feat: expand i18n and Vietnamese source support

[nguyen-hung-dev]

Summary

• add a lightweight frontend i18n provider with zh-CN and en locale files
• migrate app shell, navigation, demo shell, and language setting to translation keys
• document the full FE/BE i18n integration roadmap in I18N_PLAN.md

Scope

• keeps zh-CN as the default/source locale
• does not translate uploaded novel content, extracted entity names, or AI-generated analysis output
• leaves backend localization for a later phase

Validation

• cd frontend && npm run build

Closes #11

[mouseart2025]

@nguyen-hung-dev 感谢这个非常有价值的贡献！i18n 基建对项目长期的国际化意义很大。

当前状态：此 PR 涉及 61 文件、+5,107/-1,129 行的大规模 refactor，当前项目处在 EMNLP 2026 ARR 投稿准备期（5/25 截稿），按内部节奏约定，这段时间我们 冻结所有非 bugfix 的功能变更，专注稳定性。即使 CI 通过，61 文件的 review 深度 + 破 UI / 打包 build 的潜在风险对投稿窗口过高。

计划：ARR 录用通知预期在 2026-08-20 前后。此 PR 将在 8/20 后 优先 review 合并。届时我会：

1. 拉 main 最新版本做 conflict resolve
2. 全量验证（Mac + Windows desktop build + live demo + 核心路径 smoke test）
3. 可能分解为 2-3 个更小 PR 以便 review（您的 I18N_PLAN.md 正好给了 phase 划分）

PR 不 close，仍保留开放状态追踪。感谢您的耐心！

[mouseart2025]

Hi @nguyen-hung-dev — thank you for the substantial work here. 13k+ lines across 100+ files with full Vietnamese support is a serious contribution and the multilingual ambition is genuinely appreciated.

Two distinct concerns bundled together

Looking carefully, this PR contains two largely independent pieces of work:

(A) Frontend i18n — well-scoped, low risk:

• frontend/src/i18n/ provider + runtime
• 3 locales (zh-CN / en / vi, 1397 lines each)
• Components migrated to translation keys
• frontend/scripts/i18n/index.mjs extraction tooling
• I18N_PLAN.md

(B) Vietnamese source pipeline — broader surface, touches L3 files:

• New files: source_language_adapter.py, source_language_heuristics.py, domain_labels.py (678 lines), entity_identity.py, utils/source_language.py
• Modified: fact_validator.py (+180), entity_aggregator.py (+157/-88), world_structure_agent.py (+58), analysis_service.py (+61/-26), visualization_service.py (+96), name_authority.py (+28)
• Vietnamese fixture + 3 new test files

Per our CLAUDE.md, those modified backend files are pipeline-critical (L3) and require impact analysis + 5-novel Chinese gold standard regression before merging.

Project timeline

We're submitting to ARR by May 25, 2026 (~2.5 weeks out). The paper's §3.7 contamination-free eval and §3.3 single-root claim both depend on the current pipeline producing reproducible numbers on the 5-novel benchmark. Modifying the L3 files in (B) before submission would mean re-running the full benchmark and updating paper tables — not feasible at this stage.

Proposed path

Could you split this into two PRs?

• PR-12A — frontend i18n only: I'll review and merge this week (after you rebase against the recent v0.71.4–v0.71.6 work).
• PR-12B — Vietnamese pipeline support: I'll give it the careful L3 review it deserves after May 25. The Vietnamese fixture work especially is well thought-out and I want to see it land properly.

Two practical notes regardless:

• This PR currently shows CONFLICTING / DIRTY against main. Recent hotfixes in v0.71.4 (alias merge), v0.71.5 (export), v0.71.6 (LM Studio support) likely touch overlapping code.
• Last CI run was Apr 23, before those changes — a fresh CI is needed.

If splitting isn't workable on your end, I'm happy to keep this PR open and we'll review post-May 25 as a single unit. Just slower.

Truly appreciate the multilingual vision and the careful structure of the Vietnamese adapter. Looking forward to landing this properly.

— Lei

---

中文 TL;DR（备注，方便后续中文读者快速理解）：

感谢非常用心的贡献。这个 PR 实际打包了两件事 — 前端 i18n（低风险，可单独合）和越南语抽取管线支持（新增 1100+ 行模块 + 改动多个 CLAUDE.md 标记的 L3 关键文件，必须做 5 本中文 gold 标准回归才能合）。论文 ARR 5/25 投稿前我没法动 L3 文件，否则要重跑全 benchmark 并更新论文 §3.7/§3.3 的数字。

建议路径：拆成 PR-12A（纯 i18n，本周可合）+ PR-12B（越南语管线，5/25 后认真 review）。如果不愿拆，整体保留到 5/25 后处理也 OK，只是慢一些。

[mouseart2025]

Thanks for the substantial contribution! This is a well-structured i18n framework (13K+ lines, complete provider + locale infrastructure).

Heads up: the project is currently in the ACL Rolling Review evaluation cycle (reviews land ~July 7-13, rebuttal closes ~July 13). To avoid disruptive main-branch changes during the reviewer reproducibility window, I'll do a careful review of this PR after the July rebuttal period ends.

Two things before merge can be considered:

1. Rebase against main — there are conflicts now (main has v0.71.x hotfixes from late May)
2. Vietnamese translation scope — happy to merge the i18n infra now and have Vietnamese as an opt-in second locale; the Chinese-first audience is the current main user base, so en + zh-CN is the core

I'll re-engage on this in mid-July. Apologies for the delay, and thanks for your patience!