Initial commit

2025-11-29 18:23:22 +08:00
commit befa40959b
10 changed files with 1318 additions and 0 deletions
--- a/commands/completion.md
+++ b/commands/completion.md
@@ -0,0 +1,137 @@
+# Completion Command
+
+Generate shell completion scripts for bash, zsh, or fish shells.
+
+## Usage
+
+```bash
+gemini-nano-banana-tool completion SHELL
+```
+
+## Arguments
+
+- `SHELL` - Shell type (bash, zsh, or fish)
+
+## Examples
+
+### Bash Completion
+
+```bash
+# Generate completion script
+gemini-nano-banana-tool completion bash
+
+# Add to ~/.bashrc
+eval "$(gemini-nano-banana-tool completion bash)"
+```
+
+### Zsh Completion
+
+```bash
+# Generate completion script
+gemini-nano-banana-tool completion zsh
+
+# Add to ~/.zshrc
+eval "$(gemini-nano-banana-tool completion zsh)"
+```
+
+### Fish Completion
+
+```bash
+# Generate completion script
+gemini-nano-banana-tool completion fish
+
+# Save to completions directory
+gemini-nano-banana-tool completion fish > \
+  ~/.config/fish/completions/gemini-nano-banana-tool.fish
+```
+
+## What Shell Completion Does
+
+Shell completion provides:
+
+- **Command Completion**: Tab to complete command names
+- **Argument Completion**: Tab to complete option names
+- **Value Completion**: Tab to complete option values (file paths, choices)
+- **Help Integration**: Shows available options while typing
+
+## Installation
+
+### Bash
+
+Add to `~/.bashrc`:
+
+```bash
+eval "$(gemini-nano-banana-tool completion bash)"
+```
+
+Then reload:
+```bash
+source ~/.bashrc
+```
+
+### Zsh
+
+Add to `~/.zshrc`:
+
+```bash
+eval "$(gemini-nano-banana-tool completion zsh)"
+```
+
+Then reload:
+```bash
+source ~/.zshrc
+```
+
+### Fish
+
+Save to completions directory:
+
+```bash
+gemini-nano-banana-tool completion fish > \
+  ~/.config/fish/completions/gemini-nano-banana-tool.fish
+```
+
+Then reload (automatic in most cases):
+```bash
+exec fish
+```
+
+## Verification
+
+Test completion by typing:
+
+```bash
+gemini-nano-banana-tool <TAB>
+```
+
+You should see available commands:
+- `completion`
+- `generate`
+- `generate-conversation`
+- `list-aspect-ratios`
+- `list-models`
+- `promptgen`
+
+## Troubleshooting
+
+If completion doesn't work:
+
+1. **Check Installation**: Verify completion script is loaded
+2. **Reload Shell**: Close and reopen terminal or source config file
+3. **Check Permissions**: Ensure completion file is readable
+4. **Check Path**: Verify `gemini-nano-banana-tool` is in PATH
+
+## Pattern
+
+This follows the industry-standard pattern used by:
+- `kubectl completion`
+- `helm completion`
+- `docker completion`
+- `gh completion`
+
+## Benefits
+
+- **Faster Typing**: Tab completion reduces typing
+- **Discover Options**: See available options without `--help`
+- **Reduce Errors**: Autocomplete prevents typos
+- **Better UX**: Professional CLI experience
--- a/commands/generate-conversation.md
+++ b/commands/generate-conversation.md
@@ -0,0 +1,182 @@
+# Generate Conversation Command
+
+Multi-turn image generation with progressive refinement. Each turn builds on previous context, allowing iterative improvements without starting over.
+
+## Usage
+
+```bash
+gemini-nano-banana-tool generate-conversation PROMPT -o OUTPUT [OPTIONS]
+```
+
+## Arguments
+
+- `PROMPT` - Text prompt for this turn (required)
+
+## Required Options
+
+- `-o, --output PATH` - Output image file path (required)
+
+## Conversation Options
+
+- `-f, --file PATH` - Conversation file (creates new if doesn't exist)
+- `-a, --aspect-ratio TEXT` - Aspect ratio (default: 1:1, only for new conversations)
+- `-m, --model TEXT` - Gemini model (default: gemini-2.5-flash-image, only for new)
+
+## Authentication Options
+
+- `--api-key TEXT` - Override API key from environment
+- `--use-vertex` - Use Vertex AI instead of Developer API
+- `--project TEXT` - Google Cloud project (for Vertex AI)
+- `--location TEXT` - Google Cloud location (for Vertex AI)
+
+## Other Options
+
+- `-v, --verbose` - Multi-level verbosity (-v INFO, -vv DEBUG, -vvv TRACE)
+
+## How It Works
+
+1. **First Turn**: Create initial image from prompt, save conversation state
+2. **Subsequent Turns**: Previous output automatically becomes reference image
+3. **Persistence**: All turns, prompts, and metadata saved to JSON file
+4. **Resume**: Load conversation file to continue refinement
+
+## Examples
+
+### Basic Multi-Turn Workflow
+
+```bash
+# Turn 1: Initial generation
+gemini-nano-banana-tool generate-conversation "A sunset over mountains" \
+  -o sunset1.png --file conversation.json
+
+# Turn 2: Refinement (loads conversation automatically)
+gemini-nano-banana-tool generate-conversation "Make the sky more orange" \
+  -o sunset2.png --file conversation.json
+
+# Turn 3: Further refinement
+gemini-nano-banana-tool generate-conversation "Add a lake in the foreground" \
+  -o sunset3.png --file conversation.json
+```
+
+### Interior Design Example
+
+```bash
+# Turn 1: Initial room
+gemini-nano-banana-tool generate-conversation \
+  "A modern minimalist living room with large windows" \
+  -o room-v1.png --file interior.json -a 16:9
+
+# Turn 2: Add furniture
+gemini-nano-banana-tool generate-conversation \
+  "Add a gray sofa and wooden coffee table" \
+  -o room-v2.png --file interior.json
+
+# Turn 3: Adjust lighting
+gemini-nano-banana-tool generate-conversation \
+  "Make the lighting warmer and add floor lamp" \
+  -o room-v3.png --file interior.json
+
+# Turn 4: Final touches
+gemini-nano-banana-tool generate-conversation \
+  "Add plants and artwork on the walls" \
+  -o room-final.png --file interior.json
+```
+
+### Product Photography Example
+
+```bash
+# Turn 1: Initial product shot
+gemini-nano-banana-tool generate-conversation \
+  "Professional product photo of wireless headphones" \
+  -o headphones-v1.png --file product.json -a 1:1
+
+# Turn 2: Adjust angle
+gemini-nano-banana-tool generate-conversation \
+  "Rotate to show the left side" \
+  -o headphones-v2.png --file product.json
+
+# Turn 3: Change background
+gemini-nano-banana-tool generate-conversation \
+  "Change background to dark gradient" \
+  -o headphones-v3.png --file product.json
+```
+
+## Conversation File Format
+
+The conversation file stores complete history in JSON:
+
+```json
+{
+  "conversation_id": "20251120_181305",
+  "model": "gemini-2.5-flash-image",
+  "aspect_ratio": "16:9",
+  "turns": [
+    {
+      "prompt": "A sunset over mountains",
+      "output_path": "/path/to/sunset1.png",
+      "reference_images": [],
+      "metadata": {
+        "token_count": 1295,
+        "resolution": "1344x768",
+        "finish_reason": "STOP"
+      },
+      "timestamp": "2025-11-20T18:13:11.428020"
+    },
+    {
+      "prompt": "Make the sky more orange",
+      "output_path": "/path/to/sunset2.png",
+      "reference_images": ["/path/to/sunset1.png"],
+      "metadata": {
+        "token_count": 1554,
+        "resolution": "1344x768",
+        "finish_reason": "STOP"
+      },
+      "timestamp": "2025-11-20T18:13:27.318416"
+    }
+  ],
+  "created_at": "2025-11-20T18:13:05.751502",
+  "updated_at": "2025-11-20T18:13:27.318430"
+}
+```
+
+## Use Cases
+
+- **Product Photography**: Iteratively adjust lighting, angles, styling
+- **Character Design**: Refine poses, clothing, expressions progressively
+- **Interior Design**: Build rooms by adding furniture and decor step by step
+- **Marketing Materials**: Test variations while maintaining consistency
+- **Concept Art**: Explore different iterations of a design
+- **Fashion E-commerce**: Try products on models or in different settings
+
+## Important Notes
+
+- **Model & Aspect Ratio**: Only set when creating new conversation (locked for subsequent turns)
+- **Reference Images**: Previous output automatically used (no manual `-i` needed)
+- **Resume Anytime**: Load conversation file to continue from any turn
+- **No File Flag**: Can use without `--file` but conversation won't be saved
+
+## Why Use Conversation Mode?
+
+- **Progressive Refinement**: Iteratively improve without losing context
+- **Experiment Safely**: Try variations while maintaining consistency
+- **Context Awareness**: Each turn references previous outputs automatically
+- **Evolution Tracking**: Complete history of prompts and changes
+- **Resume Anytime**: Continue conversations across sessions
+
+## Cost Information
+
+Each turn costs the same as a single generation:
+- **Flash Model**: ~$0.039 per turn
+- **Pro Model 1K/2K**: ~$0.134 per turn
+- **Pro Model 4K**: ~$0.24 per turn
+
+Cost is tracked per turn in conversation file metadata.
+
+## Authentication
+
+Set environment variable:
+```bash
+export GEMINI_API_KEY='your-api-key'
+```
+
+Get API key: https://aistudio.google.com/app/apikey
--- a/commands/generate.md
+++ b/commands/generate.md
@@ -0,0 +1,194 @@
+# Generate Command
+
+Generate images from text prompts with optional reference images using Google Gemini models.
+
+**Alias:** `generate-image` (can be used interchangeably with `generate`)
+
+## Usage
+
+```bash
+gemini-nano-banana-tool generate [PROMPT] -o OUTPUT [OPTIONS]
+gemini-nano-banana-tool generate-image [PROMPT] -o OUTPUT [OPTIONS]  # Alias
+```
+
+## Arguments
+
+- `PROMPT` - Text prompt (positional, mutually exclusive with `-f` and `-s`)
+
+## Required Options
+
+- `-o, --output PATH` - Output image file path (required)
+
+## Prompt Input Options (mutually exclusive)
+
+- `PROMPT` - Positional argument (default)
+- `-f, --prompt-file PATH` - Read prompt from file
+- `-s, --stdin` - Read prompt from stdin (for piping)
+
+## Image Options
+
+- `-i, --image PATH` - Reference image (can be used up to 3 times for Flash, 14 for Pro)
+- `-a, --aspect-ratio TEXT` - Aspect ratio (default: 1:1)
+- `-m, --model TEXT` - Gemini model (default: gemini-2.5-flash-image)
+- `-r, --resolution TEXT` - Resolution quality for Pro model (1K/2K/4K)
+
+## Authentication Options
+
+- `--api-key TEXT` - Override API key from environment
+- `--use-vertex` - Use Vertex AI instead of Developer API
+- `--project TEXT` - Google Cloud project (for Vertex AI)
+- `--location TEXT` - Google Cloud location (for Vertex AI)
+
+## Other Options
+
+- `-v, --verbose` - Multi-level verbosity (-v INFO, -vv DEBUG, -vvv TRACE)
+
+## Examples
+
+### Basic Text-to-Image
+
+```bash
+# Simple generation (both commands work)
+gemini-nano-banana-tool generate "A cat wearing a wizard hat" -o cat.png
+gemini-nano-banana-tool generate-image "A cat wearing a wizard hat" -o cat.png
+
+# With aspect ratio
+gemini-nano-banana-tool generate "Mountain landscape" -o landscape.png -a 16:9
+```
+
+### Prompt from File or Stdin
+
+```bash
+# From file
+gemini-nano-banana-tool generate -f prompt.txt -o output.png
+
+# From stdin (piping)
+echo "Beautiful sunset" | gemini-nano-banana-tool generate -o sunset.png -s
+
+# With promptgen
+gemini-nano-banana-tool promptgen "wizard cat" | \
+  gemini-nano-banana-tool generate -o cat.png -s -a 16:9
+```
+
+### Image Editing with Reference Images
+
+```bash
+# Single reference image
+gemini-nano-banana-tool generate "Add a birthday hat" -o edited.png \
+  -i original.jpg
+
+# Multiple reference images
+gemini-nano-banana-tool generate "Put the dress on the model in garden" \
+  -o fashion.png -i dress.jpg -i model.jpg
+
+# Up to 3 references (Flash) or 14 (Pro)
+gemini-nano-banana-tool generate "Combine these styles" -o result.png \
+  -i ref1.jpg -i ref2.jpg -i ref3.jpg
+```
+
+### Different Aspect Ratios
+
+```bash
+# Square (Instagram post)
+gemini-nano-banana-tool generate "Modern design" -o square.png -a 1:1
+
+# Widescreen (YouTube thumbnail)
+gemini-nano-banana-tool generate "Epic scene" -o wide.png -a 16:9
+
+# Vertical (Instagram story)
+gemini-nano-banana-tool generate "Portrait" -o vertical.png -a 9:16
+
+# Cinematic (ultra-wide)
+gemini-nano-banana-tool generate "Sci-fi panorama" -o cinema.png -a 21:9
+```
+
+### Model Selection and Resolution
+
+```bash
+# Default Flash model (fast, cost-effective)
+gemini-nano-banana-tool generate "Your prompt" -o output.png
+
+# Pro model (higher quality)
+gemini-nano-banana-tool generate "Your prompt" -o output.png \
+  -m gemini-3-pro-image-preview
+
+# Pro with 4K resolution (maximum quality)
+gemini-nano-banana-tool generate "Your prompt" -o output.png \
+  -m gemini-3-pro-image-preview -r 4K
+```
+
+### Verbosity Levels
+
+```bash
+# Normal (warnings only)
+gemini-nano-banana-tool generate "test" -o output.png
+
+# Verbose (show operations)
+gemini-nano-banana-tool generate "test" -o output.png -v
+
+# Debug (detailed info)
+gemini-nano-banana-tool generate "test" -o output.png -vv
+
+# Trace (full HTTP logs)
+gemini-nano-banana-tool generate "test" -o output.png -vvv
+```
+
+## Output Format
+
+Returns JSON with generation details:
+
+```json
+{
+  "output_path": "output.png",
+  "model": "gemini-2.5-flash-image",
+  "aspect_ratio": "16:9",
+  "resolution": "1344x768",
+  "resolution_quality": "1K",
+  "reference_image_count": 0,
+  "token_count": 1295,
+  "estimated_cost_usd": 0.0389,
+  "metadata": {
+    "finish_reason": "STOP",
+    "safety_ratings": null
+  }
+}
+```
+
+## Cost Information
+
+Automatic cost tracking based on token usage:
+
+- **Flash Model**: ~$0.039 per image (1,290 tokens × $0.00003)
+- **Pro Model 1K/2K**: ~$0.134 per image (1,120 tokens × $0.00012)
+- **Pro Model 4K**: ~$0.24 per image (2,000 tokens × $0.00012)
+
+Cost is always included in output JSON as `estimated_cost_usd`.
+
+## Supported Aspect Ratios
+
+- `1:1` - Square (1024×1024)
+- `16:9` - Widescreen (1344×768)
+- `9:16` - Vertical (768×1344)
+- `4:3` - Traditional (1184×864)
+- `3:4` - Portrait (864×1184)
+- `3:2` - DSLR (1248×832)
+- `2:3` - Portrait photo (832×1248)
+- `21:9` - Cinematic (1536×672)
+- `4:5` - Instagram portrait (896×1152)
+- `5:4` - Medium format (1152×896)
+
+## Authentication
+
+Set environment variable:
+```bash
+export GEMINI_API_KEY='your-api-key'
+```
+
+Get API key: https://aistudio.google.com/app/apikey
+
+For Vertex AI:
+```bash
+export GOOGLE_GENAI_USE_VERTEXAI=true
+export GOOGLE_CLOUD_PROJECT='your-project-id'
+export GOOGLE_CLOUD_LOCATION='us-central1'
+```
--- a/commands/list-aspect-ratios.md
+++ b/commands/list-aspect-ratios.md
@@ -0,0 +1,94 @@
+# List Aspect Ratios Command
+
+List supported aspect ratios with resolutions and use cases.
+
+## Usage
+
+```bash
+gemini-nano-banana-tool list-aspect-ratios
+```
+
+## Output
+
+```
+Available Aspect Ratios:
+  1:1    (1024x1024)  - Square (Instagram post, social media)
+  16:9   (1344x768)   - Widescreen (YouTube thumbnail, desktop)
+  9:16   (768x1344)   - Vertical (Instagram story, TikTok, mobile)
+  4:3    (1184x864)   - Traditional (classic photography)
+  3:4    (864x1184)   - Portrait orientation
+  3:2    (1248x832)   - DSLR photography
+  2:3    (832x1248)   - Portrait photography
+  21:9   (1536x672)   - Cinematic (ultra-wide)
+  4:5    (896x1152)   - Instagram portrait
+  5:4    (1152x896)   - Medium format photography
+```
+
+## Usage with Generate Command
+
+```bash
+# Square (Instagram post)
+gemini-nano-banana-tool generate "Modern design" -o square.png -a 1:1
+
+# Widescreen (YouTube thumbnail)
+gemini-nano-banana-tool generate "Epic scene" -o wide.png -a 16:9
+
+# Vertical (Instagram story)
+gemini-nano-banana-tool generate "Portrait" -o vertical.png -a 9:16
+
+# Cinematic (ultra-wide)
+gemini-nano-banana-tool generate "Sci-fi panorama" -o cinema.png -a 21:9
+```
+
+## Common Platform Aspect Ratios
+
+### Social Media
+
+- **Instagram Post**: 1:1 (square)
+- **Instagram Story**: 9:16 (vertical)
+- **Instagram Portrait**: 4:5
+- **Twitter Post**: 16:9 or 1:1
+- **Facebook Post**: 1:1 or 16:9
+- **LinkedIn Post**: 1:1 or 16:9
+
+### Video Platforms
+
+- **YouTube Thumbnail**: 16:9
+- **YouTube Banner**: 16:9 (wide)
+- **TikTok**: 9:16 (vertical)
+- **YouTube Shorts**: 9:16 (vertical)
+
+### Photography
+
+- **DSLR Standard**: 3:2
+- **Medium Format**: 5:4
+- **Classic Film**: 4:3
+- **Portrait**: 2:3 or 3:4
+
+### Displays
+
+- **Desktop/Laptop**: 16:9
+- **Ultrawide Monitor**: 21:9
+- **Mobile Portrait**: 9:16
+- **Tablet**: 4:3
+
+## Choosing an Aspect Ratio
+
+**For social media content:**
+- Use 1:1 for maximum compatibility
+- Use 9:16 for stories and vertical video
+- Use 4:5 for Instagram portrait posts
+
+**For professional photography:**
+- Use 3:2 for DSLR standard
+- Use 5:4 for medium format look
+- Use 2:3 for traditional portrait
+
+**For video/cinema:**
+- Use 16:9 for standard widescreen
+- Use 21:9 for cinematic ultra-wide
+- Use 9:16 for vertical mobile video
+
+**For presentations/displays:**
+- Use 16:9 for modern displays
+- Use 4:3 for traditional presentations
--- a/commands/list-models.md
+++ b/commands/list-models.md
@@ -0,0 +1,78 @@
+# List Models Command
+
+List available Gemini image generation models with descriptions.
+
+## Usage
+
+```bash
+gemini-nano-banana-tool list-models
+```
+
+## Output
+
+```
+Available Gemini Image Generation Models:
+  • gemini-2.5-flash-image (default) - Fast, high-quality image generation
+  • gemini-3-pro-image-preview - Advanced model with higher quality and more features
+```
+
+## Model Details
+
+### Flash Model (gemini-2.5-flash-image)
+
+**Default Model**
+
+- **Speed**: Fast (seconds per image)
+- **Quality**: High quality
+- **Resolution**: Fixed ~1024p
+- **Cost**: $0.00003 per token (~$0.039 per image)
+- **Reference Images**: Up to 3
+- **Best For**: Quick iterations, cost-effective production, high-volume generation
+
+### Pro Model (gemini-3-pro-image-preview)
+
+**Advanced Model**
+
+- **Speed**: Slower (better quality)
+- **Quality**: Maximum quality
+- **Resolution**: Variable (1K/2K/4K)
+- **Cost**: $0.00012 per token (~$0.134-0.24 per image)
+- **Reference Images**: Up to 14
+- **Best For**: Professional assets, high-quality requirements, complex scenes
+
+## Usage with Generate Command
+
+```bash
+# Default Flash model
+gemini-nano-banana-tool generate "Your prompt" -o output.png
+
+# Pro model
+gemini-nano-banana-tool generate "Your prompt" -o output.png \
+  -m gemini-3-pro-image-preview
+
+# Pro model with 4K resolution
+gemini-nano-banana-tool generate "Your prompt" -o output.png \
+  -m gemini-3-pro-image-preview -r 4K
+```
+
+## Cost Comparison
+
+| Model | Typical Cost | Speed | Quality | Resolution |
+|-------|-------------|-------|---------|------------|
+| Flash | $0.039 | Fast | High | Fixed ~1024p |
+| Pro 1K/2K | $0.134 | Medium | Higher | 1K-2K |
+| Pro 4K | $0.24 | Slower | Maximum | Up to 4K |
+
+## Choosing a Model
+
+**Use Flash when:**
+- Prototyping and testing
+- High-volume generation
+- Cost is a concern
+- Speed matters
+
+**Use Pro when:**
+- Final production images
+- Complex scenes with fine details
+- Higher resolution needed
+- Professional/commercial work
--- a/commands/promptgen.md
+++ b/commands/promptgen.md
@@ -0,0 +1,111 @@
+# Promptgen Command
+
+Transform simple descriptions into detailed, optimized image generation prompts using Gemini 2.0 Flash AI.
+
+## Usage
+
+```bash
+gemini-nano-banana-tool promptgen DESCRIPTION [OPTIONS]
+```
+
+## Arguments
+
+- `DESCRIPTION` - Simple description to enhance (e.g., "wizard cat", "cyberpunk city")
+
+## Options
+
+- `-t, --template TEXT` - Apply specialized template (photography, character, scene, food, abstract, logo)
+- `-s, --style TEXT` - Style hint (photorealistic, artistic, minimalist, etc.)
+- `-o, --output PATH` - Save prompt to file instead of stdout
+- `--json` - Output in JSON format
+- `-v, --verbose` - Show analysis and reasoning
+- `--list-templates` - Show all available templates
+
+## Examples
+
+### Basic Usage
+
+```bash
+# Generate detailed prompt from simple description
+gemini-nano-banana-tool promptgen "wizard cat"
+
+# With template for better results
+gemini-nano-banana-tool promptgen "wizard cat" --template character
+
+# Save to file for reuse
+gemini-nano-banana-tool promptgen "sunset landscape" -o prompt.txt
+```
+
+### Pipeline: Generate Prompt → Create Image
+
+```bash
+# Single pipeline: description → optimized prompt → image
+gemini-nano-banana-tool promptgen "wizard cat in magical library" | \
+  gemini-nano-banana-tool generate --stdin -o wizard-cat.png -a 16:9
+```
+
+### With Templates
+
+```bash
+# Photography template - technical camera details
+gemini-nano-banana-tool promptgen "mountain landscape" --template photography
+
+# Food template - plating and lighting
+gemini-nano-banana-tool promptgen "pasta dish" --template food
+
+# Scene template - foreground/midground/background
+gemini-nano-banana-tool promptgen "cyberpunk city" --template scene
+```
+
+### JSON Output (for automation)
+
+```bash
+# Get structured output
+gemini-nano-banana-tool promptgen "sunset" --json
+
+# Output format:
+# {
+#   "original": "sunset",
+#   "prompt": "A breathtaking golden hour sunset...",
+#   "template": null,
+#   "style": null,
+#   "tokens_used": 156
+# }
+```
+
+## Available Templates
+
+- `photography` - Professional photography with technical details (aperture, focal length, lighting)
+- `character` - Character design with pose, attire, expression
+- `scene` - Scene composition with foreground/midground/background layers
+- `food` - Food photography with plating, garnish, lighting
+- `abstract` - Abstract art with shapes, colors, patterns
+- `logo` - Logo design with typography, symbolism, brand identity
+
+## Why Use Promptgen?
+
+Creating effective image generation prompts requires knowledge of:
+- Photography and composition terminology
+- Lighting and technical details
+- Artistic styles and techniques
+- Color theory and palettes
+
+Promptgen automates this expertise using AI, transforming simple ideas into detailed, effective prompts.
+
+## Output Modes
+
+- **Plain text** (default) - Ready for piping to generate command
+- **JSON** (`--json`) - Structured output for scripts and automation
+- **Verbose** (`-v`) - Shows AI analysis and reasoning process
+
+## Cost
+
+Promptgen uses Gemini 2.0 Flash for prompt generation:
+- Cost: ~$0.001-0.003 per prompt optimization
+- Much cheaper than trial-and-error generation
+
+## Authentication
+
+Requires `GEMINI_API_KEY` or `GOOGLE_API_KEY` environment variable.
+
+Get your API key: https://aistudio.google.com/app/apikey