feat(sight): add strategy breakdown to token savings page

[husterL9]

Description

Add per-strategy visualization and breakdown to the token savings dashboard, enabling users to see which optimization strategies (response compression, schema compression, command rewriting, TOON encoding) contribute most to their token savings.

Changes:

• Frontend: Add strategy pie chart to savings summary card; add strategy column with colored badges and hover tooltips to optimization items table
• Backend: Extend /api/token-savings response with strategy_breakdown array; add session-scoped /api/token-savings/session/{id} endpoint
• Tests: Update AtifViewerPage and TokenSavingsPage tests

Related Issue

closes #984

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional change)
• Performance improvement
• CI/CD or build changes

Scope

• sight (agentsight)

Checklist

• I have read the Contributing Guide
• My code follows the project's code style
• I have added tests that prove my fix is effective or that my feature works
• I have updated the documentation accordingly
• For sight: cargo test pass
• Lock files are up to date (Cargo.lock)

Testing

• cargo test passes (2 passed, 28 ignored, 0 failed)
• Manual verification: token savings page displays strategy pie chart and strategy column correctly with test data

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

linyizhou seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[jfeng18]

感谢 @husterL9 这个策略拆分!特别赞这次补齐了后端 Rust 测试 👍 复核后几个建议:

建议处理:

1. 节省率口径不一致:get_session_savings 用 compounded/(actual+compounded)(handlers.rs:872),而列表页 get_token_savings 用 saved/actual(:1182)——同一 session 在 AtifViewer 卡片与列表页会显示不同的百分比,建议统一口径。
2. 饼图重复切片:strategy_breakdown 按原始 operation 聚合(:703)、用 label 展示(:724),多个未知 operation 会生成多个都叫"其他优化"的切片;建议按 strategy label/id 聚合。
3. 全表扫描:get_session_savings 用 list_sessions_for_savings(0, i64::MAX) 取全部 session 再 .find 单条,每次打开 AtifViewer 都会全表聚合;建议加按 session_id 的单条查询。

小点: compress-toon/compress-schema 归入 tool/mcp 桶的依据建议加注释说明。

另:本 PR 与 #990/#1007 改了同一批文件(handlers.rs / TokenSavingsPage.tsx),合并需注意顺序与 rebase;三者共往 handlers.rs 加了约千行,建议考虑抽出 server/token_savings.rs 收口。

[husterL9]

感谢 review，已全部修复并 force push：

1. 节省率口径统一 — get_session_savings 改为 compounded / total_tokens，与列表页一致
2. 饼图重复切片 — 按 strategy_label 聚合，未知 operation 合并为单个「其他优化」切片
3. 全表扫描 — 新增 get_session_for_savings(session_id) 方法，按 session_id 直接 GROUP BY 查询
4. 分类注释 — 补充了 compress-toon / compress-schema 归桶的说明
5. 模块拆分 — 提取 server/token_savings.rs（~800 行），handlers.rs 瘦身

PTAL 🙏

[jfeng18]

Thanks for the solid work on strategy breakdown! Two items to address:

P1 — Test env pollution

token_savings.rs L668/702/740/773: set_var("HOME", ...) is never restored after each test. Since ENV_MUTEX only guards within this module, other tests in the same binary (e.g. genai::default_path(), tokenless::default_stats_path()) may observe the stale value under parallel execution, causing intermittent CI failures.

P2 — Pie chart colors always fall back to grey

StrategyBreakdown.strategy is populated with the Chinese label (e.g. "Schema 压缩") via grand_strategy_map key (L377), but the frontend STRATEGY_CONFIG is keyed by operation name (compress-schema). The lookup STRATEGY_CONFIG[b.strategy] never matches, so all pie slices render in the default #9ca3af grey.

[jfeng18]

Both issues addressed. HOME env var is now properly saved/restored in all 4 test functions, and StrategyBreakdown.strategy correctly uses the operation key for known ops. LGTM.