@nockdev/awf 6.2.0 → 6.2.2

package/.agent/skills/ai-ml/gemini-live/data/live-api.yaml

@@ -0,0 +1,103 @@
+# =============================================================================
+# live-api.yaml — Gemini Live API Patterns
+# =============================================================================
+# 8 patterns for real-time WebSocket streaming with Gemini
+# Source: https://ai.google.dev/gemini-api/docs/live
+# =============================================================================
+
+version: "1.0.0"
+total_patterns: 8
+
+overview:
+  description: "Low-latency, real-time voice and video interactions with Gemini"
+  transport: "WebSocket"
+  input_formats:
+    audio: "PCM 16-bit, 16kHz, mono"
+    video: "Supported camera streams"
+    text: "UTF-8 text"
+  output_audio: "24kHz sample rate"
+  features:
+    - "Voice Activity Detection (VAD)"
+    - "Tool use and function calling"
+    - "Session management"
+    - "Ephemeral tokens"
+    - "Multi-tool support"
+
+approaches:
+  server_to_server:
+    desc: "Backend proxies client streams to Live API"
+    pros:
+      - "Full security control"
+      - "API key stays server-side"
+      - "Can add business logic"
+    cons:
+      - "Added latency from proxy"
+      - "Server resource requirements"
+    use_case: "Production apps with security requirements"
+
+  client_to_server:
+    desc: "Frontend connects directly to Live API"
+    pros:
+      - "Lowest latency"
+      - "Simpler architecture"
+    cons:
+      - "Requires ephemeral tokens"
+      - "Limited security control"
+    use_case: "Real-time demos, low-latency requirements"
+
+voice_activity_detection:
+  desc: "Automatic detection of when user starts/stops speaking"
+  behavior: "Model processes accumulated audio when silence detected"
+  customization: "Configurable silence threshold"
+
+session_management:
+  desc: "Maintain conversation state across multiple exchanges"
+  features:
+    - "Session persistence"
+    - "Context window management"
+    - "Automatic reconnection"
+  max_duration: "Configurable (model dependent)"
+
+ephemeral_tokens:
+  desc: "Short-lived tokens for client-side auth"
+  flow:
+    - "Server generates ephemeral token from API key"
+    - "Token sent to client"
+    - "Client uses token to connect directly"
+    - "Token expires after set duration"
+  security: "Never expose API key to client"
+
+partner_integrations:
+  - name: "Pipecat (Daily)"
+    desc: "Real-time AI chatbot framework"
+    url: "https://docs.pipecat.ai/guides/features/gemini-live"
+  - name: "LiveKit"
+    desc: "Agents with Gemini Live"
+    url: "https://docs.livekit.io/agents/models/realtime/plugins/gemini/"
+  - name: "ADK (Google)"
+    desc: "Agent Development Kit with Live API"
+    url: "https://google.github.io/adk-docs/streaming/"
+  - name: "Fishjam"
+    desc: "Live video/audio streaming"
+    url: "https://docs.fishjam.io/tutorials/gemini-live-integration"
+  - name: "Voximplant"
+    desc: "Phone call integration"
+    url: "https://voximplant.com/products/gemini-client"
+
+api_usage:
+  python: |
+    from google import genai
+    client = genai.Client(api_key="API_KEY")
+
+    async with client.aio.live.connect(model="gemini-2.0-flash-live-001") as session:
+        # Send audio
+        await session.send(audio_data, end_of_turn=True)
+        
+        # Receive response
+        async for response in session.receive():
+            if response.data:
+                play_audio(response.data)
+
+# =============================================================================
+# _DOMYH Awesome Code • Live API v1.0_
+# =============================================================================

package/.agent/skills/ai-ml/gemini-media-gen/META.yaml

@@ -0,0 +1,56 @@
+name: gemini-media-gen
+version: "1.0.0"
+display: Gemini Media Generation
+category: ai-ml
+tier: 1
+priority: 2
+
+desc: "Gemini native media generation: Nano Banana image gen (Flash + Pro 4K), Veo 3.1 video gen (8s 720p-4K + audio), Files API integration. Prompt optimization, multi-ref images, search grounding, batch generation. 36+ patterns."
+
+triggers:
+  file_patterns: []
+  keywords:
+    - nano banana
+    - image generation
+    - generate image
+    - veo
+    - video generation
+    - generate video
+    - gemini image
+    - gemini video
+    - text to image
+    - text to video
+    - image editing
+    - 4k image
+    - search grounding image
+    - files api
+    - media generation
+    - aspect ratio
+    - image prompt
+  intents:
+    - generate image
+    - create video
+    - edit image
+    - media generation
+    - batch image creation
+
+caps:
+  - Nano Banana Flash (speed, 1024px max)
+  - Nano Banana Pro (4K, Thinking, 14 ref images, Search grounding)
+  - Veo 3.1 (8s video, 720p-4K, native audio)
+  - Video extension & frame control
+  - Files API (20GB/project, 48h retention, free)
+  - Batch image generation
+  - Multi-reference image compositing
+  - Text rendering in images
+  - Image editing (inpainting, style transfer)
+
+related_skills:
+  - prompt-engineering
+  - domyh-design
+
+data_files:
+  - data/image-models.yaml
+  - data/image-prompts.yaml
+  - data/video-generation.yaml
+  - data/files-api.yaml

package/.agent/skills/ai-ml/gemini-media-gen/SKILL.md

@@ -0,0 +1,128 @@
+# Gemini Media Generation
+
+> 🎨 **Native image + video generation via Gemini API**
+> **Models**: Nano Banana (Flash/Pro) + Veo 3.1 | **Patterns**: 36+
+
+---
+
+## Quick Reference
+
+| What You Need                | Data File               | Patterns |
+| ---------------------------- | ----------------------- | -------- |
+| Model selection & configs    | `image-models.yaml`     | 8        |
+| Image prompt optimization    | `image-prompts.yaml`    | 12       |
+| Video generation & prompting | `video-generation.yaml` | 10       |
+| Files API management         | `files-api.yaml`        | 6        |
+
+---
+
+## Model Selection Guide
+
+| Feature              | Nano Banana Flash                           | Nano Banana Pro                           |
+| -------------------- | ------------------------------------------- | ----------------------------------------- |
+| **Speed**            | ⚡ Fast                                     | 🐢 Slower                                 |
+| **Max resolution**   | 1024×1024                                   | **4K (4096×4096)**                        |
+| **Reference images** | Up to 5                                     | **Up to 14**                              |
+| **Thinking**         | ❌                                          | ✅                                        |
+| **Search grounding** | ❌                                          | ✅                                        |
+| **Text rendering**   | Basic                                       | **Enhanced**                              |
+| **SynthID**          | ✅ (watermark)                              | ✅                                        |
+| **Model ID**         | `gemini-2.0-flash-preview-image-generation` | `gemini-2.5-pro-preview-image-generation` |
+
+**Decision**: Use Flash for speed/prototyping, Pro for quality/production.
+
+---
+
+## Image Generation Rules
+
+1. **Describe the scene** — Don't list keywords
+   - ✅ "A golden retriever playing fetch on a sandy beach at sunset, waves crashing in the background"
+   - ❌ "golden retriever, beach, sunset, waves, fetch, sand"
+
+2. **Aspect Ratios**: `1:1` (default), `16:9` (landscape), `9:16` (portrait), `4:3`, `3:4`
+
+3. **Safety**: All images include invisible SynthID watermark. Safety filters block harmful/CSAM content.
+
+4. **Text in images**: Wrap desired text in quotes — `'Sale 50% Off'`
+
+5. **Vietnamese support**: Gemini supports Vietnamese text in prompts natively
+
+---
+
+## Reference Image Strategies
+
+| Strategy                  | Max Refs          | Use Case                                        |
+| ------------------------- | ----------------- | ----------------------------------------------- |
+| **Style transfer**        | 1-3               | Apply reference art style to new content        |
+| **Subject reference**     | 1                 | Keep consistent character/product across images |
+| **Multi-ref composition** | Up to 14 (Pro)    | Combine elements from multiple references       |
+| **Inpainting**            | 1 (mask required) | Edit specific regions of existing image         |
+
+---
+
+## Veo 3.1 Video Generation
+
+| Parameter           | Option          | Description                           |
+| ------------------- | --------------- | ------------------------------------- |
+| **Duration**        | 5-8 seconds     | Fixed short clips                     |
+| **Resolution**      | 720p, 1080p, 4K | Output quality                        |
+| **Audio**           | Native          | AI-generated sound effects & dialogue |
+| **Input**           | Text or Image   | Text-to-video or image-to-video       |
+| **Negative prompt** | Yes             | "Don't include X in the video"        |
+
+### Video Prompt Elements
+
+```
+Subject: [who/what is in the scene]
+Action: [what is happening]
+Style: [cinematic, documentary, anime, etc.]
+Camera: [pan left, zoom in, tracking shot, etc.]
+Audio: [ambient sounds, dialogue, music style]
+Lighting: [golden hour, neon, soft diffused, etc.]
+```
+
+### Video Extension
+
+Feed last frame of generated video to extend the scene for another 5-8 seconds.
+
+### Frame Control
+
+Use reference image as first/last frame to control video composition.
+
+---
+
+## Files API Quick Reference
+
+| Feature             | Value                            |
+| ------------------- | -------------------------------- |
+| **Storage**         | 20GB per project                 |
+| **Per-file max**    | 2GB                              |
+| **Retention**       | 48 hours                         |
+| **Cost**            | Free                             |
+| **Supported types** | Images, video, audio, PDFs, text |
+
+```python
+# Upload
+file = client.files.upload(file="large_video.mp4")
+
+# Check status (for video processing)
+while file.state == "PROCESSING":
+    file = client.files.get(name=file.name)
+
+# Use in generation
+response = model.generate_content([file, "Describe this video"])
+
+# Delete
+client.files.delete(name=file.name)
+```
+
+---
+
+## HSA Integration
+
+| Domain | Query Examples                          |
+| ------ | --------------------------------------- |
+| Image  | "nano banana 4K pro reference images"   |
+| Video  | "veo 3.1 camera movement prompt"        |
+| Files  | "files API upload video processing"     |
+| Prompt | "image generation describe scene style" |

package/.agent/skills/ai-ml/gemini-media-gen/data/files-api.yaml

@@ -0,0 +1,96 @@
+# =============================================================================
+# files-api.yaml — Gemini Files API Patterns
+# =============================================================================
+# 6 patterns for file upload, management, and multimodal usage
+# Source: https://ai.google.dev/gemini-api/docs/files
+# =============================================================================
+
+version: "1.0.0"
+total_patterns: 6
+
+overview:
+  storage: "20GB per project"
+  max_file_size: "2GB"
+  retention: "48 hours (auto-deleted)"
+  cost: "Free"
+  download: "Cannot download uploaded files"
+
+supported_types:
+  images: ["image/jpeg", "image/png", "image/gif", "image/webp", "image/heic"]
+  video:
+    [
+      "video/mp4",
+      "video/mpeg",
+      "video/quicktime",
+      "video/webm",
+      "video/x-msvideo",
+    ]
+  audio:
+    [
+      "audio/aac",
+      "audio/flac",
+      "audio/mp3",
+      "audio/m4a",
+      "audio/ogg",
+      "audio/wav",
+    ]
+  documents: ["application/pdf", "text/plain", "text/csv", "text/html"]
+
+operations:
+  upload:
+    desc: "Upload a file for use in generation"
+    python: |
+      from google import genai
+      client = genai.Client(api_key="API_KEY")
+      file = client.files.upload(file="video.mp4")
+      print(f"URI: {file.uri}, State: {file.state}")
+    javascript: |
+      const file = await ai.files.upload({ file: "video.mp4" });
+      console.log(`URI: ${file.uri}, State: ${file.state}`);
+
+  check_status:
+    desc: "Wait for video/large file processing"
+    python: |
+      import time
+      while file.state.name == "PROCESSING":
+          time.sleep(5)
+          file = client.files.get(name=file.name)
+      if file.state.name == "FAILED":
+          raise ValueError("File processing failed")
+
+  use_in_generation:
+    desc: "Reference uploaded file in model request"
+    python: |
+      response = client.models.generate_content(
+          model="gemini-2.5-flash",
+          contents=[file, "Summarize this video in 3 bullet points"]
+      )
+
+  list_files:
+    desc: "List all uploaded files"
+    python: |
+      for f in client.files.list():
+          print(f"{f.name}: {f.size_bytes} bytes, {f.state}")
+
+  get_metadata:
+    desc: "Get file metadata by name"
+    python: |
+      file = client.files.get(name="files/abc123")
+      print(f"Name: {file.display_name}, Types: {file.mime_type}")
+
+  delete_file:
+    desc: "Delete a file before 48h auto-expiry"
+    python: |
+      client.files.delete(name=file.name)
+
+best_practices:
+  - "Upload large files (>20MB) via Files API instead of inline"
+  - "Video processing takes 30s-5min depending on size"
+  - "Check state before using: PROCESSING → ACTIVE → use in request"
+  - "Delete sensitive files immediately after use"
+  - "Use display_name for easier file management"
+  - "Files API works across all Gemini models"
+
+# =============================================================================
+# _DOMYH Awesome Code • Files API v1.0_
+# =============================================================================

package/.agent/skills/ai-ml/gemini-media-gen/data/image-models.yaml

@@ -0,0 +1,112 @@
+# =============================================================================
+# image-models.yaml — Nano Banana Image Generation Models
+# =============================================================================
+# 8 patterns for model selection, configuration, and API usage
+# Source: https://ai.google.dev/gemini-api/docs/image-generation
+# =============================================================================
+
+version: "1.0.0"
+total_patterns: 8
+
+models:
+  flash:
+    id: "gemini-2.0-flash-preview-image-generation"
+    display: "Nano Banana Flash"
+    tier: "speed"
+    max_resolution: "1024x1024"
+    max_reference_images: 5
+    thinking: false
+    search_grounding: false
+    text_rendering: "basic"
+    synthid: true
+    best_for:
+      - "Rapid prototyping"
+      - "Batch generation"
+      - "Low-latency applications"
+      - "Standard quality images"
+    config:
+      response_modalities: ["TEXT", "IMAGE"]
+      temperature: 1.0 # Recommended for image gen
+      top_p: null # Not supported for image gen
+      top_k: null # Not supported for image gen
+
+  pro:
+    id: "gemini-2.5-pro-preview-image-generation"
+    display: "Nano Banana Pro"
+    tier: "quality"
+    max_resolution: "4096x4096"
+    max_reference_images: 14
+    thinking: true
+    search_grounding: true
+    text_rendering: "enhanced"
+    synthid: true
+    best_for:
+      - "Production-quality images"
+      - "4K resolution output"
+      - "Complex multi-reference compositing"
+      - "Accurate text rendering"
+      - "Search-grounded image generation"
+    config:
+      response_modalities: ["TEXT", "IMAGE"]
+      temperature: 1.0
+      thinking_config:
+        thinking_budget: 4096 # Optional, for complex prompts
+
+aspect_ratios:
+  - ratio: "1:1"
+    desc: "Square (default)"
+    use_case: "Social media, profile pictures, icons"
+  - ratio: "16:9"
+    desc: "Wide landscape"
+    use_case: "Desktop wallpapers, YouTube thumbnails, presentations"
+  - ratio: "9:16"
+    desc: "Tall portrait"
+    use_case: "Mobile wallpapers, Instagram stories, TikTok covers"
+  - ratio: "4:3"
+    desc: "Standard landscape"
+    use_case: "Photo prints, blog images"
+  - ratio: "3:4"
+    desc: "Standard portrait"
+    use_case: "Product photos, book covers"
+
+safety_settings:
+  description: "All generated images include invisible SynthID watermark"
+  blocked_content:
+    - "CSAM (Child Sexual Abuse Material)"
+    - "Violent or graphic content"
+    - "Hate speech imagery"
+  person_generation:
+    default: "ALLOW_ADULT"
+    options:
+      - "DONT_ALLOW"
+      - "ALLOW_ADULT"
+
+api_usage:
+  python: |
+    from google import genai
+    client = genai.Client(api_key="YOUR_API_KEY")
+    response = client.models.generate_content(
+        model="gemini-2.0-flash-preview-image-generation",
+        contents="A serene Japanese garden with cherry blossoms",
+        config=genai.types.GenerateContentConfig(
+            response_modalities=["TEXT", "IMAGE"]
+        )
+    )
+    # Access image: response.candidates[0].content.parts[0].inline_data
+  javascript: |
+    const response = await ai.models.generateContent({
+        model: "gemini-2.0-flash-preview-image-generation",
+        contents: "A serene Japanese garden with cherry blossoms",
+        config: { responseModalities: ["TEXT", "IMAGE"] }
+    });
+
+batch_generation:
+  description: "Generate multiple images in sequence"
+  tips:
+    - "Use chat mode to maintain style consistency"
+    - "Reference previous images: 'Generate another image like the previous one but...'"
+    - "Flash model for batch, Pro for final selection"
+
+# =============================================================================
+# _DOMYH Awesome Code • Image Models v1.0_
+# =============================================================================

package/.agent/skills/ai-ml/gemini-media-gen/data/image-prompts.yaml

@@ -0,0 +1,131 @@
+# =============================================================================
+# image-prompts.yaml — Image Generation Prompt Patterns
+# =============================================================================
+# 12 patterns for crafting effective image generation prompts
+# =============================================================================
+
+version: "1.0.0"
+total_patterns: 12
+
+golden_rule:
+  principle: "Describe the SCENE, don't list KEYWORDS"
+  good: "A golden retriever playing fetch on a sandy beach at sunset, waves crashing in the background"
+  bad: "golden retriever, beach, sunset, waves, fetch, sand"
+
+prompt_structure:
+  template: "[Subject] [Action] [Setting] [Style] [Lighting] [Details]"
+  elements:
+    - name: "Subject"
+      desc: "Who or what is the main focus"
+      example: "A young woman in a red dress"
+    - name: "Action"
+      desc: "What is happening"
+      example: "walking through a field of sunflowers"
+    - name: "Setting"
+      desc: "Where the scene takes place"
+      example: "in a vast countryside under open sky"
+    - name: "Style"
+      desc: "Art direction and aesthetic"
+      examples:
+        - "photorealistic"
+        - "oil painting"
+        - "watercolor"
+        - "digital art"
+        - "anime"
+        - "pencil sketch"
+        - "3D render"
+        - "isometric"
+        - "flat design"
+        - "pixel art"
+    - name: "Lighting"
+      desc: "Light quality and direction"
+      examples:
+        - "golden hour sunlight"
+        - "soft diffused studio light"
+        - "dramatic chiaroscuro"
+        - "neon glow"
+        - "moonlit"
+        - "backlit silhouette"
+    - name: "Details"
+      desc: "Additional specifications"
+      examples:
+        - "shallow depth of field"
+        - "wide angle lens"
+        - "bird's eye view"
+        - "macro close-up"
+
+text_rendering:
+  description: "Render text inside generated images"
+  method: "Wrap desired text in quotes within prompt"
+  example: "A neon sign that says 'Open 24 Hours' hanging in a rainy alley"
+  tips:
+    - "Use Pro model for best text accuracy"
+    - "Keep text short (1-5 words best)"
+    - "Specify font style: 'bold', 'handwritten', 'neon', 'serif'"
+    - "Describe text placement: 'centered', 'at the bottom', 'on the sign'"
+
+reference_image_patterns:
+  style_transfer:
+    desc: "Apply art style from reference to new content"
+    max_refs: 3
+    prompt: "Generate an image in the same style as the reference: [content description]"
+
+  subject_consistency:
+    desc: "Keep same character/product across images"
+    max_refs: 1
+    prompt: "Generate this character [reference] in a new scene: [scene description]"
+    tip: "Describe the subject's key features in text too for best consistency"
+
+  multi_reference:
+    desc: "Combine elements from multiple references"
+    max_refs: 14 # Pro only
+    prompt: "Combine the style of [ref1], subject from [ref2], background from [ref3]"
+
+  inpainting:
+    desc: "Edit specific region of existing image"
+    requires: "Mask (white = edit, black = keep)"
+    prompt: "Replace [masked area] with [new content]"
+
+style_keywords:
+  photographic:
+    - "35mm film photography"
+    - "DSLR, f/1.4 bokeh"
+    - "polaroid instant photo"
+    - "long exposure light trails"
+    - "infrared photography"
+  artistic:
+    - "impressionist oil painting"
+    - "Japanese woodblock print (ukiyo-e)"
+    - "Art Nouveau poster"
+    - "pop art halftone"
+    - "baroque still life"
+  digital:
+    - "Unreal Engine 5 render"
+    - "octane render, 8K"
+    - "low poly 3D model"
+    - "vaporwave aesthetic"
+    - "cyberpunk neon city"
+
+quality_enhancers:
+  - "highly detailed"
+  - "sharp focus"
+  - "professional lighting"
+  - "award-winning photography"
+  - "4K resolution" # Use with Pro model
+
+negative_prompts:
+  description: "What to AVOID in the image (Nano Banana supports)"
+  examples:
+    - "blurry, out of focus"
+    - "watermark, signature"
+    - "deformed, distorted"
+    - "low quality, pixelated"
+
+vietnamese_support:
+  description: "Gemini supports Vietnamese prompts natively"
+  example: "Một cô gái Việt Nam mặc áo dài trắng đang đi dưới hàng cây phượng vĩ đỏ"
+  tip: "Vietnamese prompts produce culturally accurate imagery"
+
+# =============================================================================
+# _DOMYH Awesome Code • Image Prompt Patterns v1.0_
+# =============================================================================