You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
StableSteering is a research documentation repository for an interactive system that studies prompt-embedding steering for text-to-image diffusion models.
5
+
<h1align="center">StableSteering</h1>
4
6
5
-
The runtime app is now GPU-only by default and expects CUDA-backed Diffusers inference.
7
+
<palign="center">
8
+
Interactive steering for diffusion image generation, from a user text prompt to preference-guided refinement.
StableSteering is a research-oriented system for interactive image generation with diffusion models.
28
23
29
-
## Folder Guides
24
+
Instead of relying on one-shot prompt rewriting, the system starts from a user text prompt, proposes multiple candidate directions, records user preferences, updates an internal steering state, and generates the next round from that evolving state.
Real Diffusers inference is GPU-only. The app now targets `cuda` explicitly for
131
-
model runs and will fail fast if a CUDA-capable GPU is not available. The
132
-
default server runtime also enforces the `diffusers` backend and never falls
133
-
back to mock automatically. The mock generator is reserved for explicit test
134
-
harnesses only.
178
+
## Banner Asset
135
179
136
-
Trace logs are persisted under `data/traces/`.
137
-
Per-session reports are written under `data/traces/sessions/<session_id>/report.html` and are also available in the app at `/sessions/{session_id}/trace-report`.
180
+
The README banner is stored at [docs/assets/readme_banner.png](./docs/assets/readme_banner.png).
138
181
139
-
Run a one-off real-model smoke test:
182
+
It can be regenerated with:
140
183
141
184
```bash
142
-
python scripts/smoke_real_diffusers.py
185
+
python scripts/generate_readme_banner.py
143
186
```
144
187
145
-
Create a full real GPU-backed example run with a readable HTML walkthrough:
188
+
The generation script expects `GEMINI_API_KEY` in the environment and uses the official Gemini image-generation API.
146
189
147
-
```bash
148
-
python scripts/create_real_e2e_example.py
149
-
```
190
+
## Diagrams And Illustrations
150
191
151
-
Build an optional release zip:
192
+
The documentation layer can include Gemini-generated illustrations to make the Markdown and published HTML easier to scan.
@@ -53,4 +57,4 @@ The zip is written to `output/releases/`.
53
57
54
58
## Release Scope
55
59
56
-
The current release line is a research-oriented MVP. It is suitable for local development, evaluation, and architectural review, but not intended as a production deployment package.
60
+
The current release line is a research-oriented MVP. It is suitable for local development, evaluation, teaching, and architectural review, but not intended as a production deployment package.
0 commit comments