Issue #696: Ingress v1 Phase 5 — docs + full guards (ingress layer v1 shipped)

[eanzhao]

Issue

Closes #696 — Ingress v1 · Phase 5 — docs + full guards + close #672 #674

Ingress layer v1 shipped. This PR is the top of the 6-PR stacked train (#699 → #703 → #704 → #705 → #709 → this).

Implementation summary

See .implement-loop/runs/implement-issue-696.md.

• Adds the "router = config actor + boundary resolver" entry to docs/canon/architecture-vocabulary.md, citing ADR-0024: Chat Route Policy — ChatRoutePolicyGAgent holds per-scope config, ChatRouteResolver is a stateless boundary decision; no new actor hop, no persisted decision.
• Full guard/test suite run and results recorded honestly in the implement summary.

Stacked-PR position

• Base: feat/2026-05-19_ingress-phase4-voice-ws (Phase 4's branch — top of the stack)
• Head: feat/2026-05-19_ingress-phase5-docs-guards
• Auto-loop iteration: codex-implement-loop / milestone 20 (Ingress v1)

Guard / test status

• docs lint (tools/docs/lint.sh) passed — 43 files, 0 errors.
• architecture_guards.sh: every guard passes except the pre-existing playground asset drift guard (cli-app.js vs demo-app.js) — unrelated, no frontend assets touched in this milestone.
• dotnet test aevatar.slnx: 6 observed failures, all pre-existing baseline (DI-registration / Ornn-NyxIdApiClient / workflow-coverage tests) — none touch ChatRouting; a 1-line docs diff cannot cause them. Per issue body, ~15 baseline failures are expected and excluded.

Remaining milestone closeout (controller / human)

The codex-implement-loop never merges PRs, so the GitHub issue-lifecycle steps in the issue body are deferred to human merge time:

• Update Unify direct chat, NyxID relay, and NyxID Responses through an agent chat router #672 / feat: voice WebSocket as 4th chat-router source_kind — wire VoicePresence into #672 routing #674 trivial-check tables; close Unify direct chat, NyxID relay, and NyxID Responses through an agent chat router #672, feat: voice WebSocket as 4th chat-router source_kind — wire VoicePresence into #672 routing #674, and phase tracking issues Ingress v1 · Phase 0 — chat_route_policy.proto + ADR #691–Ingress v1 · Phase 5 — docs + full guards + close #672 #674 #696 once this stack lands.
• Update the project MEMORY.md ingress-v1 entry from "plan" to "shipped".

🤖 Generated by codex-implement-loop. Reviewer is a Claude subagent (see PR comments for round-N review reports).

[eanzhao]

Review of PR #710 — round 1

Verdict: pass
Issue: #696 — Ingress v1 · Phase 5 — docs + full guards + close #672 #674
Head: feat/2026-05-19_ingress-phase5-docs-guards @ fcf8b6f
Base: feat/2026-05-19_ingress-phase4-voice-ws
Reviewed by: Claude subagent (codex-implement-loop)

Verdict rationale (one paragraph)

The in-repo deliverable is satisfied cleanly. The diff is a single +1/-0 change to docs/canon/architecture-vocabulary.md, adding one "Router" row to the §1.1 "已经存在但容易和上面混淆的词" table. The entry accurately ties router to the "config actor + boundary resolver" shape, names the two shipped types (ChatRoutePolicyGAgent, ChatRouteResolver), correctly states the invariants ("不是新增 actor hop" / "也不持久化决策"), and links ADR-0024 via a path (../adr/0024-chat-route-policy.md) that resolves to a real file. All claims are faithful to ADR-0024 (D1 three-part form / no router actor, D2 transient decisions) and to what shipped in Phases 2-4. The canon file frontmatter is intact, the change is docs-only with no drive-by edits, and the single commit in the three-dot diff is the expected one. The recorded test/guard failures are plausibly pre-existing baseline (DI-registration, Ornn NyxIdApiClient, workflow coverage, playground tsc not found) and cannot be caused by a one-line docs change. Deferred GitHub-lifecycle actions are an intentional controller scoping decision, not a finding.

Findings

F1 — Vocabulary row accurately reflects ADR-0024 and shipped code

• Severity: comment
• Dimension: issue-intent
• Location: docs/canon/architecture-vocabulary.md:41
• Evidence: | Router | 新的转发 Actor / hot-path 调度中心 | 在 ingress 语境里，router = **config actor + boundary resolver**：ChatRoutePolicyGAgent 只作为 per-scope 配置权威持有策略，ChatRouteResolver 只在入口边界做无状态、瞬时决策；它不是新增 actor hop，也不持久化决策。见 [ADR-0024 ...](../adr/0024-chat-route-policy.md)。 |
• Why it's a problem: Not a problem — this is a positive confirmation. The row matches ADR-0024 §D1 ("Three-part form, no router actor"; ChatRoutePolicyGAgent = per-scope config aggregate; ChatRouteResolver = stateless library function) and §D2 ("Decisions are transient ... MUST NOT be persisted"). Both ChatRoutePolicyGAgent.cs and ChatRouteResolver exist in the stack's earlier-phase worktrees. The row was placed in the correct table (§1.1, aevatar 术语 | 不等价于 | 区分), following the established three-column "easily confused terms" pattern rather than the §1 deepening-vocabulary table — a sound editorial choice.
• What would change your verdict: nothing; informational.

What's good (mandatory on pass)

• Docs-only, single file, single commit (fcf8b6fa) — exactly the controller-scoped deliverable; no "顺手清理" violation of the issue's 关键不变量 ("不在 phase 5 内做任何顺手清理").
• Canon frontmatter (title/status/owner) intact, satisfying CLAUDE.md "文档系统" ("所有 docs/canon/ 和 docs/adr/ 文件必须有 YAML frontmatter，包含 title、status、owner 字段"); ADR link path resolves to a real file; no edit to auto-generated docs/README.md.
• Implement summary is honest: it records the dotnet test and architecture_guards.sh failures and correctly attributes them to pre-existing baseline (DI-registration, Ornn NyxIdApiClient, playground tsc not found) outside the docs change, with no attempt to disable tests to go green.

REVIEW_VERDICT:pass:Docs-only router vocabulary entry accurate vs ADR-0024; clean scoped diff

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.43%. Comparing base (f4e9bdf) to head (fcf8b6f).

@@                           Coverage Diff                            @@
##           feat/2026-05-19_ingress-phase4-voice-ws     #710   +/-   ##
========================================================================
  Coverage                                    82.43%   82.43%           
========================================================================
  Files                                          944      944           
  Lines                                        59886    59886           
  Branches                                      7843     7843           
========================================================================
  Hits                                         49369    49369           
  Misses                                        7133     7133           
  Partials                                      3384     3384           

Flag	Coverage Δ
ci	82.43% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.