This update introduces two new runners, Codex CLI and Claude Code CLI, enhancing the evaluation harness capabilities. The README and relevant files have been updated to reflect these additions, including new options for specifying the runner. Additionally, dependencies have been updated to their latest versions, and the package version has been bumped to 0.0.8.

This commit introduces a comprehensive TypeScript SDK for the web-codegen-scorer CLI, enabling programmatic access to its functionalities. Additionally, detailed documentation for both the SDK and CLI has been added, including usage examples and command references. The .gitignore file has been updated to exclude the .env file.

This commit updates the Codex, Gemini, and Claude Code CLI runners to return detailed output structures, including command information and error handling improvements. Additionally, a new smoke test script has been introduced to validate the functionality of each CLI runner, ensuring they respond correctly to prompts and handle errors gracefully. The package.json file has also been updated to include a new script for running smoke tests.

…in smoke test

This commit enhances the Codex CLI runner by adding detailed logging for schema JSON and command results, aiding in debugging. Additionally, the default Codex model has been updated to 'gpt-5-codex', and the Claude model has been simplified to 'sonnet'. The smoke test script has been adjusted to handle structured responses more gracefully and includes error logging for better visibility during test execution.

…test execution

This commit improves the schemaToPrettyJson function to ensure it correctly handles definitions in the JSON schema. Additionally, it updates the smoke test script to enable the execution of both Gemini and Claude CLI tests, enhancing test coverage and functionality.

This commit refines the model handling in the eval-cli by making the model option optional and updating the default model selection logic. It also enhances the CLI options for better clarity and consistency, including correcting the description for the model option. Additionally, the default models for the Codex and Claude runners have been simplified, and unnecessary constants have been removed from the configuration. Overall, these changes improve the usability and maintainability of the CLI.

…and CLI

This commit deletes the TypeScript SDK documentation, CLI reference, usage examples, and related files, as they are no longer relevant. The removal streamlines the documentation and focuses on the current implementation, ensuring clarity and reducing maintenance overhead.