Summary

• All 7 workflow model functions (e.g. read_btc_1_hour_price_forecast) accessed result.numbers["key"] directly without checking if the key exists, causing a bare KeyError with no context when the workflow output is missing the expected field
• Added _extract_number() helper that validates key existence before access and raises a descriptive KeyError listing available keys
• read_workflow_wrapper() had a broad except Exception that converted every error (including KeyError) into a generic RuntimeError, losing the original error type and traceback
• Now catches KeyError separately so callers can distinguish "missing output key" from "network/contract failure"
• Added from e to both exception paths to preserve the full exception chain for debugging

Changed files

• src/opengradient/workflow_models/workflow_models.py — added _extract_number() helper, replaced all 7 raw x.numbers["key"] accesses
• src/opengradient/workflow_models/utils.py — added KeyError handler, added from e chaining, updated docstring
• tests/workflow_models_test.py — new test file with 7 tests

Test plan

• test_extracts_valid_key — happy path extraction
• test_raises_on_missing_key — KeyError with descriptive message
• test_raises_on_missing_numbers_attr — handles result without .numbers
• test_reports_available_keys — error message lists available keys
• test_keyerror_preserved_not_converted_to_runtime — KeyError stays KeyError
• test_runtime_error_chains_original — __cause__ is set via from e
• test_success_returns_workflow_output — happy path returns correct output
• All existing tests pass