Expose model preferences in Context.sample for flexible model selection.

[davenpi]

This PR addresses #524 by exposing the model_preferences parameter in the Context.sample method. This allows server authors to specify model selection preferences—such as model hints or a full ModelPreferences object—when requesting LLM sampling from the client.

Key changes

• Added model_preferences parameter to Context.sample, supporting:
  • A string (single model hint)
  • A list of strings (multiple model hints)
  • A ModelPreferences object (for advanced use)
• Added robust input validation and conversion logic.
• Updated documentation and usage examples to reflect the new parameter.
• Added unit tests for all supported input types and error cases.

Docs

• Updated the docs to include the new parameter

Let me know if you'd like any changes or more tests!

[Copilot]

Pull Request Overview

This PR exposes a new model_preferences parameter in Context.sample to allow callers to specify model selection hints or full preferences, with input validation and updated docs/tests.

• Introduced model_preferences parameter in Context.sample (supports str, list[str], ModelPreferences, or None)
• Added _parse_model_preferences helper with validation and conversion logic
• Updated tests and documentation to cover the new parameter

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
tests/server/test_context.py	Added fixture, imports, and tests for parsing model_preferences inputs
src/fastmcp/server/context.py	Added model_preferences argument to sample and parsing helper method
docs/servers/context.mdx	Updated example usage and method signature to include model_preferences

Comments suppressed due to low confidence (1)

tests/server/test_context.py:86

• Add a test case for None input to verify that _parse_model_preferences(None) returns None without error.

def test_parse_model_preferences_invalid_type(self, context):

[jlowin]

Thanks @davenpi! Made a minor tweak to add typing to the helper function arg, otherwise LGTM

[davenpi]

Awesome. Thanks!