jiviteshhp/EarningsScribe

fix: route Qwen inference through Featherless

fix: remove bad import, keep HEAD support on health endpoint

fix: allow HEAD requests on /health for UptimeRobot

fix: restore script tag broken by about section insertion

trigger: rebuild space

fix: IS_GPU true, add about section with metrics

fix: correct chroma path for Docker, pin chromadb==0.4.24

debug: print download path and files

fix: re-fetch collection after download, guard retrieve on empty index

feat: download pre-built chroma index from HF dataset on startup

debug: log chroma collection count

feat: track all chroma files via git lfs

feat: add chroma index via git lfs

fix: use correct HF router URL

fix: use base model via HF router

fix: switch to HF router API

debug: add HF API logging

feat: switch to HF Inference API, remove local model loading

fix: normalize JSON output, fix tokenizer, fix rag path, fix fetch URL

use Qwen chat format prompt

increase max_new_tokens to 1024

restore exact training prompt and generation params

increase max_new_tokens and improve prompt

shorten prompt to fix empty generation

fix generation - use greedy decoding on CPU

add raw output logging and fix dtype warning

fix API URL for HF Spaces

fix pip conflicts - minimal version pins

upgrade peft to 0.18.1 to match adapter

fix all deployment errors

load tokenizer from base model

fix numpy chromadb compatibility

initial deployment

initial commit