fix: allow multi-image uploads by updating global size limits

remove the duplicate text from the modelfile 1-37 lines

enforce 10MB image size limit across all upload endpoints

update the recommendation colors for both text and hex

update the recommendation size for both upper and lower to get an answer in the end even if it is wrong by increas the size numbers

increase the timeout of the recommendation from 60s to 180s to let the llm move from cpu,gpu to the ram in the first request

changing the logic of rendering upper and lower garmrent at time

changing tryon logic

change the smartfilt llm model from lamma3.1(8B) to qwen2.5:3b

update the readme file

change the model from llama3.2:3b to llama3.1 (8B)

updating the pants size table part in the modelfile by fixing the size guide table for the llm

adding a pants size rather than the pants math part in the modelfile by adding a size guide table for the llm

update the pants math part in the modelfile by adding some examples to the llm to give a good answer in math

feat(recommendation): map recommendation colors to hex codes and fix prompt overrides

Update recommendation color by removing the system prompt from the service, so now the llm will just use the omdel file as a guide

Update recommendation color JSON format and the modelfile sturcutre for Hugging Face

Update recommendation JSON format for Hugging Face on router and schemas

Update recommendation service and modelfile prompts with new JSON format for Hugging Face

fix: use printf for GCP secret to preserve multi-line JSON formatting

fix: pass numpy array to avatar estimator to resolve 'Image object has no attribute shape' error

feat: add GCP credential handling and root home page

fix: un-ignore skin tone model and force add it for LFS tracking

refactor: move sam_3d_body inside avatar/ for cleaner structure, resolve via PYTHONPATH

fix: restore sam_3d_body as top-level package, clean avatar module, upgrade Dockerfile to pytorch devel base

fix: install sam_3d_body from source + lazy avatar imports to prevent crash

fix: add zstd required by Ollama install script

fix: add missing system build deps for pycocotools and pyrender

Initial clean commit with LFS for Hugging Face