Update README.md

README.md

@@ -5,36 +5,31 @@ colorFrom: blue
 colorTo: gray
 sdk: gradio
 sdk_version: 5.49.1
-python_version: '3.10'
-torch_version: 2.1.2
-cuda_version: '12.1'
 app_file: app.py
 pinned: false
 ---
 
 ## Loomvale Image Lab — SDXL cinematic generator
 
-- Paste **Ambience + 5 Scenes** or **Pull from Google Sheet** (row’s `ImagePrompt_Ambience` + `ImagePrompt_Scenes`).
-- Generates **5 images** per run on **ZeroGPU** and (optionally) writes the image URLs to the sheet column **“AI generated images”** and marks **Assistant = Done**.
+- Paste **Ambience + 5 Scenes** or **Pull from Google Sheet** (`ImagePrompt_Ambience` + `ImagePrompt_Scenes`).
+- Generates **5 images** per run on **ZeroGPU** and (optionally) writes the image URLs back to column **“AI generated images”** and marks **Assistant = Done**.
 
-**Follow us on Instagram** → [@theloomvale](https://instagram.com/theloomvale)  
-Shop the template / automation kit → https://loomvale.gumroad.com/l/social-app
+**Follow [@theloomvale](https://instagram.com/theloomvale)**  
+Shop the template → [loomvale.gumroad.com/l/social-app](https://loomvale.gumroad.com/l/social-app)
 
-### Secrets to set (Settings → Variables & secrets → “New secret”)
-- `SHEET_ID` — your Google Sheet ID
-- `SHEET_NAME` — `Pipeline` (or leave default)
-- `GOOGLE_CREDENTIALS_JSON` — service account JSON
-- `HF_TOKEN` — optional (only needed for private models). Public SDXL works without it.
+### 🔐 Secrets
+- `SHEET_ID`
+- `SHEET_NAME` = `Pipeline`
+- `GOOGLE_CREDENTIALS_JSON`
+- `HF_TOKEN` (optional, only for private models)
 
-### ZeroGPU tips
-- First run can take longer (model download). Subsequent runs reuse cached weights.
-- If you see **HfHubHTTPError**, just try again — the app retries and uses local snapshots after the first success.
-- Runtime is configured for **Gradio 5.49.1** with **PyTorch 2.1.2 / CUDA 12.1**, matching the combinations that the [ZeroGPU guide](https://huggingface.co/docs/hub/spaces-zerogpu) lists as available and the [framework matrix](https://huggingface.co/docs/hub/spaces-gpus#frameworks) documents.
-- Additional dependency pins live in [`requirements.txt`](./requirements.txt); keep them in sync when upgrading the Space SDK (notably `huggingface_hub==0.39.4` to satisfy Gradio’s requirement and `spaces==0.42.0` for Space helpers).
-- Avoid `torch.compile` in this environment — the [PyTorch compilation tutorial](https://docs.pytorch.org/tutorials/intermediate/torch_compile_tutorial.html) highlights GPU warm-up costs that are prohibitive for short ZeroGPU jobs.
+### ⚡ ZeroGPU tips
+- First run downloads weights; later runs are fast.
+- If you hit `HfHubHTTPError`, retry once — cached models load instantly after.
+- No manual Torch/CUDA pins are needed; ZeroGPU provides optimized PyTorch automatically.
 
-### Space file layout
-- [`app.py`](./app.py) — the Gradio entrypoint referenced by `app_file` in the metadata frontmatter. Keep this filename unless you also update the frontmatter.
-- [`requirements.txt`](./requirements.txt) — Python dependency pins installed automatically by the Space runtime.
-- [`README.md`](./README.md) — doubles as the Space “Configuration” tab; the YAML frontmatter above must stay at the top of the file.
-- [`.gitattributes`](./.gitattributes) — optional but recommended for line-ending normalisation and future Large File Storage hooks.
+### 📁 Space layout
+- `app.py` → Gradio UI entrypoint  
+- `requirements.txt` → Python deps  
+- `README.md` → Space metadata + config  
+- `.gitattributes` → line-ending normalization / LFS future-proofing