ltcai 11.6.0 → 11.9.0

package/README.md

@@ -11,23 +11,24 @@
 [![CI Status](https://github.com/TaeSooPark-PTS/LatticeAI/actions/workflows/ci.yml/badge.svg?branch=main)](https://github.com/TaeSooPark-PTS/LatticeAI/actions/workflows/ci.yml)
 [![License](https://img.shields.io/badge/license-MIT-green)](LICENSE)
 
-![v11.6.0 Living Brain walkthrough](output/release/v11.6.0/gifs/v11.6.0-living-brain-walkthrough.gif)
+![v11.9.0 Living Brain walkthrough](output/release/v11.9.0/gifs/v11.9.0-living-brain-walkthrough.gif)
 
 Chat, files, folders, notes, and web pages all flow into one durable knowledge
-graph on your computer. Any model — local MLX or cloud — can speak with that
-memory. Nothing leaves your machine without explicit consent.
+graph on your computer. The default is local. Cloud is optional — an
+OAuth-authenticated CLI (`agy` / `grok`) or an API key you configure — and
+nothing leaves your machine without explicit consent.
 
-대화·파일·폴더·웹페이지가 전부 내 컴퓨터 안의 지식 그래프로 쌓이고, 어떤
-모델이든 그 기억을 이어받아 대화합니다.
+대화·파일·폴더·웹페이지가 전부 내 컴퓨터 안의 지식 그래프로 쌓입니다.
+기본은 로컬이고, 클라우드는 선택입니다 (OAuth CLI 지원).
 
 ## What You Can Do
 
 | | |
 | --- | --- |
-| **See your Brain's story in time** — a growth curve, an activity heatmap, and each day's story, rewindable to any past moment ![Brain Chronicle](output/release/v11.6.0/screenshots/13-chronicle.png) | **Chat with a Brain that remembers** — every conversation grows durable, source-linked memory ![Brain Chat](output/release/v11.6.0/screenshots/04-brain-chat-home.png) |
-| **See how knowledge connects** — a real relationship graph, not a file list ![Memory Graph](output/release/v11.6.0/screenshots/05-memory-graph.png) | **Capture anything** — files, whole folders, notes, screenshots, web pages ![Capture](output/release/v11.6.0/screenshots/06-capture.png) |
-| **Automate with review** — agent changes become proposals you approve first ![Review Center](output/release/v11.6.0/screenshots/12-review-center.png) | **Pick a model in one click** — recommended local models for your hardware ![Recommended Models](output/release/v11.6.0/screenshots/02-recommended-models.png) |
-| **Watch a file become memory** — three named steps, not a pipeline diagram ![Material to memory](output/release/v11.6.0/screenshots/11-knowledge-journey.png) | **Say how much it may do alone** — one dial in plain words; dangerous actions stay blocked either way ![Settings](output/release/v11.6.0/screenshots/08-system.png) |
+| **See your Brain's story in time** — a growth curve, an activity heatmap, and each day's story, rewindable to any past moment ![Brain Chronicle](output/release/v11.9.0/screenshots/13-chronicle.png) | **Chat with a Brain that remembers** — every conversation grows durable, source-linked memory ![Brain Chat](output/release/v11.9.0/screenshots/04-brain-chat-home.png) |
+| **See how knowledge connects** — a real relationship graph, not a file list ![Memory Graph](output/release/v11.9.0/screenshots/05-memory-graph.png) | **Capture anything** — files, whole folders, notes, screenshots, web pages ![Capture](output/release/v11.9.0/screenshots/06-capture.png) |
+| **Automate with review** — agent changes become proposals you approve first ![Review Center](output/release/v11.9.0/screenshots/12-review-center.png) | **Pick a model in one click** — recommended local models for your hardware ![Recommended Models](output/release/v11.9.0/screenshots/02-recommended-models.png) |
+| **Watch a file become memory** — three named steps, not a pipeline diagram ![Material to memory](output/release/v11.9.0/screenshots/11-knowledge-journey.png) | **Say how much it may do alone** — one dial in plain words; dangerous actions stay blocked either way ![Settings](output/release/v11.9.0/screenshots/08-system.png) |
 
 ## Why Lattice AI
 
@@ -58,67 +59,49 @@ First-run flow — wake the Brain, pick the owner, load a recommended model:
 
 | | | |
 | --- | --- | --- |
-| ![Login](output/release/v11.6.0/screenshots/01-login.png) | ![Model install](output/release/v11.6.0/screenshots/03-install-load-progress.png) | ![Model library](output/release/v11.6.0/screenshots/07-model-library.png) |
+| ![Login](output/release/v11.9.0/screenshots/01-login.png) | ![Model install](output/release/v11.9.0/screenshots/03-install-load-progress.png) | ![Model library](output/release/v11.9.0/screenshots/07-model-library.png) |
 
 Screenshot index and capture notes:
-[output/release/v11.6.0/SCREENSHOT_INDEX.md](output/release/v11.6.0/SCREENSHOT_INDEX.md)
+[output/release/v11.9.0/SCREENSHOT_INDEX.md](output/release/v11.9.0/SCREENSHOT_INDEX.md)
 
 ## Current Release
 
-The current release is **11.6.0 — One Door**:
-
-The product now has one front door, and it is Rust. `lattice-host` serves
-**420 operations across 41 route families** at the paths they always had,
-and the Python package is no longer a web application at all — it is a
-pure-compute **AI worker** with **28 routes**: LLM inference and streaming,
-embedding, extraction, parsing, four document renderers, speech-to-text,
-multimodal description, the model and engine catalog, `sysinfo`, and
-`/health`. Nine crates hold the product; the door forwards exactly the 28
-and answers `404 {"detail":"Not Found"}` for anything else, from a
-committed allowlist that a drift gate regenerates and compares.
-
-- **Every write is native.** The knowledge-graph write engine moved into
-  `lattice-core`: ingest, curation, provenance, taxonomy and the vector
-  queue. It is held to Python's bytes by a **32-step row-parity battery**
-  that dumps every table after every step with zero tolerated differences,
-  and by a schema comparison over all **67 objects** in `sqlite_master`
-  (indexes and FTS shadow tables included). Seventeen graph tables changed
-  owner from the worker to Rust, and the single-writer invariant is now a
-  test rather than a convention.
-- **The surface is replayed, not re-described.** **1,487 recorded HTTP
-  cases** across twelve committed fixture files — captured from the real
-  Python app while it still served them — are replayed against the native
-  routes, status line and body compared. The retrieval, chunking, agent
-  kernel and agent-loop golden families are unchanged and still green.
-- **298 Python files and 73,617 lines were deleted.** What remains is 127
-  files and about 20,900 lines, all of it compute, at **100.00% statement
-  and branch coverage** with `fail_under=100` and no new pragmas.
-- **What the port found.** The Python oracle's own bugs are documented
-  rather than smoothed over: four graph-write divergences (a `kgv2_*` view
-  that updates `type` on conflict where `nodes` does not; two writers that
-  serialise without `sort_keys`; a hash computed from CPython's default
-  JSON separators), a documented redaction rule that never fired, and a
-  change-proposal kind whitelist that could never approve a folder
-  reorganization. Three known-wrong behaviours were ported **as they are**
-  so the surface does not change under users mid-release — they are listed
-  in [RELEASE_NOTES_v11.6.0.md](RELEASE_NOTES_v11.6.0.md).
-- **What was removed, and why.** The **Telegram bridge** is gone: it lived
-  in the platform code that became the worker, and a bridge with no product
-  server to bridge to is not a feature. **SSO/OIDC login and callback
-  flows** are gone with it — the configuration surface remains, and
-  password login is native. Both are consequences of the worker boundary,
-  not decisions taken on their own merits.
-
-Full disclosure of gaps, ported bugs and removed surfaces:
-[RELEASE_NOTES_v11.6.0.md](RELEASE_NOTES_v11.6.0.md)
-
-Expected artifacts for 11.6.0 release must use exact filenames:
-
-- `dist/ltcai-11.6.0-py3-none-any.whl`
-- `dist/ltcai-11.6.0.tar.gz`
-- `ltcai-11.6.0.tgz`
-- `dist/ltcai-11.6.0.vsix`
-- `src-tauri/target/release/bundle/dmg/Lattice AI_11.6.0_aarch64.dmg`
+The current release is **11.9.0 — Working Order**:
+
+11.8.0 took weight off the door. This release puts the remaining Current
+stubs and the half-wired lanes in working order. The door is unchanged:
+`lattice-host` still serves **420 operations across 41 route families**,
+and the AI worker is still **19 routes** (`/worker/sysinfo` gained
+additive `capabilities` and `python_version`). Cloud is optional; the
+default is local; OAuth CLI (`agy` / `grok`) is supported.
+
+- **Thirteen documented-Current stubs now answer.** Model
+  recommendations probe RAM and the worker catalog. Setup scan/auto are
+  real probes; install is real-or-manual (brew/pip stay manual).
+  Computer-use status reads worker capabilities. `/agent/eval` is a
+  deterministic skill eval. `/agents/api/run` is a live single-agent
+  pass. Automation mines `conversation_messages`. Workflow run is a
+  per-step executor; `build`/`deploy_project` run governed scripts.
+- **Hybrid cloud is wired.** ReviewSink and a shape-only egress audit
+  are bound in production. Dual credentials: `api_key` (mock-verified
+  only) and `cli_oauth` (live OAuth E2E at zero billing). Escalation is
+  `auto` / `manual` / `always`; `local_only` on the request always wins.
+- **MCP is a real JSON-RPC server** at `POST /mcp`. Chat file generation
+  — the v9.2.0 headline deleted in the 11.6.0 port — is restored. The
+  8GB-tier default (gemma-4-e2b) actually runs.
+
+What this release does not close — 2B agent-run quality, the mock-only
+`api_key` path, restore needing a restart, brew/pip staying manual,
+`/mcp` outside the OpenAPI contract, the ad-hoc-signed dmg — is listed
+in [RELEASE_NOTES_v11.9.0.md](RELEASE_NOTES_v11.9.0.md).
+
+Expected artifacts for 11.9.0 release must use exact filenames:
+
+- `dist/ltcai-11.9.0-py3-none-any.whl`
+- `dist/ltcai-11.9.0.tar.gz`
+- `ltcai-11.9.0.tgz`
+- `dist/ltcai-11.9.0.vsix`
+- `src-tauri/target/release/bundle/dmg/Lattice AI_11.9.0_aarch64.dmg`
 
 Do not use wildcard artifact uploads. Package registry publishing remains owner-run.
 
@@ -131,21 +114,24 @@ every product route, owns every write to the Brain, and supervises a Python
 **AI worker** it reaches over loopback for the things a model does — inference,
 embedding, extraction, parsing, rendering, speech-to-text. The React/Vite
 frontend and the Tauri desktop shell sit on top of that one door. Local-first
-by default — cloud model calls, downloads, Brain Network, and update checks are
-opt-in.
+by default — cloud is optional (OAuth CLI or an API key you configure),
+and downloads, Brain Network, and update checks are opt-in.
 
 See [ARCHITECTURE.md](ARCHITECTURE.md) for details and
 [docs/DEVELOPMENT.md](docs/DEVELOPMENT.md) for the developer workflow
-(`npm install && npm run dev`, validation via `npm run lint`,
-`npm run test:unit`, `npm run test:visual`).
+(`npm start` / `bin/ltcai.js` is the product; `npm run dev` is the
+19-route worker). Validation via `npm run lint`, `npm run test:unit`,
+`npm run test:visual`.
 
 ## Known Limitations
 
 - External package registries are owner-published and can lag behind GitHub.
 - SQLite is the live local Brain store. The optional PostgreSQL/pgvector
-  migration tooling is not part of the 11.6.0 worker.
+  migration tooling is not part of the 11.9.0 worker.
 - Docker, model downloads, cloud model calls, Brain Network, and update checks
-  require explicit user action.
+  require explicit user action. Cloud is optional: `cli_oauth` (`agy` /
+  `grok`) was live-checked at zero billing; the `api_key` path is
+  mock-verified only.
 - **The Telegram bridge was removed in 11.6.0** — it lived in the platform code
   that became the AI worker. **SSO/OIDC login and callback flows were removed
   with it**; the configuration surface remains and password login is native.
@@ -157,10 +143,26 @@ See [ARCHITECTURE.md](ARCHITECTURE.md) for details and
   required dependency as of 11.6.0 (it used to be an undeclared lazy import
   that raised a 500; `ltcai[pdf]` remains as an empty alias for older
   install instructions).
-- Known gaps carried openly into this release — upload extraction enrichment is
-  UTF-8-text only, supplied vectors cover the primary ingest node, and the
-  pointer-control tools still execute in the worker — are listed with their
-  reasons in [RELEASE_NOTES_v11.6.0.md](RELEASE_NOTES_v11.6.0.md).
+- The multimodal image and video analysis functions have no HTTP door as of
+  11.8.0: their only route wrapped a native image ingest that was never
+  built. The observation code stays in Brain Core under unit test, and its
+  module header says so.
+- The Python coverage gate is a **line floor of 90** since 11.8.0 (the
+  100%-lines-and-branches gate was removed). 11.9.0 measures 99.97%; the
+  enforced floor is the smaller claim.
+- **2B agent-run quality is gated honestly.** gemma-4-e2b writes the
+  requested file and can still fail the critic on the summary.
+- Restore in a long-lived process may serve pre-restore bytes until
+  connections recycle — restart after restore. brew/pip setup items stay
+  manual. `/mcp` is a real JSON-RPC server and is outside the OpenAPI
+  contract by design.
+- The macOS dmg is ad-hoc signed — effectively unsigned — so first launch
+  needs the usual Gatekeeper step.
+- Pointer-control tools still execute in the worker. Remaining honest gaps
+  (`open_keys` pending-only, no Self-Model refiner, `PART_OF` left on
+  delete, review events silent without an owner, KG-api ingest text-only,
+  two store cycles per review mutation) are listed with their reasons in
+  [RELEASE_NOTES_v11.9.0.md](RELEASE_NOTES_v11.9.0.md).
 
 ## Release History
 
@@ -171,6 +173,9 @@ same boundary. Earlier notes stay in the tree as `RELEASE_NOTES_v*.md` files.
 
 | Version | Theme |
 | --- | --- |
+| 11.9.0 | Working Order |
+| 11.8.0 | Travel Light |
+| 11.7.0 | Clean Sweep |
 | 11.6.0 | One Door |
 | 11.5.2 | Tight Ship |
 | 11.5.1 | Rust Full Loop |

package/docs/BENCHMARKS.md

@@ -1,8 +1,15 @@
 # Model Robustness Benchmarks
 
 > Status: reference 2026-07-21
-
-Methodology and results for `scripts/bench_models.py`, a harness that measures
+>
+> **Dated snapshot.** The harness this page documents no longer exists in the
+> tree: `scripts/bench_models.py` went with the Python agent loop when 11.5.1
+> ported it to Rust, and `latticeai.core.agent_eval` / `scripts/agent_eval.py`
+> followed (the last of them deleted in 11.8.0). The methodology and the
+> honesty notes below are kept as the record of how these numbers were produced
+> and what they did and did not mean — not as instructions you can run today.
+
+Methodology and results for `scripts/bench_models.py`, a harness that measured
 how well the agent loop turns model output into executable actions across model
 quality tiers. Companion to [CI & Release Gates](CI_AND_RELEASE_GATES.md).
 

package/docs/CHANGELOG.md

@@ -4,6 +4,207 @@ The top entry is either the current unreleased main-branch work or the current
 release line. Older entries are historical and may describe behavior as it
 existed at that release.
 
+## [11.9.0] - 2026-08-17 — Working Order
+
+### Added
+- 이중 클라우드 자격증명: `<data_dir>/cloud_provider.json` →
+  `LATTICEAI_CLOUD_API_KEY` → CLI 자동감지(`agy` / `grok`). `api_key`는
+  OpenAI 호환 어댑터(모의 서버만 검증, 실과금 없음). `cli_oauth`는 로컬
+  OAuth CLI(`agy` → gemini-3.7-flash, `grok` → grok-4.6). 중립 temp cwd,
+  120초 타임아웃.
+- `GET /api/cloud/status` — `{configured, mode, provider, model, detail}`.
+- 하이브리드 에스컬레이션 `always` / `auto`(기본) / `manual`
+  (`hybrid_policy.json`). `auto`는 로컬 모델 없음 / 로컬 컨텍스트가 얇음
+  / 명시적 `/cloud `·`클라우드:` 접두사일 때만. 요청
+  `network_mode:"local_only"`가 항상 이김.
+- 호스트가 Review Center + egress 감사 싱크를 바인딩. 클라우드 턴이
+  `kg_cloud_expansion` 제안을 올리고 형태만 있는 egress 레코드를 남김
+  (provider / model / reason, 내용 없음).
+- `POST /mcp` streamable-HTTP JSON-RPC 실서버(`initialize` /
+  `tools/list` / `tools/call`). 큐레이트된 안전 도구 + 스키마가 파싱된
+  스킬 7개. `/mcp/call`이 실제로 디스패치. `/mcp/install`은 정직
+  (스킬/플러그인 활성화, 원격은 수동).
+- SPA 클라우드 투명성: ☁️ 클라우드 답변 칩, 컴포저 경계 힌트, 「이번
+  대화는 로컬만」 토글, System 패널 provider 상태 행.
+- 채팅 파일 생성 복원(v9.2.0 헤드라인, 11.6.0 포트에서 삭제됨). 확장자별
+  앵커 프롬프트, judge + 1회 교정, 정직한 sanitize 배지, 프로젝트는
+  최대 3파일 순차, 워커 렌더 시임의 실제 docx/pdf.
+- `/worker/sysinfo`에 `capabilities`(pointer_tools)와 `python_version`
+  (가산, 라우트 수는 19 유지).
+- 통합 로컬 승인 토큰(`LocalApprovals`) — `/permissions/approve`에서
+  상환.
+
+### Changed
+- 문서에만 있던 13개 Current 스텁이 실동작: `/models/recommendations`,
+  `/setup/scan`+`auto`, `/setup/install`, computer-use 상태,
+  `/agent/eval`, `/agents/api/run`, 자동화 패턴/제안, 워크플로 run·
+  resume, 리뷰 `run_now`, `build`/`deploy_project`, 백업 blob.
+- 클라우드 모델은 provider 설정에서 오고, 로컬 MLX id를 쓰지 않음.
+  로컬 모델이 없어도 클라우드 턴이 가능.
+- 하이브리드 토큰 프레임은 도착하는 대로 스트림. `stream: false`는
+  JSON.
+- 채팅 쓰기의 스코프가 비면 `workspace_id = personal`. null workspace
+  읽기는 personal 가시성.
+- 2B(gemma-4-e2b-it-4bit) compact 프로파일: 태그 스트립·balanced·
+  truncated_close·labeled·fence_rescue 파서, v10.8.0 salvage 삼종
+  복원, COMPACT 컨텍스트 <2k, EXECUTE 온도 고정(compact 0.1 /
+  standard 0.2). MLX가 temperature + stop string을 존중.
+- 백업 = `VACUUM INTO` 스냅샷 + blob + 정직한 매니페스트 + 원자적 복원.
+  export가 edges/chunks를 실음.
+- `POST /agent`가 네이티브 루프에 닿고 제품 정책 표를 실음.
+
+### Fixed
+- 라이브 감사 N1–N9: 에이전트 루프 호스트 바인딩, 빈 Brain 메모리
+  API 500(`conversation_messages` 부트스트랩), chat/memory/chronicle/
+  command가 지식을 봄, brain health의 빈 100점, 백업/복원 정직성,
+  export의 없는 컬럼 조회, 폴더 ingest 승인, 보이스 메모 텍스트 저장.
+- 워크플로 resume 승인 게이트. `write_file`이 비어 있는 content를
+  교정 에러로 거절. critic 판정 복구(verdict-shape + 보수적 토큰,
+  fail-closed 유지).
+- 이전에 고장 나 있던 22항목을 신선한 Brain·실모델로 재검증.
+
+### Known issues
+- 2B 에이전트 루프의 *내용 품질*은 정직하게 게이트된다. gemma-4-e2b는
+  요청한 파일을 기계적으로 쓰지만(마감 테이프 5/5) 요약이 critic에서
+  떨어져 FAILED/NEEDS_REVIEW로 끝날 수 있다. 퍼널은 에이전트 런에 더
+  큰 모델을 권한다.
+- `api_key` 클라우드 경로는 모의 서버만 검증했다. 계약 테스트는 있고
+  실호출은 없다(과금 예산 없음).
+- 오래 떠 있는 프로세스에서 복원하면 커넥션이 재활용되기 전까지
+  복원 전 바이트를 줄 수 있다 — 재시작이 필요하다.
+- brew/pip 셋업 항목은 설계상 수동이다.
+- `/mcp`는 설계상 OpenAPI 제품 계약 밖이다.
+- DMG는 ad-hoc 서명(미서명)이다.
+- 11.8.0이 열어 둔 항목(커버리지 바닥 라인 90, 멀티모달 이미지/비디오
+  HTTP 문 없음, 11.7.0 leftover)은 그대로다. 전문은
+  [RELEASE_NOTES_v11.9.0.md](../RELEASE_NOTES_v11.9.0.md).
+
+## [11.8.0] - 2026-08-16 — Travel Light
+
+### Added
+- 골든 드리프트 가드: 축소된 agent 대표행 세트가 커널의 새 판정 클래스를
+  덮지 않으면 실패. 축소가 조용한 커버리지 손실이 되지 않게 하는 장치.
+- 삭제된 아홉 워커 라우트에 대한 네거티브 테스트 — 게이트웨이가 더 이상
+  전달하지 않음을 Rust KEEP 표와 allowlist 양쪽에서 고정.
+- `rust/fixtures/chunking/FROZEN.md`. 이제 모든 픽스처 계열이 FROZEN.md를
+  가짐.
+- `PhaseBudgets` 토큰 상한(MIN 128 / MAX 8192). 이전에는 하한만 있었음.
+- 모든 GitHub Actions 워크플로에 `concurrency` 그룹과 `timeout-minutes`.
+- Brain Chat Home 재설계: 히어로 컴포저, 3배 커진 LivingBrain(1440에서
+  60px → 179px)과 금빛/옥빛 성장 링, 준비도에 묶인 '기억이 자라고 있어요'
+  캡션, 풀캔버스 그리드, 바닥 연속성 바(지난 대화·현황·기억 지도·기능),
+  `prefers-reduced-motion` 폴백.
+
+### Changed
+- 워커 표면 **28 → 19 라우트**. `rust/fixtures/worker_allowlist.json`도 19.
+- 로컬 lint 체인 **13 → 10 게이트**. 순서와 남은 열 개는 그대로.
+- Python 커버리지 게이트가 `fail_under = 100`(문+분기) → **라인 90**.
+  분기 측정은 꺼짐. 실측은 이번 릴리스에서도 100%.
+- `ci.yml` 중복 스텝 제거 — 4레그 매트릭스는 유지하되 OpenAPI ·
+  product-readiness · 확장 테스트는 3.11+ubuntu 레그에서만, 커버리지 레그는
+  `pytest tests/`를 두 번 돌리지 않음. `release.yml`은 릴리스 스텝만.
+  dependency-audit은 cron 전용, visual은 push+nightly(PR 제외),
+  e2e-sidecar는 nightly 전용.
+- Rust: `lattice-{platform,retrieval,ingest,jobs}`의 약 191개 파일 blanket
+  `#![allow]` 헤더 제거, 약 650건의 clippy/rustc 진단을 원인에서 수정.
+  워크스페이스 수준 허용 0건 추가, 남긴 억제는 이유가 붙은 지역
+  `#[allow(clippy::too_many_arguments)]` 8개.
+- 통합 테스트 파일(= 링크되는 테스트 바이너리) **98 → 56**
+  (lattice-platform 43 → 11). 테스트 함수 삭제 없이 주제별 통합.
+- `ROLE_CAPABILITIES`가 `lattice-auth` 단일 출처. `lattice-agent`가 의존.
+- agent 골든 `decisions__strict` · `calls`가 702 → **171 대표행**
+  (등가류당 한 행).
+- product-readiness 증거를 Rust 픽스처로 재지정. 판정은 COMPLETE 10/10.
+
+### Fixed
+- `SessionStore`가 조회 미스 시 `sessions.json`을 다시 읽음. v11.6.0부터
+  writer는 `lattice-auth`인데 워커는 `__init__`에서 한 번만 읽어, **워커
+  기동 이후의 로그인이 워커에게 보이지 않았음** —
+  `trusted_local_owner`에서는 조용히, `LATTICEAI_REQUIRE_AUTH=true`에서는
+  파일 안에 있는 토큰에 대해 401. `mtime_ns`/`size` 가드 + 1초 throttle을
+  같은 락 아래에서 적용해 추측 버스트가 읽기 버스트가 되지 않게 함.
+  테스트 9개 추가.
+
+### Removed
+- 호출자 0인 워커 라우트 아홉 개: `GET /api/embeddings/providers`,
+  `POST /tools/read_document`, `GET /tools/pdf_pages`,
+  `POST /worker/multimodal/describe`, `GET /api/ingestion/multimodal`,
+  `POST /models/switch/{model_id}`, `DELETE /models/unload-all`,
+  `POST /engines/pull-model`, `GET /api/capture/voice/status`. 파일로는
+  `latticeai/api/{tools,local_files,voice_capture}.py`와
+  `lattice_brain/ingestion/pipeline.py`. `pypdfium2` 의존성도 함께.
+- `latticeai/core/agent_permission.py`; 죽은 보안 헬퍼(`hash_password` /
+  `verify_password`, `check_ip_rate_limit`, `configure_trusted_proxies`,
+  `client_ip`, `bytes_match_extension`); `_kg_common/text.py`의 죽은 청커와
+  호출자 0 함수 9개; 렌더 시임의 `_safe_filename` 이중 살균.
+- 얼어붙은 픽스처의 생성기 2종(`generate_agent_parity_fixtures.py`,
+  `generate_chunking_parity_fixtures.py`),
+  `scripts/{brain_quality_eval,agent_eval,check_python,bench_agent_smoke}.py`,
+  `scripts/check_legacy_debt.mjs`(드리프트한 거울 — 파이썬 테스트가 권위).
+- `.github/workflows/agent-smoke.yml` — 모델 없는 러너에서 이중 fail-open.
+- Rust 죽은 코드: `workspace_scope` 모듈, `WORKSPACE_OS_VERSION` 상수,
+  호출자 0 항목 16개.
+- agent 골든 `decisions__trusted` · `decisions__bypass` 그리드(각 702행) —
+  모든 판정 클래스를 덮는 이름 붙은 단위 테스트로 대체.
+- 죽은 프론트 컴포넌트 `FeedbackState.tsx`, `DepthEmergence.tsx`.
+
+### Known issues
+- 커버리지 **강제 바닥이 100 → 90(라인)** 으로 내려갔고 분기 게이트가
+  없어졌습니다. 실측은 100%지만 게이트가 지키는 것은 바닥입니다.
+- `lattice_brain` 멀티모달의 이미지/비디오 절반에 HTTP 문이 없습니다
+  (관찰 함수는 Brain Core에 남아 단위 테스트를 받고, 모듈 헤더에 명시).
+- DMG는 ad-hoc 서명(미서명)입니다.
+- 삭제된 라우트의 메시지 카탈로그 키는 얼어붙은 패리티 픽스처에 의도적으로
+  남습니다 — 픽스처는 녹화 당시의 표면을 고정하는 기록입니다.
+- `tests/visual/mock_server`가 고아 mock 라우트 하나를 아직 서빙합니다
+  (현재 증거가 해시로 묶여 있어 다음 캡처 사이클에 정리).
+- 11.7.0이 열어 둔 항목은 그대로입니다. 전문은
+  [RELEASE_NOTES_v11.8.0.md](../RELEASE_NOTES_v11.8.0.md).
+
+## [11.7.0] - 2026-08-15 — Clean Sweep
+
+### Added
+- 네이티브 훅 싱크(`HooksStore` → `HookSink`): 사용자 `pre_tool` /
+  `post_tool`이 네이티브 도구에서 발화하고 `hooks_runs.json`에 같은
+  13키 레코드를 남김(상한 100).
+- `review_item_created` / `review_item_updated` 타임라인 이벤트 —
+  모든 변이 리뷰 경로에서 저장 후 1회. `run_now`는 쓰기가 없어 무발화.
+- vault-watch 폴러: enable 시 기동, 마지막 watch disable 시 정지,
+  재시작 시 resume. 감시 `.pdf`/`.docx`는 `/worker/parse`.
+- 좌초 워커 경로 정적 게이트(디코이 증명). 부채 레지스터 비움.
+- 전면 시각 재설계: elevation 사다리, gradient/border 토큰, 유리 제거,
+  토큰 네이티브 cytoscape. 13 스크린 클레임.
+
+### Changed
+- `workspace_os.json` writer 통합 — `WorkspaceOsStore` 하나 +
+  디렉터리 키 레지스트리 + retrieval/agent 포트. 스탬프는
+  `CARGO_PKG_VERSION`.
+- 네 ingest 문(upload / browser-tab / garden note / chat turn)이
+  청크 배치 임베드를 붙임. `(model_id, dim)` 불일치 시 벡터 없음.
+- `/upload/document`가 바이너리 본문을 `/worker/parse`로 보냄.
+- 보안보내기 xlsx가 `/worker/render/xlsx`를 사용.
+- UI 표면이 먹빛 밤 / 한지 낮 elevation 언어. 번들 ~103 KiB gzip.
+
+### Fixed
+- command-search knowledge 그룹이 `matches`를 읽어 결과를 반환
+  (오라클은 `results`를 읽어 항상 비었음).
+- 리뷰 스누즈 offset-aware datetime → 200, 잘못된 `until` → 422.
+- 이중 거절 → 409 (형제 dismiss와 바이트 동일).
+- Self-Model 쓰기 5경로가 은퇴 시임으로 전면 정지 — 네이티브 복구,
+  녹화된 본문 9건 바이트 동일. `resolve_contradiction`이 빈 stamps로
+  "적용됨"을 주장하던 거짓말 포함.
+- chat `ingest_generated` 스키마 400 (11.5.2에서도 깨져 있었음)과
+  vault-watch 미배선.
+- chronicle `@today` 달력 결합, briefing/insights/garden/proactive/
+  health/quality 신선도 퓨즈. 시계 시임 + falsifier + 4시간대.
+
+### Known issues
+- `open_keys`는 `pending`만 (Python은 snoozed도). 추출 refiner 없음.
+  `delete_node`는 `PART_OF`를 남김. owner 없는 리뷰 이벤트는 침묵.
+  KG-api ingest는 텍스트 전용. 리뷰 변이 = 스토어 사이클 2회.
+  snooze 422 detail은 영문 리터럴. 전문은
+  [RELEASE_NOTES_v11.7.0.md](../RELEASE_NOTES_v11.7.0.md).
+
 ## [11.6.0] - 2026-08-15 — One Door
 
 ### Added