[namrataghadi-galileo]

Summary

• Add the galileo.decision_models evaluator for invoking a pinned Galileo Decision Models scorer version.
• Send structured Galileo records and authenticated execution context to the scorer-invoke endpoint, then apply the configured threshold and operator to the returned score.
• Keep Decision Models context and runtime-flag definitions within its own package for now; these can be consolidated with Luna later.

Scope

• User-facing: Registers galileo.decision_models with required scorer_id and scorer_version_id, plus optional label, threshold, operator, payload field, and timeout settings.
• Internal: Adds a Decision Models client, request/response models, record and selector handling, and tests for invocation, serialization, comparison, and error cases.
• Out of scope: The separate TypeSafe Jev example remains unchanged.

Risk and Rollout

• Risk level: Medium. This adds a new external evaluator integration and requires the Galileo scorer-invoke runtime, API secret, and invoke URL to be configured.
• Rollback: Remove the galileo.decision_models entry point and evaluator package implementation.

Testing

• Added or updated automated tests.
• Galileo package Ruff lint passed.
• Galileo package mypy typecheck passed.
• All PR CI checks passed, including ci, codecov/patch, sdk-ts-ci, ui-ci, validate, and build-and-push-server.

[codecov]

Codecov Report

❌ Patch coverage is 97.04918% with 9 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...ontrol_evaluator_galileo/decision_models/client.py	94.23%	9 Missing ⚠️

📢 Thoughts on this report? Let us know!