Skip to content

Commit c894d31

Browse files
committed
Change default to SDXL-Turbo and robot animal examples
Default model: SD-Turbo → SDXL-Turbo - Better quality (⭐⭐⭐ vs ⭐⭐) - Only 4s slower (17s vs 13s) - More impressive demos - Better showcases LLM enhancement Demo theme: Robot animals - "robot puppy", "robot kitten", "robot owl" - SD handles these well (no logo/text issues) - Adorable and engaging for demos - Shows enhancement clearly (simple → cute + detailed) Files updated: - src/gaia/agents/sd/mixin.py: default="SDXL-Turbo" - src/gaia/agents/sd/agent.py: default="SDXL-Turbo" - src/gaia/llm/lemonade_client.py: default="SDXL-Turbo" - src/gaia/cli.py: default="SDXL-Turbo", updated help text - tests/unit/test_sd_mixin.py: Expect SDXL-Turbo - docs/guides/sd.mdx: Robot animal examples, updated mermaid - docs/playbooks/sd-agent/index.mdx: Remove AMD logo reference Perfect demo: gaia sd "robot puppy" (17s, adorable result)
1 parent 366673e commit c894d31

7 files changed

Lines changed: 41 additions & 41 deletions

File tree

docs/guides/sd.mdx

Lines changed: 31 additions & 31 deletions
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@ description: "Transform simple prompts into professional images with LLM-powered
77
**CLI Command:** `gaia sd` - Your AI assistant for professional-quality images on Ryzen AI
88
</Info>
99

10-
# From "a robot" to Masterpiece in 15 Seconds
10+
# From "a robot puppy" to Masterpiece in 17 Seconds
1111

1212
The `gaia sd` command is **not just another Stable Diffusion wrapper**. It's an AI-powered prompt enhancement engine that transforms casual descriptions into professional-quality images.
1313

@@ -22,11 +22,11 @@ The `gaia sd` command is **not just another Stable Diffusion wrapper**. It's an
2222
```mermaid
2323
%%{init: {'theme':'base', 'themeVariables': { 'primaryColor':'#ED1C24', 'primaryTextColor':'#fff', 'primaryBorderColor':'#C8171E', 'lineColor':'#F4484D', 'secondaryColor':'#2d2d2d', 'tertiaryColor':'#f5f5f5', 'fontSize':'16px', 'fontFamily': 'system-ui, -apple-system, sans-serif'}}}%%
2424
graph LR
25-
A["You Type:<br/><b>'a robot'</b>"] --> B["🧠 LLM Brain<br/><small>Analyzes Intent</small>"]
25+
A["You Type:<br/><b>'robot puppy'</b>"] --> B["🧠 LLM Brain<br/><small>Analyzes Intent</small>"]
2626
2727
B --> C["✨ Enhancer<br/><small>Adds Keywords</small>"]
2828
29-
C --> D["Enhanced Prompt:<br/><i>'futuristic robot,<br/>studio lighting,<br/>metallic finish,<br/>highly detailed, 8K'</i>"]
29+
C --> D["Enhanced Prompt:<br/><i>'cute robotic puppy,<br/>glowing LED eyes,<br/>soft lighting,<br/>detailed, 4K'</i>"]
3030
3131
D --> E["🖼️ Stable Diffusion<br/><small>on Ryzen AI</small>"]
3232
@@ -43,11 +43,11 @@ graph LR
4343
```
4444

4545
**The 3-Second Story:**
46-
1. 🧠 LLM analyzes what you want
47-
2. ✨ Adds pro keywords (lighting, style, quality, composition)
48-
3. 🖼️ SD generates with the enhanced prompt
46+
1. 🧠 LLM analyzes "robot puppy" → understands cute + robotic theme
47+
2. ✨ Adds keywords: glowing LED eyes, metallic fur texture, soft lighting, detailed
48+
3. 🖼️ SD generates professional-quality robot puppy
4949

50-
**Result:** 2-word input20+ keyword professional prompt → High-quality image
50+
**Result:** Simple ideaEnhanced prompt → Adorable high-quality image
5151

5252
---
5353

@@ -76,7 +76,7 @@ photorealistic, shallow depth of field, 8K"
7676
```
7777
</CodeGroup>
7878

79-
**You type 2 words, the LLM adds 15+ keywords!** No SD expertise needed.
79+
**You type "robot puppy", the LLM adds lighting, materials, details!** No SD expertise needed.
8080

8181
---
8282

@@ -92,10 +92,10 @@ photorealistic, shallow depth of field, 8K"
9292

9393
<Step title="Generate!">
9494
```bash
95-
gaia sd "sunset over mountains"
95+
gaia sd "robot puppy"
9696
```
9797

98-
Watch the LLM enhance your prompt and SD generate the image.
98+
Watch the LLM enhance your prompt and SD generate an adorable robot animal!
9999
</Step>
100100
</Steps>
101101

@@ -105,9 +105,9 @@ photorealistic, shallow depth of field, 8K"
105105

106106
| You Type | LLM Enhances To | Result |
107107
|----------|----------------|--------|
108-
| **"robot"** | "futuristic robot, metallic chrome, studio lighting, sci-fi, detailed, 8K" | Professional render |
108+
| **"robot puppy"** | "cute robotic puppy, glowing LED eyes, metallic fur, soft lighting, playful, detailed, 4K" | Adorable robot animal |
109+
| **"robot owl"** | "mechanical owl, glowing amber eyes, bronze feathers, steampunk, detailed gears, 8K" | Majestic robot bird |
109110
| **"sunset"** | "vibrant sunset, golden hour, dramatic clouds, wide angle, volumetric lighting, 4K" | Cinematic scene |
110-
| **"cat"** | "fluffy orange cat, soft lighting, detailed fur, photorealistic, bokeh, 8K" | Photo-quality |
111111

112112
**Pattern:** Simple → Professional in every case.
113113

@@ -119,20 +119,20 @@ The LLM knows how each model works best:
119119

120120
| Model | Speed | What LLM Does |
121121
|-------|-------|---------------|
122-
| **SD-Turbo** | ~13s | Focuses on key elements, concise enhancement |
123-
| **SDXL-Turbo** | ~17s | Adds artistic style keywords, detailed lighting |
124-
| **SD-1.5** | ~88s | Balanced approach, traditional keywords |
122+
| **SDXL-Turbo** (default) | ~17s | Adds artistic style keywords, detailed lighting, best balance |
123+
| **SD-Turbo** | ~13s | Focuses on key elements only, concise enhancement |
124+
| **SD-1.5** | ~88s | Balanced traditional approach |
125125
| **SDXL-Base-1.0** | ~9min | Natural language, photorealistic focus, keyword weights |
126126

127127
```bash
128-
# Fast iteration
129-
gaia sd "test concept" --sd-model SD-Turbo
128+
# Default (good quality, fast)
129+
gaia sd "robot puppy"
130130

131-
# Better quality
132-
gaia sd "final design" --sd-model SDXL-Turbo
131+
# Even faster (lower quality)
132+
gaia sd "quick test" --sd-model SD-Turbo
133133

134-
# Photorealistic
135-
gaia sd "product photo" --sd-model SDXL-Base-1.0
134+
# Photorealistic (slow)
135+
gaia sd "robot kitten, photorealistic" --sd-model SDXL-Base-1.0
136136
```
137137

138138
**The LLM adjusts its strategy based on which model you pick!**
@@ -148,17 +148,17 @@ gaia sd -i
148148
Have a conversation about your image:
149149

150150
```
151-
You: a dragon
152-
Agent: [enhances] → Generates epic dragon
151+
You: robot puppy
152+
Agent: [enhances] → Generates cute robotic puppy
153153
154-
You: make it blue
155-
Agent: [enhances with "blue scales"] → Blue dragon
154+
You: make it playful
155+
Agent: [enhances with "playful pose, wagging tail"] → Playful robot puppy
156156
157-
You: add fire
158-
Agent: [enhances with "breathing fire, flames"] → Dragon with fire
157+
You: add glowing eyes
158+
Agent: [enhances with "bright LED eyes, glowing"] → Glowing-eyed version
159159
160-
You: more dramatic lighting
161-
Agent: [enhances lighting] → Cinematic version
160+
You: more futuristic
161+
Agent: [enhances with "holographic elements, sleek"] → Futuristic design
162162
```
163163

164164
The agent remembers context and refines based on feedback!
@@ -274,7 +274,7 @@ Peer-reviewed best practices, automatically applied to your prompts.
274274

275275
**Try it now:**
276276
```bash
277-
gaia sd "your idea here"
277+
gaia sd "robot puppy"
278278
```
279279

280-
Watch the LLM work its magic!
280+
Watch the LLM transform it into an adorable robotic companion! 🤖🐶

docs/playbooks/sd-agent/index.mdx

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -294,7 +294,7 @@ You: Create an image of a sunset over mountains
294294
Agent: [Enhances prompt and generates image]
295295
Image saved to: ./generated_images/sunset_over_mountains_SD-Turbo_20260129_150000.png
296296
297-
You: Make a robot with AMD logo
297+
You: Make a friendly robot assistant
298298
Agent: [Generates with enhanced prompt]
299299
Image saved to: ./generated_images/robot_with_AMD_logo_SD-Turbo_20260129_150030.png
300300
```

src/gaia/agents/sd/agent.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -23,7 +23,7 @@ class SDAgentConfig:
2323
"""Configuration for SD Agent."""
2424

2525
# SD settings
26-
sd_model: str = "SD-Turbo"
26+
sd_model: str = "SDXL-Turbo"
2727
output_dir: str = ".gaia/cache/sd/images"
2828
prompt_to_open: bool = True
2929

src/gaia/agents/sd/mixin.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -75,7 +75,7 @@ def init_sd(
7575
self,
7676
base_url: str = "http://localhost:8000",
7777
output_dir: Optional[str] = None,
78-
default_model: str = "SD-Turbo",
78+
default_model: str = "SDXL-Turbo",
7979
default_size: Optional[str] = None,
8080
default_steps: Optional[int] = None,
8181
default_cfg: Optional[float] = None,

src/gaia/cli.py

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1008,8 +1008,8 @@ def main():
10081008
"--sd-model",
10091009
dest="sd_model",
10101010
choices=["SD-1.5", "SD-Turbo", "SDXL-Base-1.0", "SDXL-Turbo"],
1011-
default="SD-Turbo",
1012-
help="SD model: SD-Turbo (fast, default), SDXL-Turbo (better quality), SDXL-Base-1.0 (photorealistic, slow)",
1011+
default="SDXL-Turbo",
1012+
help="SD model: SDXL-Turbo (fast, good quality, default), SD-Turbo (faster but lower quality), SDXL-Base-1.0 (photorealistic, slow)",
10131013
)
10141014
sd_parser.add_argument(
10151015
"--size",

src/gaia/llm/lemonade_client.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1620,7 +1620,7 @@ def embeddings(
16201620
def generate_image(
16211621
self,
16221622
prompt: str,
1623-
model: str = "SD-Turbo",
1623+
model: str = "SDXL-Turbo",
16241624
size: Optional[str] = None,
16251625
steps: Optional[int] = None,
16261626
cfg_scale: Optional[float] = None,

tests/unit/test_sd_mixin.py

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -64,7 +64,7 @@ def test_init_sd_custom_config(self, tmp_path, mock_lemonade_client):
6464
default_steps=8,
6565
)
6666

67-
assert mixin.sd_default_model == "SD-Turbo"
67+
assert mixin.sd_default_model == "SDXL-Turbo"
6868
assert mixin.sd_default_size == "512x512"
6969
assert mixin.sd_default_steps == 8
7070

@@ -138,8 +138,8 @@ def test_generate_image_success(self, tmp_path, mock_lemonade_client):
138138

139139
assert result["status"] == "success"
140140
assert Path(result["image_path"]).exists()
141-
assert result["model"] == "SD-Turbo" # Default model
142-
assert result["size"] == "512x512" # Auto-selected for SD-Turbo
141+
assert result["model"] == "SDXL-Turbo" # Default model
142+
assert result["size"] == "512x512" # Auto-selected for SDXL-Turbo
143143
assert "generation_time_ms" in result
144144

145145
def test_generate_image_tracks_history(self, tmp_path, mock_lemonade_client):
@@ -213,7 +213,7 @@ def test_load_model_called_before_generation(self, tmp_path, mock_lemonade_clien
213213
# Verify load_model was called with the model name
214214
mock_lemonade_client.load_model.assert_called()
215215
call_args = mock_lemonade_client.load_model.call_args
216-
assert call_args[0][0] == "SD-Turbo" # Default model
216+
assert call_args[0][0] == "SDXL-Turbo" # Default model
217217

218218

219219
class TestSDToolsMixinHealthCheck:

0 commit comments

Comments
 (0)