Fix ESM service dependencies: add biopython and requests to requirements.txt

fix(startup): avoid importing inverse_folding at module load; use local esm/source; add scipy

fix(deps): add scipy required by esm.inverse_folding.gvp_transformer

fix(import): force local esm/source via sys.path & PYTHONPATH; start via script

build: drop fair-esm; use local esm via PYTHONPATH and script path startup

build: restore fair-esm for esm.pretrained on Space

chore(build): remove fair-esm; use local esm source via PYTHONPATH

fix(build): add fair-esm/torch deps to root requirements to provide esm.pretrained

chore(esM mcp): update requirements for local esm source (torch,fairscale,numpy,tqdm)

feat(esm-mcp): enable variant effect & fixed-backbone; align adapter returns

feat(mcp): add get_prediction_pdb tool (latest/by name) and simplify to read-only

revert(mcp_service): restore file to 05544d4 version

feat(esm mcp): get_prediction_pdb supports optional save_to_path on server

feat(esm mcp): add get_prediction_pdb tool to download saved PDB (supports latest)

feat(esm mcp): save predictions to /data/predictions with fallback

chore(esm mcp): add requests and biopython to requirements for predict_structure

revert: remove incorrect root path handler

fix: correct FastMCP API usage for root path

feat: add root path handler for health check

fix: set default port to 7860 for HF Spaces

build(docker): switch base image to python:3.11

Use python:3.11-slim and upgrade pip for fastmcp

Add fastmcp runtime dependency

Run FastMCP HTTP on 7860; optionalize ESM imports; expose health

Track *.png and *.p with Git LFS

Track binaries with Git LFS

Add esm folder and minimal app

initial commit