for Nano Banana (optional)')