change tokens and add better error handling

apply hard_constraint for receding hairline and hair_type factor, rework score percentage for hairstyles

fix qwen response think

try to fix thinking for qwen

change model

change model and trait_summary function

reformat prompt for llm analysis

rebuild _build_face_analysis_llm prompt and function

change of LLM model and system prompt for face_analysis

fixing llm face analysis errors

add debug and fix trait_summary

adding debug to address face analysis issue

trying to fix face_analysis issue

revert prepare_trait_summary to debug error for face analysis with llm

change trait_summary function

fix face_analysis translation

implement EN/PL translations across backend and frontend components

fix: declaration of lang in main and pipeline

add lang param to analyse endpoint and LLM face analysis

slight correction of model responses

add llm analysis module with llama

add hair type and hairline compatibility

change face analysis function and add hard constraints

change face analysis component by introducing importance

add missing trait dims and filter in main

add missing trait explanations and reformat face analysis component

fix no explanation for low score hairstyle

fix penalty for certain sensitive features

refactor recommendation scoring with missing-feature penalties

change trait explanation in recommender module

Deploy backend API to HF Space