environment.yml

name: river-level
channels:
  - conda-forge
  - pytorch
  - defaults
dependencies:
  # Add additional Conda dependencies here
  # Version lock critical packages using '=' ie:
  # - packageName=0.123
  - python>=3.8
  - flake8                  # Linter for style adherence
  - mypy                    # Static type checker
  - pytest                  # Unit testing framework
  - boto3                   # AWS S3 interaction
  - s3fs                    # AWS S3 interaction TODO: Convert to only boto3 OR only s3fs (Issue #30)
  - holidays=0.24           # Holiday calendar for time series forecasting. Version locked to avoid upstream issue.
  - pytorch=1.13.1          # Core underlying neural network library
  - pytorch-lightning=1.9.3 # High level interface for pytorch      
  - u8darts=0.23       # Core Timeseries forecasting library
  - pip                     # Package installer for dependencies not available through Conda
  - pip:
    # Add pip packages which are not available through conda here
    # Note: to version lock a pip package, use '==' not '=' ie:
    # - packageName==0.123
    - dataretrieval       # Package for fetching NWIS river level dataa
    - simplekml           # Weather location KML generation
    - pyarrow             # Parquet file creation
    - scikit-learn==1.3   # ML library used internally by darts
    - dataretrieval # Package for fetching NWIS river level dataa
    - simplekml     # Weather location KML generation
    - pyarrow       # Parquet file creation
    - openmeteo-requests==1.1.0 # Handling Open Meteo HTTP Request
    - requests-cache==1.1.1 # Improves performance of HTTP Requests
    - retry-requests==2.0.0 # For retrying failed requests
    - openmeteo-sdk==1.7.0 # For creating a WeatherApiResponse object