Document AI model pipeline

README.md

@@ -13,6 +13,8 @@ pinned: false
 
 Paste an HTML game, describe asset roles like `player`, `background`, or `enemy`, and generate a rewritten version of the game with image data embedded directly into the HTML.
 
-The app interprets the shared art style into explicit per-role image prompts, then generates and embeds game assets as base64 data URIs.
+The app interprets the shared art style and pasted game code into explicit per-role image prompts, then generates and embeds game assets as base64 data URIs.
 
-If the Space has an `HF_TOKEN` secret, it will try the Hugging Face image model in `HF_IMAGE_MODEL` first. The default model is `black-forest-labs/FLUX.1-schnell`. Without that secret, the app uses a local style-aware Pillow fallback that varies watercolor, vector/neon, claymation, oil-painting, anime, and pixel-art looks.
+If the Space has an `HF_TOKEN` secret, it first uses the text model in `HF_PROMPT_MODEL` to interpret the code, roles, and theme into image prompts. The default prompt model is `Qwen/Qwen2.5-Coder-7B-Instruct`.
+
+It then tries the image model in `HF_IMAGE_MODEL` for each role. The default image model is `black-forest-labs/FLUX.1-schnell`. Without `HF_TOKEN`, or if an API call fails, the app uses a local style-aware Pillow fallback that varies watercolor, vector/neon, claymation, oil-painting, anime, pixel-art looks, and common game sprite silhouettes.