Add Grammar-aware constrained decoding

[Ki-Seki]

Fixes #165

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Copilot]

Pull request overview

This PR adds an include_grammar option to model calls so that GIM queries can optionally include tag-level grammar information (currently the regex attribute) directly in the prompt sent to LLM backends, addressing the need for grammar visibility described in #165.

Changes:

• Add include_grammar: bool = False to OpenAI and vLLM model call APIs and plumb it through the shared _call / _acall path.
• Extend prompt construction (get_outlines_model_input) to use a new Query.to_string_with_grammar() representation when include_grammar=True.
• Add/adjust tests to cover the string transformation behavior for include_grammar.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/gimkit/contexts.py	Adds Query.to_string_with_grammar() to include regex in serialized tags.
src/gimkit/models/utils.py	Adds include_grammar plumbing to optionally serialize queries with grammar for model input.
src/gimkit/models/base.py	Threads include_grammar through shared sync/async call helpers.
src/gimkit/models/openai.py	Exposes include_grammar on OpenAI wrappers and forwards it into shared call path.
src/gimkit/models/vllm.py	Exposes include_grammar on vLLM wrappers and forwards it into shared call path.
tests/models/test_utils.py	Adds explicit assertions for include_grammar affecting prompt string generation.
tests/models/test_openai.py	Updates usage to pass include_grammar=True (currently without asserting behavior).
tests/models/test_vllm.py	Updates usage to pass include_grammar=True (currently without asserting behavior).