docs: generate version-scoped llms.txt artifacts

[jujn]

Summary

Close #2185
This PR replaces the root-level LLM documentation artifacts with version-scoped outputs:

• /v1/llms.txt
• /v1/llms-full.txt
• /v2/llms.txt
• /v2/llms-full.txt

The goal is to avoid mixing AgentScope Java 1.x and 2.x documentation in a single llms-full.txt, so AI coding tools can ingest the correct documentation set for the target version.

Changes

• Add a local Sphinx build-finished extension to generate version-scoped llms.txt and llms-full.txt files from docs/_toc.yml.
• Remove the default sphinx_llms_txt integration to stop generating root-level /llms.txt and /llms-full.txt.
• Filter generated LLM docs to English v1/en and v2/en pages, excluding blogs and community pages.
• Update AI coding docs and the AI context menu to use versioned LLM documentation URLs.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[oss-maintainer]

Summary

Documentation update — generates version-scoped llms.txt artifacts for better LLM context management.

Findings

• [Info] Adds 287 lines of documentation across 6 files. Structure looks good.

Verdict

LGTM.

---

Automated review by github-manager

[AgentScopeJavaBot]

🤖 AI Review

This PR replaces root-level llms.txt/llms-full.txt with version-scoped outputs (/v1/llms.txt, /v2/llms.txt, etc.) to prevent mixing v1 and v2 documentation in a single LLM index file. The implementation is a new Sphinx build-finished extension (versioned_llms.py) that parses _toc.yml and generates per-version artifacts, removing the third-party sphinx-llms-txt dependency.

The overall architecture is clean and consistent with existing extensions like root_index_notice.py. However, there's one notable gap: the v2 root: page (v2/en/intro) defined in _toc.yml is not parsed by _load_ordered_docs(), which only handles - file: lines. This means the v2 intro page won't appear in v2/llms.txt or v2/llms-full.txt, unlike v1 where the intro is listed as a chapter entry.

Additionally, the text.find("\\n---", 4) in _extract_frontmatter() could match --- within frontmatter values, causing premature truncation. Using python-frontmatter library or a regex-based parser would be more robust.

[AgentScopeJavaBot]

🤖 AI Review

This PR replaces root-level llms.txt/llms-full.txt with version-scoped outputs (/v1/llms.txt, /v2/llms.txt, etc.) to prevent mixing v1 and v2 documentation in a single LLM index file. The implementation is a new Sphinx build-finished extension (versioned_llms.py) that parses _toc.yml and generates per-version artifacts, removing the third-party sphinx-llms-txt dependency.

The overall architecture is clean and consistent with existing extensions like root_index_notice.py. However, there's one notable gap: the v2 root: page (v2/en/intro) defined in _toc.yml is not parsed by _load_ordered_docs(), which only handles - file: lines. This means the v2 intro page won't appear in v2/llms.txt or v2/llms-full.txt, unlike v1 where the intro is listed as a chapter entry.

Additionally, the text.find("\\n---", 4) in _extract_frontmatter() could match --- within frontmatter values, causing premature truncation. Using python-frontmatter library or a regex-based parser would be more robust.

[AgentScopeJavaBot]

[major] The _toc.yml v2 entry root: v2/en/intro is not parsed by _load_ordered_docs(), which only matches - file: lines. This means the v2 intro page will be missing from v2/llms.txt and v2/llms-full.txt. Consider adding logic to also parse root: lines at the start of the method, or confirm this is intentional and add a comment explaining why.

[AgentScopeJavaBot]

[minor] text.find("\\n---", 4) can match --- substrings within frontmatter values (e.g., a multi-line description containing --- separators), causing premature frontmatter truncation. Consider using a regex like re.search(r'\\n---\\s*\\n', text[4:]) or the python-frontmatter library for more robust parsing.