Use tokenizer options for chat template kwargs

[jennyf19]

Use tokenizer options for chat template kwargs

Summary

• route typed chat template kwargs through the existing tokenizer UpdateOptions API
• remove the parallel GenAI ApplyChatTemplateWithOptions C and C++ surface
• document chat_template_kwargs as a serialized JSON-object tokenizer option
• update the chat template C API test to configure kwargs through UpdateOptions
• pin ONNX Runtime Extensions to Route chat template kwargs through tokenizer options onnxruntime-extensions#1108 merge commit
  7192ff774619fb37358783714ec996a438bfdb05

Why

This is the cleanup discussed after #2421 merged. The existing tokenizer options
API is the forward-compatible surface for extending tokenizer behavior without
adding new C API signatures.

microsoft/onnxruntime-extensions#1108 now makes legacy ApplyChatTemplate
consume validated, typed chat_template_kwargs, while preserving core request
values such as messages, tools, and add_generation_prompt.

The GenAI-specific ApplyChatTemplateWithOptions API was introduced by #2421
but has not shipped in a release. Removing it here avoids establishing a
redundant public surface before release.

Validation

Fork-only native validation passed against the merged Extensions commit:

https://github.com/jennyf19/onnxruntime-genai/actions/runs/33018340933

The run built the native C++ tests with guidance enabled and passed:

CAPITests.ChatTemplate

No production Foundry package pin is changed by this PR. Foundry will consume
the final shape after a GenAI .Foundry package containing this change is
published.

[azure-pipelines]

Azure Pipelines:
There may be pipelines that require an authorized user to comment /azp run to run.

[Copilot]

Pull request overview

This PR consolidates chat-template keyword arguments (chat_template_kwargs) onto the existing tokenizer UpdateOptions surface, removing the parallel GenAI-specific ApplyChatTemplateWithOptions API while documenting the new option and updating C API tests accordingly. It also updates the pinned ONNX Runtime Extensions commit to pick up the upstream validated/typed kwargs support.

Changes:

• Route typed chat_template_kwargs through OgaUpdateTokenizerOptions / Tokenizer::UpdateOptions and use the legacy ApplyChatTemplate call path.
• Remove the C/C++ ApplyChatTemplateWithOptions API surface from GenAI headers and C API implementation.
• Update documentation and tests, and pin onnxruntime_extensions to the specified merge commit.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
test/c_api_tests.cpp	Updates the chat template test to set kwargs via UpdateOptions and call ApplyChatTemplate.
src/ort_genai.h	Removes the C++ convenience wrapper ApplyChatTemplateWithOptions.
src/ort_genai_c.h	Documents chat_template_kwargs as a tokenizer option and removes the C API ApplyChatTemplateWithOptions declaration.
src/ort_genai_c.cpp	Removes the OgaTokenizerApplyChatTemplateWithOptions C API implementation.
src/models/model.h	Removes Tokenizer::ApplyChatTemplateWithOptions declaration.
src/models/model.cpp	Removes Tokenizer::ApplyChatTemplateWithOptions implementation (now relies on OrtxApplyChatTemplate + options).
cmake/deps.txt	Pins onnxruntime_extensions to commit 7192ff774619fb37358783714ec996a438bfdb05.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Baiju Meswani (baijumeswani)]

/azp run macOS arm64 Build

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).