Rainbow Tensor 1.1.1

Rainbow Tensor 1.1.1 makes preview work more predictable and validates the packages that users install.

Highlights:

• Integer arguments consistently accept Python and NumPy integers and custom index values, while rejecting booleans and floats before reading array data.
• Repeated advanced-index selections use shared-coordinate lookups instead of repeated candidate scans.
• max_total_terms=100000 bounds the combined calculation cost of visible outputs, alongside max_terms=10000 for each output.
• Skipped output values do not grow the cache.
• Source archives include all 39 SVG reference fixtures. Both wheel and source installations are checked in isolated environments before publication.

Compatibility:

• Set both max_terms=None and max_total_terms=None to remove calculation limits.
• Evaluation metadata now includes planned output_count, total_terms, and reason fields, with scope="per_output_cell_and_preview".
• Source-panel reads are separate from output calculations. Term limits are not fixed runtime or memory guarantees for arbitrary backends.

Install:

pip install --upgrade rainbow-tensor==1.1.1
pip install --upgrade "rainbow-tensor[interactive]==1.1.1"

Validation covers all 978 test cases across local environments, real CPU checks for PyTorch, JAX, and TensorFlow, fresh package installations, and the updated tutorial.