An example of configuring a Flask-SQLAlchemy session for a relfected Table object from a legacy database.

flask_sqlalchemy_secondary_db_reflection.py

import os
from flask import Flask
from flask_sqlalchemy import SQLAlchemy
from sqlalchemy import MetaData, Table

app = Flask()
db = SQLAlchemy()
db.init_app(app)

SQLALCHEMY_DATABASE_URI = os.environ.get('DATABASE_1_URI')

# Create a Flask-SQLAlchemy bind for the second database
SQLALCHEMY_BINDS = {'db2': os.environ.get('DATABASE_2_URI')}

# Need to give SQLAlchemy object the current app context
with app.app_context():
    # Create an engine for the db2 bind
    engine = db.get_engine(app, bind='db2')

    # Create a new session using the db2-bound engine.  This is a Flask-SQLAlchemy
    # SignallingSession and should behave the same way as db.session (have not tested yet but docs say it will)
    Session = db.create_scoped_session({'bind': engine})
    db2_session = Session()

meta = MetaData(engine)
reflected_table = Table('table_to_reflect', meta, autoload=True, autoload_with=engine)

db2_session.query(reflected_table).first()  # etc...