[alcholiclg]

Change Summary

Long-running agents can repeat unsuccessful actions, claim completion despite failed checks, and accumulate tool outputs that are no longer needed.

This PR adds an opt-in experimental Decision mode that uses a separate decision model to review execution and select context while retaining the user's chosen main model.

• Add typed decision requests and responses with a System One-compatible client, validation, timeouts and resource cleanup.
• Add bounded execution and completion checks that can request reflection when evidence indicates a problem.
• Select redundant tool outputs for compression. Use LLM summaries when selection is insufficient, and retain original outputs for later retrieval.
• Add WebUI mode selection and decision-model settings. Lock the mode after execution begins and preserve it across reloads, resumes and forks.
• Add SDK/TUI integration, English and Chinese documentation, regression tests and an opt-in live smoke script.

Related issue number

Checklist

• The pull request title is a good summary of the changes - it will be used in the changelog
• Unit tests for the changes exist
• Run pre-commit install and pre-commit run --all-files before git commit, and passed lint check.
• Documentation reflects the changes where applicable