Spaces:
Sleeping
Sleeping
Commit History
-
Switch LLM stack to Groq + Gemini hybrid; fix resume parse, PDF layout, UI
Strategy_Test +1: Claude Opus 4.8- config/settings.py: provider-aware get_llm() routes groq/* and google/* by model-id prefix (NVIDIA stays the fallback default); new parse_model() tier; Gemini thinking disabled by default (thinking_budget=0) -> tailor ~70s to ~9s. - phase1_parse/parse_resume.py: parse on parse_model(); stronger extraction prompt; deterministic education/phone backfill for fields both LLMs tend to drop. - phase4_tailor/render_pdf.py: tailored PDF now matches the source resume layout (black name, blue links + section rules, two-column skills, blue-italic roles). - frontend: sticky action row so Tailor / Download PDF stay visible in the internally-scrolling cards. - requirements.txt: add langchain-groq, langchain-google-genai. - scripts/compare_providers.py: provider quality+speed benchmark harness. Deploy config (Space settings): Secrets: GROQ_API_KEY, GOOGLE_API_KEY Variables: NVIDIA_MODEL_PARSE/FAST=groq/llama-3.3-70b-versatile, NVIDIA_MODEL_TAILOR=google/gemini-3.5-flas…6aea80d
-
Deploy Resume Buddy (FastAPI + React + Nemotron)
Strategy_Test +1: Claude Opus 4.8Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>abe42b3