coderouter-cli 2.5.0__tar.gz → 2.5.1__tar.gz

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/CHANGELOG.md

@@ -6,6 +6,38 @@ versioning follows [SemVer](https://semver.org/).
 
 ---
 
+## [v2.5.1] — 2026-05-22 (MLX backend + docs reorganization)
+
+Patch release: a third Launcher backend, a reorganized documentation
+tree, and a security fix.
+
+### Added
+
+- **MLX backend** for the Launcher (`launcher_gui.py` and
+  `coderouter/ingress/launcher_routes.py`): `mlx` joins `llama.cpp` and
+  `vllm`, aimed at Apple Silicon users. Launches
+  `python -m mlx_lm.server --model <m> --port <p>`. The backend
+  selectors (desktop GUI combobox / Web `<select>`) gain an `mlx`
+  option, and the binary-not-found error messages are now
+  backend-agnostic.
+
+### Changed
+
+- **`docs/` reorganized** into role-based folders — `start/`, `guides/`,
+  `backends/`, `concepts/` — with a new bilingual (JA/EN) master index
+  at `docs/README.md` including a quick "what to read" table. Internal
+  cross-links, `README.md` / `README.en.md`, and code/config path
+  references were updated to the new layout.
+- **`plan.md` restructured**: deduplicated, version ordering fixed,
+  sections compressed (1747 → 721 lines).
+
+### Security
+
+- **starlette `1.0.0` → `1.0.1`** (`uv.lock`): fixes PYSEC-2026-161,
+  which failed the `cve-audit` CI job.
+
+---
+
 ## [v2.5.0] — 2026-05-22 (Launcher — llama.cpp / vllm GUI)
 
 Browser-based process manager for local inference backends, integrated

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: coderouter-cli
-Version: 2.5.0
+Version: 2.5.1
 Summary: Local-first, free-first, fallback-built-in LLM router. Claude Code / OpenAI compatible.
 Project-URL: Homepage, https://github.com/zephel01/CodeRouter
 Project-URL: Repository, https://github.com/zephel01/CodeRouter
@@ -54,7 +54,7 @@ Description-Content-Type: text/markdown
 </p>
 
 <p align="center">
-  <a href="./README.en.md">English</a> · <strong>日本語</strong> · <a href="./docs/quickstart.md">10 分で動かす</a> · <a href="./docs/architecture.md">設計詳細</a>
+  <a href="./README.en.md">English</a> · <strong>日本語</strong> · <a href="./docs/start/quickstart.md">10 分で動かす</a> · <a href="./docs/concepts/architecture.md">設計詳細</a>
 </p>
 
 ---
@@ -121,7 +121,7 @@ ANTHROPIC_BASE_URL=http://localhost:8088 ANTHROPIC_AUTH_TOKEN=dummy claude
 | codex / gemini-cli + Ollama 直繋ぎで動いてる | オプション — フォールバックが欲しいなら |
 | Claude API を直接叩いてて問題ない | 不要 |
 
-詳細は → [要否判定ガイド](./docs/when-do-i-need-coderouter.md)
+詳細は → [要否判定ガイド](./docs/start/when-do-i-need-coderouter.md)
 
 ---
 
@@ -186,7 +186,7 @@ launcher:
           "--max-model-len": 4096
 ```
 
-詳細 → [Launcher ガイド](./docs/launcher.md)
+詳細 → [Launcher ガイド](./docs/backends/launcher.md)
 
 ---
 
@@ -213,7 +213,7 @@ providers:
     api_key_env: OPENROUTER_API_KEY
 ```
 
-もっと詳しい設定 → [利用ガイド](./docs/usage-guide.md) · [設計詳細](./docs/architecture.md)
+もっと詳しい設定 → [利用ガイド](./docs/guides/usage-guide.md) · [設計詳細](./docs/concepts/architecture.md)
 
 ---
 
@@ -221,15 +221,15 @@ providers:
 
 | やりたいこと | ドキュメント |
 |---|---|
-| すぐ動かす | [Quickstart](./docs/quickstart.md) |
-| 使いこなす | [利用ガイド](./docs/usage-guide.md) |
-| 無料で回す | [無料枠ガイド](./docs/free-tier-guide.md) |
-| llama.cpp / vllm を GUI で起動 | [Launcher ガイド](./docs/launcher.md) |
-| 詰まった | [トラブルシューティング](./docs/troubleshooting.md) |
-| 設計を知りたい | [アーキテクチャ詳細](./docs/architecture.md) |
+| すぐ動かす | [Quickstart](./docs/start/quickstart.md) |
+| 使いこなす | [利用ガイド](./docs/guides/usage-guide.md) |
+| 無料で回す | [無料枠ガイド](./docs/guides/free-tier-guide.md) |
+| llama.cpp / vllm を GUI で起動 | [Launcher ガイド](./docs/backends/launcher.md) |
+| 詰まった | [トラブルシューティング](./docs/guides/troubleshooting.md) |
+| 設計を知りたい | [アーキテクチャ詳細](./docs/concepts/architecture.md) |
 | 全リリース履歴 | [CHANGELOG](./CHANGELOG.md) |
 
-English: [Quickstart](./docs/quickstart.en.md) · [Usage guide](./docs/usage-guide.en.md) · [Free-tier](./docs/free-tier-guide.en.md) · [Troubleshooting](./docs/troubleshooting.en.md)
+English: [Quickstart](./docs/start/quickstart.en.md) · [Usage guide](./docs/guides/usage-guide.en.md) · [Free-tier](./docs/guides/free-tier-guide.en.md) · [Troubleshooting](./docs/guides/troubleshooting.en.md)
 
 ---
 
@@ -239,10 +239,10 @@ English: [Quickstart](./docs/quickstart.en.md) · [Usage guide](./docs/usage-gui
 
 | 症状 | 原因 | 詳細 |
 |---|---|---|
-| 401 エラー | API キー未設定 / `.env` に `export` 忘れ | [§1](./docs/troubleshooting.md#1-起動設定で踏みやすい-5-つの罠-v162-追加) |
-| 返信が空 / 意味不明 | Ollama の `num_ctx` が 2048 に切り詰め | [§3](./docs/troubleshooting.md#3-ollama-初心者--サイレント失敗-5-症状-v07-c) |
-| `<think>` タグが漏れる | `output_filters: [strip_thinking]` を付ける | [§3](./docs/troubleshooting.md#3-ollama-初心者--サイレント失敗-5-症状-v07-c) |
-| Claude Code でツール呼び出しがおかしい | tool-call 修復が効いてない | [§4](./docs/troubleshooting.md#4-claude-code-連携で踏みやすい罠-v162-追加) |
+| 401 エラー | API キー未設定 / `.env` に `export` 忘れ | [§1](./docs/guides/troubleshooting.md#1-起動設定で踏みやすい-5-つの罠-v162-追加) |
+| 返信が空 / 意味不明 | Ollama の `num_ctx` が 2048 に切り詰め | [§3](./docs/guides/troubleshooting.md#3-ollama-初心者--サイレント失敗-5-症状-v07-c) |
+| `<think>` タグが漏れる | `output_filters: [strip_thinking]` を付ける | [§3](./docs/guides/troubleshooting.md#3-ollama-初心者--サイレント失敗-5-症状-v07-c) |
+| Claude Code でツール呼び出しがおかしい | tool-call 修復が効いてない | [§4](./docs/guides/troubleshooting.md#4-claude-code-連携で踏みやすい罠-v162-追加) |
 
 `http://localhost:8088/dashboard` を開いておくと、ほとんどの問題が見て 10 秒でわかります。
 
@@ -268,7 +268,7 @@ CodeRouter は backend ルーター層として独立して動きます。`OPENA
 
 ## Security
 
-シークレットは環境変数に置きます。[`docs/security.md`](./docs/security.md) に完全な方針と報告手順があります。
+シークレットは環境変数に置きます。[`docs/security.md`](./docs/guides/security.md) に完全な方針と報告手順があります。
 
 ## License
 

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/README.en.md

@@ -13,7 +13,7 @@
 </p>
 
 <p align="center">
-  <strong>English</strong> · <a href="./README.md">日本語</a> · <a href="./docs/quickstart.en.md">Get started in 10 min</a> · <a href="./docs/architecture.md">Architecture</a>
+  <strong>English</strong> · <a href="./README.md">日本語</a> · <a href="./docs/start/quickstart.en.md">Get started in 10 min</a> · <a href="./docs/concepts/architecture.md">Architecture</a>
 </p>
 
 ---
@@ -80,7 +80,7 @@ That's it. Claude Code works as usual, but your local Ollama is answering behind
 | codex / gemini-cli + Ollama works fine | Optional — if you want fallback |
 | Using Claude API directly, no issues | Not needed |
 
-Full decision matrix → [Do I need CodeRouter?](./docs/when-do-i-need-coderouter.en.md)
+Full decision matrix → [Do I need CodeRouter?](./docs/start/when-do-i-need-coderouter.en.md)
 
 ---
 
@@ -145,7 +145,7 @@ launcher:
           "--max-model-len": 4096
 ```
 
-Details → [Launcher guide](./docs/launcher.md)
+Details → [Launcher guide](./docs/backends/launcher.md)
 
 ---
 
@@ -172,7 +172,7 @@ providers:
     api_key_env: OPENROUTER_API_KEY
 ```
 
-More detail → [Usage guide](./docs/usage-guide.en.md) · [Architecture](./docs/architecture.md)
+More detail → [Usage guide](./docs/guides/usage-guide.en.md) · [Architecture](./docs/concepts/architecture.md)
 
 ---
 
@@ -180,15 +180,15 @@ More detail → [Usage guide](./docs/usage-guide.en.md) · [Architecture](./docs
 
 | Goal | Document |
 |---|---|
-| Get running fast | [Quickstart](./docs/quickstart.en.md) |
-| Use it well | [Usage guide](./docs/usage-guide.en.md) |
-| Run for free | [Free-tier guide](./docs/free-tier-guide.en.md) |
-| Launch llama.cpp / vllm via GUI | [Launcher guide](./docs/launcher.md) |
-| Stuck? | [Troubleshooting](./docs/troubleshooting.en.md) |
-| Understand the design | [Architecture](./docs/architecture.md) |
+| Get running fast | [Quickstart](./docs/start/quickstart.en.md) |
+| Use it well | [Usage guide](./docs/guides/usage-guide.en.md) |
+| Run for free | [Free-tier guide](./docs/guides/free-tier-guide.en.md) |
+| Launch llama.cpp / vllm via GUI | [Launcher guide](./docs/backends/launcher.md) |
+| Stuck? | [Troubleshooting](./docs/guides/troubleshooting.en.md) |
+| Understand the design | [Architecture](./docs/concepts/architecture.md) |
 | Full release history | [CHANGELOG](./CHANGELOG.md) |
 
-日本語: [Quickstart](./docs/quickstart.md) · [利用ガイド](./docs/usage-guide.md) · [無料枠ガイド](./docs/free-tier-guide.md) · [トラブルシューティング](./docs/troubleshooting.md)
+日本語: [Quickstart](./docs/start/quickstart.md) · [利用ガイド](./docs/guides/usage-guide.md) · [無料枠ガイド](./docs/guides/free-tier-guide.md) · [トラブルシューティング](./docs/guides/troubleshooting.md)
 
 ---
 
@@ -198,10 +198,10 @@ More detail → [Usage guide](./docs/usage-guide.en.md) · [Architecture](./docs
 
 | Symptom | Cause | Details |
 |---|---|---|
-| 401 error | API key not set / missing `export` in `.env` | [§1](./docs/troubleshooting.en.md#1-five-startup--config-gotchas-added-in-v162) |
-| Empty / garbage replies | Ollama `num_ctx` truncated to 2048 | [§3](./docs/troubleshooting.en.md#3-ollama-beginner--5-silent-fail-symptoms-v07-c) |
-| `<think>` tags leaking | Add `output_filters: [strip_thinking]` | [§3](./docs/troubleshooting.en.md#3-ollama-beginner--5-silent-fail-symptoms-v07-c) |
-| Tool calls misbehaving in Claude Code | Tool-call repair not kicking in | [§4](./docs/troubleshooting.en.md#4-claude-code-integration-gotchas-added-in-v162) |
+| 401 error | API key not set / missing `export` in `.env` | [§1](./docs/guides/troubleshooting.en.md#1-five-startup--config-gotchas-added-in-v162) |
+| Empty / garbage replies | Ollama `num_ctx` truncated to 2048 | [§3](./docs/guides/troubleshooting.en.md#3-ollama-beginner--5-silent-fail-symptoms-v07-c) |
+| `<think>` tags leaking | Add `output_filters: [strip_thinking]` | [§3](./docs/guides/troubleshooting.en.md#3-ollama-beginner--5-silent-fail-symptoms-v07-c) |
+| Tool calls misbehaving in Claude Code | Tool-call repair not kicking in | [§4](./docs/guides/troubleshooting.en.md#4-claude-code-integration-gotchas-added-in-v162) |
 
 Open `http://localhost:8088/dashboard` while debugging — most issues become visible in 10 seconds.
 
@@ -227,7 +227,7 @@ CodeRouter runs as an independent backend router layer. Point any project's `OPE
 
 ## Security
 
-Secrets go in env vars, not config files. See [`docs/security.en.md`](./docs/security.en.md) for the full policy and reporting instructions.
+Secrets go in env vars, not config files. See [`docs/security.en.md`](./docs/guides/security.en.md) for the full policy and reporting instructions.
 
 ## License
 

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/README.md

@@ -13,7 +13,7 @@
 </p>
 
 <p align="center">
-  <a href="./README.en.md">English</a> · <strong>日本語</strong> · <a href="./docs/quickstart.md">10 分で動かす</a> · <a href="./docs/architecture.md">設計詳細</a>
+  <a href="./README.en.md">English</a> · <strong>日本語</strong> · <a href="./docs/start/quickstart.md">10 分で動かす</a> · <a href="./docs/concepts/architecture.md">設計詳細</a>
 </p>
 
 ---
@@ -80,7 +80,7 @@ ANTHROPIC_BASE_URL=http://localhost:8088 ANTHROPIC_AUTH_TOKEN=dummy claude
 | codex / gemini-cli + Ollama 直繋ぎで動いてる | オプション — フォールバックが欲しいなら |
 | Claude API を直接叩いてて問題ない | 不要 |
 
-詳細は → [要否判定ガイド](./docs/when-do-i-need-coderouter.md)
+詳細は → [要否判定ガイド](./docs/start/when-do-i-need-coderouter.md)
 
 ---
 
@@ -145,7 +145,7 @@ launcher:
           "--max-model-len": 4096
 ```
 
-詳細 → [Launcher ガイド](./docs/launcher.md)
+詳細 → [Launcher ガイド](./docs/backends/launcher.md)
 
 ---
 
@@ -172,7 +172,7 @@ providers:
     api_key_env: OPENROUTER_API_KEY
 ```
 
-もっと詳しい設定 → [利用ガイド](./docs/usage-guide.md) · [設計詳細](./docs/architecture.md)
+もっと詳しい設定 → [利用ガイド](./docs/guides/usage-guide.md) · [設計詳細](./docs/concepts/architecture.md)
 
 ---
 
@@ -180,15 +180,15 @@ providers:
 
 | やりたいこと | ドキュメント |
 |---|---|
-| すぐ動かす | [Quickstart](./docs/quickstart.md) |
-| 使いこなす | [利用ガイド](./docs/usage-guide.md) |
-| 無料で回す | [無料枠ガイド](./docs/free-tier-guide.md) |
-| llama.cpp / vllm を GUI で起動 | [Launcher ガイド](./docs/launcher.md) |
-| 詰まった | [トラブルシューティング](./docs/troubleshooting.md) |
-| 設計を知りたい | [アーキテクチャ詳細](./docs/architecture.md) |
+| すぐ動かす | [Quickstart](./docs/start/quickstart.md) |
+| 使いこなす | [利用ガイド](./docs/guides/usage-guide.md) |
+| 無料で回す | [無料枠ガイド](./docs/guides/free-tier-guide.md) |
+| llama.cpp / vllm を GUI で起動 | [Launcher ガイド](./docs/backends/launcher.md) |
+| 詰まった | [トラブルシューティング](./docs/guides/troubleshooting.md) |
+| 設計を知りたい | [アーキテクチャ詳細](./docs/concepts/architecture.md) |
 | 全リリース履歴 | [CHANGELOG](./CHANGELOG.md) |
 
-English: [Quickstart](./docs/quickstart.en.md) · [Usage guide](./docs/usage-guide.en.md) · [Free-tier](./docs/free-tier-guide.en.md) · [Troubleshooting](./docs/troubleshooting.en.md)
+English: [Quickstart](./docs/start/quickstart.en.md) · [Usage guide](./docs/guides/usage-guide.en.md) · [Free-tier](./docs/guides/free-tier-guide.en.md) · [Troubleshooting](./docs/guides/troubleshooting.en.md)
 
 ---
 
@@ -198,10 +198,10 @@ English: [Quickstart](./docs/quickstart.en.md) · [Usage guide](./docs/usage-gui
 
 | 症状 | 原因 | 詳細 |
 |---|---|---|
-| 401 エラー | API キー未設定 / `.env` に `export` 忘れ | [§1](./docs/troubleshooting.md#1-起動設定で踏みやすい-5-つの罠-v162-追加) |
-| 返信が空 / 意味不明 | Ollama の `num_ctx` が 2048 に切り詰め | [§3](./docs/troubleshooting.md#3-ollama-初心者--サイレント失敗-5-症状-v07-c) |
-| `<think>` タグが漏れる | `output_filters: [strip_thinking]` を付ける | [§3](./docs/troubleshooting.md#3-ollama-初心者--サイレント失敗-5-症状-v07-c) |
-| Claude Code でツール呼び出しがおかしい | tool-call 修復が効いてない | [§4](./docs/troubleshooting.md#4-claude-code-連携で踏みやすい罠-v162-追加) |
+| 401 エラー | API キー未設定 / `.env` に `export` 忘れ | [§1](./docs/guides/troubleshooting.md#1-起動設定で踏みやすい-5-つの罠-v162-追加) |
+| 返信が空 / 意味不明 | Ollama の `num_ctx` が 2048 に切り詰め | [§3](./docs/guides/troubleshooting.md#3-ollama-初心者--サイレント失敗-5-症状-v07-c) |
+| `<think>` タグが漏れる | `output_filters: [strip_thinking]` を付ける | [§3](./docs/guides/troubleshooting.md#3-ollama-初心者--サイレント失敗-5-症状-v07-c) |
+| Claude Code でツール呼び出しがおかしい | tool-call 修復が効いてない | [§4](./docs/guides/troubleshooting.md#4-claude-code-連携で踏みやすい罠-v162-追加) |
 
 `http://localhost:8088/dashboard` を開いておくと、ほとんどの問題が見て 10 秒でわかります。
 
@@ -227,7 +227,7 @@ CodeRouter は backend ルーター層として独立して動きます。`OPENA
 
 ## Security
 
-シークレットは環境変数に置きます。[`docs/security.md`](./docs/security.md) に完全な方針と報告手順があります。
+シークレットは環境変数に置きます。[`docs/security.md`](./docs/guides/security.md) に完全な方針と報告手順があります。
 
 ## License
 

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/coderouter/cli.py

@@ -61,7 +61,7 @@ def _build_parser() -> argparse.ArgumentParser:
             "binding the server. Repeat to layer multiple files. By "
             "default, file values do NOT override variables already in "
             "the environment (the shell `export` wins). See "
-            "docs/troubleshooting.md §5 for 1Password / direnv / sops "
+            "docs/guides/troubleshooting.md §5 for 1Password / direnv / sops "
             "integration recipes."
         ),
     )
@@ -116,7 +116,7 @@ def _build_parser() -> argparse.ArgumentParser:
             "POSIX file mode (0600 expected), .gitignore coverage, "
             "and git-tracking state. Bare `--check-env` (no PATH) "
             "looks for `./.env` then `~/.coderouter/.env`. "
-            "See docs/troubleshooting.md §5 for the threat model."
+            "See docs/guides/troubleshooting.md §5 for the threat model."
         ),
     )
     doctor.add_argument(

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/coderouter/config/capability_registry.py

@@ -109,7 +109,7 @@ class RegistryCapabilities(BaseModel):
             "harness. ``degraded`` = the model over-eagerly invokes "
             "tools/skills when given Claude Code's system prompt — e.g. "
             "Llama-3.3-70B treating small talk like ``こんにちは`` as "
-            "``Skill(hello)`` invocations (see docs/troubleshooting.md "
+            "``Skill(hello)`` invocations (see docs/guides/troubleshooting.md "
             "§4-1 for the symptom log). ``ok`` = explicitly verified "
             "clean. ``None`` = no opinion (treated as ``ok`` at the "
             "startup check)."

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/coderouter/data/model-capabilities.yaml

@@ -35,7 +35,7 @@
 #                                       Claude Code's agentic-coding harness;
 #                                       "degraded" triggers a startup WARN when
 #                                       the provider is on a `claude-code-*`
-#                                       chain. See docs/troubleshooting.md §4-1.
+#                                       chain. See docs/guides/troubleshooting.md §4-1.
 #
 # First-match semantics: rules within a file are evaluated top-to-bottom
 # per flag; the first rule whose glob matches AND declares that flag
@@ -153,7 +153,7 @@ rules:
   # Llama-3.3-70B (verified 2026-04-24 against NVIDIA NIM) rewrites
   # ``こんにちは`` into ``Skill(hello)`` invocations and fabricates
   # ``AskUserQuestion("What is your name?")`` elicitations — see
-  # docs/articles/note-nvidia-nim.md §6-2 + docs/troubleshooting.md §4-1.
+  # docs/articles/note/note-nvidia-nim.md §6-2 + docs/guides/troubleshooting.md §4-1.
   #
   # Glob coverage: NIM uses ``meta/llama-3.3-70b-instruct``, OpenRouter
   # uses ``meta-llama/llama-3.3-70b-instruct``, some local servers use

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/coderouter/ingress/launcher_routes.py

@@ -1,13 +1,13 @@
 """Launcher routes — ``GET /launcher`` + ``/api/launcher/*``.
 
-llama.cpp / vllm プロセス管理 UI。
+llama.cpp / vllm / mlx プロセス管理 UI。
 
 設計方針:
 - ダッシュボードと同じ "1ファイル完結" スタイル (Tailwind CDN + inline JS)
 - プロセスレジストリは app.state.launcher に持たせる (再起動で消えるが意図通り)
 - option_profiles は providers.yaml の launcher: セクションで管理 → コード変更不要で拡張可
 - 複数プロセスの同時起動に対応 (UUID ベースの ID 管理)
-- llama.cpp / vllm どちらも同じ key-value args スキーマで統一
+- llama.cpp / vllm / mlx いずれも同じ key-value args スキーマで統一
 
 エンドポイント:
   GET  /launcher                   → HTML UI
@@ -62,7 +62,7 @@ class ManagedProcess:
 
     id: str
     name: str
-    backend: str         # "llama.cpp" | "vllm"
+    backend: str         # "llama.cpp" | "vllm" | "mlx"
     model_path: str
     port: int
     options: dict[str, Any]
@@ -150,6 +150,7 @@ def _scan_models(model_dirs: list[str]) -> list[dict[str, Any]]:
 _BACKEND_DEFAULTS: dict[str, str] = {
     "llama.cpp": "llama-server",
     "vllm": "python",
+    "mlx": "python",          # mlx_lm.server (Apple Silicon 向け)
 }
 
 
@@ -312,8 +313,17 @@ def _build_cmd(
             "--model", model_path,
             "--port", str(port),
         ]
+    elif backend == "mlx":
+        cmd = [
+            exe, "-m", "mlx_lm.server",
+            "--model", model_path,
+            "--port", str(port),
+        ]
     else:
-        raise ValueError(f"Unknown backend: {backend!r}. Expected 'llama.cpp' or 'vllm'.")
+        raise ValueError(
+            f"Unknown backend: {backend!r}. "
+            "Expected 'llama.cpp', 'vllm' or 'mlx'."
+        )
 
     for flag, val in options.items():
         if isinstance(val, bool):
@@ -720,6 +730,7 @@ _LAUNCHER_HTML = r"""<!doctype html>
         <select id="f-backend" onchange="onBackendChange()">
           <option value="llama.cpp">llama.cpp</option>
           <option value="vllm">vllm</option>
+          <option value="mlx">mlx</option>
         </select>
         <div id="binary-hint" class="mt-1 text-xs text-slate-500 min-h-[1.2rem]"></div>
       </div>
@@ -961,7 +972,7 @@ _LAUNCHER_HTML = r"""<!doctype html>
     // Enable/disable launch button based on binary availability
     if (!info.found) {
       btn.disabled = true;
-      showLaunchErr(`⚠ "${esc(info.resolved)}" が見つかりません。llama.cpp をインストールするか、providers.yaml の launcher.backends.llama\\.cpp.binary にフルパスを設定してください。`);
+      showLaunchErr(`⚠ "${esc(info.resolved)}" が見つかりません。選択中のバックエンド (${esc(backend)}) をインストールするか、providers.yaml の launcher.backends.${esc(backend)}.binary にフルパスを設定してください。`);
     } else {
       btn.disabled = false;
       // Clear error only if it was a binary-not-found error

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/coderouter/logging.py

@@ -779,7 +779,7 @@ def log_output_filter_applied(
 # v1.7-B: chain-claude-code-suitability-degraded log shape
 #
 # Motivation (plan.md §11.B.4 #2):
-#   v1.6.2 documented in docs/troubleshooting.md §4-1 that putting
+#   v1.6.2 documented in docs/guides/troubleshooting.md §4-1 that putting
 #   Llama-3.3-70B at the head of a Claude-Code-facing chain causes
 #   over-eager tool invocation (small talk like ``こんにちは`` getting
 #   rewritten to ``Skill(hello)`` calls). Docs alone require the operator
@@ -805,7 +805,7 @@ def log_output_filter_applied(
 _DEFAULT_CLAUDE_CODE_SUITABILITY_HINT: str = (
     "move the degraded provider(s) to the tail of the chain or replace "
     "with an agentic-coding-tuned model (e.g. qwen3-coder-480b-a35b-instruct); "
-    "see docs/troubleshooting.md §4-1"
+    "see docs/guides/troubleshooting.md §4-1"
 )
 
 

{coderouter_cli-2.5.0 → coderouter_cli-2.5.1}/coderouter/routing/capability.py

@@ -312,7 +312,7 @@ def anthropic_request_has_cache_control(request: AnthropicRequest) -> bool:
 # v1.7-B: claude_code_suitability startup check
 #
 # Motivation (plan.md §11.B.4 #2):
-#   v1.6.2 documented in docs/troubleshooting.md §4-1 the "Llama-3.3-70B
+#   v1.6.2 documented in docs/guides/troubleshooting.md §4-1 the "Llama-3.3-70B
 #   over-eagerly invokes Skill() for small talk under Claude Code"
 #   symptom. v1.7-B promotes that hint from prose-only to a structured
 #   automatic startup WARN: at app startup we scan every profile whose

coderouter_cli-2.5.1/docs/README.md

@@ -0,0 +1,94 @@
+# CodeRouter ドキュメント / Documentation
+
+CodeRouter の公開ドキュメント索引です。「やりたいこと」から読むべきページを引けるよう整理しています。
+Index of CodeRouter's public documentation — find the right page by what you want to do.
+
+> 開発者向けの内部メモ・記事原稿は `inside/` と `articles/` にあります(ローカル専用 / gitignored、公開リポジトリには含まれません)。
+> Developer-internal notes and article drafts live in `inside/` and `articles/` (local-only, not shipped in the public repo).
+
+---
+
+## こういう時はこれを読む / Quick start by goal
+
+| やりたいこと / Goal | 読むもの / Read |
+|---|---|
+| 今すぐ動かしたい / Get running now | [start/quickstart](start/quickstart.md) |
+| 自分に必要か知りたい / Is it for me? | [start/when-do-i-need-coderouter](start/when-do-i-need-coderouter.md) |
+| 無料で運用したい / Run for free | [guides/free-tier-guide](guides/free-tier-guide.md) |
+| 機能を一通り知りたい / Learn the features | [guides/usage-guide](guides/usage-guide.md) |
+| エラーで詰まった / Something broke | [guides/troubleshooting](guides/troubleshooting.md) |
+| ローカル LLM を起動したい / Launch a local LLM | [backends/launcher-quickstart](backends/launcher-quickstart.md) |
+| APIキー・機密の扱い / Secrets & security | [guides/security](guides/security.md) |
+| 仕組みを理解したい / Understand the design | [concepts/architecture](concepts/architecture.md) |
+
+---
+
+## 構成 / Layout
+
+```
+docs/
+├── start/             はじめに / Getting started
+├── guides/            使い方ガイド / How-to guides
+├── backends/          ローカルLLMバックエンド / Local LLM backends
+├── concepts/          設計・内部動作 / Architecture & internals
+├── designs/           設計ドキュメント / Design docs
+├── retrospectives/    リリース振り返り / Release retrospectives
+├── evidence/          検証ログ / Verification logs
+├── openrouter-roster/ OpenRouter モデル一覧 / OpenRouter model roster
+└── assets/            画像など / Images
+```
+
+各ドキュメントは日本語版 (`.md`) と英語版 (`.en.md`) が揃っているものがあります。
+Many documents have a Japanese version (`.md`) and an English version (`.en.md`).
+
+---
+
+## 1. はじめに / Getting started — `start/`
+
+初めて CodeRouter に触れる人向け。 / For first-time users.
+
+- **quickstart** — 最短セットアップで動かす / Get running in one sitting · [日本語](start/quickstart.md) · [English](start/quickstart.en.md)
+- **when-do-i-need-coderouter** — 自分に必要かを判断する / Decide whether you need it · [日本語](start/when-do-i-need-coderouter.md) · [English](start/when-do-i-need-coderouter.en.md)
+
+## 2. 使い方ガイド / How-to guides — `guides/`
+
+日常的に使いこなすためのガイド。 / Day-to-day usage.
+
+- **usage-guide** — 機能を一通り使いこなす / Full feature guide · [日本語](guides/usage-guide.md) · [English](guides/usage-guide.en.md)
+- **free-tier-guide** — NVIDIA NIM × OpenRouter Free でコストゼロ運用 / Zero-cost operation · [日本語](guides/free-tier-guide.md) · [English](guides/free-tier-guide.en.md)
+- **troubleshooting** — つまずいたときの解決集 / Fixing problems · [日本語](guides/troubleshooting.md) · [English](guides/troubleshooting.en.md)
+- **security** — シークレット管理とセキュリティ方針 / Secrets handling & security posture · [日本語](guides/security.md) · [English](guides/security.en.md)
+
+## 3. ローカル LLM バックエンド / Local LLM backends — `backends/`
+
+ローカル推論バックエンドの導入・起動・接続。 / Installing, launching, and connecting local inference backends.
+
+- **launcher-quickstart** — バックエンド導入から起動までの最短手順 / Install a backend and launch · [日本語](backends/launcher-quickstart.md)
+- **launcher** — ブラウザから使う Web 版ランチャー / Web launcher · [日本語](backends/launcher.md)
+- **launcher-gui** — デスクトップ GUI ランチャー / Desktop GUI launcher · [日本語](backends/launcher-gui.md)
+- **llamacpp-direct** — llama.cpp に直結する / Connect llama.cpp directly · [日本語](backends/llamacpp-direct.md) · [English](backends/llamacpp-direct.en.md)
+- **lmstudio-direct** — LM Studio に直結する / Connect LM Studio directly · [日本語](backends/lmstudio-direct.md) · [English](backends/lmstudio-direct.en.md)
+- **hf-ollama-models** — HuggingFace 配布モデルを Ollama で使う / Use HF models via Ollama · [日本語](backends/hf-ollama-models.md)
+- **gguf_dl** — GGUF モデルのダウンロードツール / GGUF download helper · [日本語](backends/gguf_dl.md)
+- **verify-ollama-0.23.1** — Ollama v0.23.1 実機検証チェックリスト / Ollama verification checklist · [日本語](backends/verify-ollama-0.23.1.md)
+
+## 4. 設計・内部動作 / Architecture & internals — `concepts/`
+
+CodeRouter の仕組みと信頼性機構。 / How CodeRouter works and its reliability mechanisms.
+
+- **architecture** — アーキテクチャ全体像 / Architecture overview · [日本語](concepts/architecture.md)
+- **context-budget** — コンテキスト予算管理 (v2.0.0) / Context budget management · [日本語](concepts/context-budget.md)
+- **drift-detection** — ドリフト検出 (v2.0-G) / Drift detection · [日本語](concepts/drift-detection.md)
+- **partial-stitch** — ストリーム途中の部分ステッチ (v2.0-H) / Mid-stream partial stitching · [日本語](concepts/partial-stitch.md)
+- **continuous-probing** — 継続プロービング (v2.0-I) / Continuous probing · [日本語](concepts/continuous-probing.md)
+
+## 5. 設計資料・記録 / Design docs & records
+
+- **designs/** — 機能の設計ドキュメント / Feature design docs ([v1.6 auto-router](designs/v1.6-auto-router.md) ほか)
+- **retrospectives/** — リリース振り返り / Release retrospectives ([v0.4](retrospectives/v0.4.md) 〜 [v1.0](retrospectives/v1.0.md))
+- **evidence/** — 実機検証ログ / Verification run logs
+- **openrouter-roster/** — OpenRouter 利用可能モデル一覧 / OpenRouter model roster — [README](openrouter-roster/README.md)
+
+---
+
+最終更新 / Last updated: 2026-05-22

{coderouter_cli-2.5.0/docs → coderouter_cli-2.5.1/docs/backends}/hf-ollama-models.md

@@ -256,5 +256,5 @@ ollama cp hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q4_K_M qwen3-coder:30b
 - bartowski (品質重視の量子化版): <https://huggingface.co/bartowski>
 - Qwen3-Coder (Alibaba 公式): <https://huggingface.co/collections/Qwen/qwen3-coder>
 - Gemma 4 (Google 公式): <https://huggingface.co/collections/google/gemma-4>
-- CodeRouter doctor 詳細: [`docs/troubleshooting.md`](./troubleshooting.md)
-- providers.yaml 全体構造: [`examples/providers.yaml`](../examples/providers.yaml)
+- CodeRouter doctor 詳細: [`docs/troubleshooting.md`](../guides/troubleshooting.md)
+- providers.yaml 全体構造: [`examples/providers.yaml`](../../examples/providers.yaml)

{coderouter_cli-2.5.0/docs → coderouter_cli-2.5.1/docs/backends}/launcher-gui.md

@@ -197,4 +197,4 @@ launcher:
 - [Launcher クイックスタート](./launcher-quickstart.md) — llama.cpp / vLLM 導入からランチャー起動までの通し手順
 - [Launcher ガイド（Web版）](./launcher.md) — `/launcher` ページ版
 - [llama.cpp 直接接続ガイド](./llamacpp-direct.md)
-- [利用ガイド](./usage-guide.md)
+- [利用ガイド](../guides/usage-guide.md)

{coderouter_cli-2.5.0/docs → coderouter_cli-2.5.1/docs/backends}/launcher-quickstart.md

@@ -183,5 +183,5 @@ ANTHROPIC_BASE_URL=http://localhost:8088 ANTHROPIC_AUTH_TOKEN=dummy claude
 
 - [Launcher ガイド（デスクトップGUI版）](./launcher-gui.md)
 - [Launcher ガイド（Web版）](./launcher.md)
-- [CodeRouter クイックスタート](./quickstart.md)
+- [CodeRouter クイックスタート](../start/quickstart.md)
 - [llama.cpp 直接接続ガイド](./llamacpp-direct.md)

{coderouter_cli-2.5.0/docs → coderouter_cli-2.5.1/docs/backends}/launcher.md

@@ -283,6 +283,6 @@ CodeRouter を再起動すると UI に反映されます。
 
 - [Launcher クイックスタート](./launcher-quickstart.md) — llama.cpp / vLLM 導入からランチャー起動までの通し手順
 - [Launcher ガイド（デスクトップGUI版）](./launcher-gui.md) — ブラウザ不要の tkinter アプリ版
-- [アーキテクチャ詳細 — Launcher セクション](./architecture.md#launcher--llamacpp--vllm-プロセス管理-v250)
-- [利用ガイド](./usage-guide.md)
+- [アーキテクチャ詳細 — Launcher セクション](../concepts/architecture.md#launcher--llamacpp--vllm-プロセス管理-v250)
+- [利用ガイド](../guides/usage-guide.md)
 - [llama.cpp 直接接続ガイド](./llamacpp-direct.md)

{coderouter_cli-2.5.0/docs → coderouter_cli-2.5.1/docs/backends}/llamacpp-direct.en.md

@@ -213,10 +213,10 @@ In the CodeRouter logs you should see `dropped=["reasoning", "reasoning_content"
 
 ## Related docs
 
-- [`troubleshooting.md` §4-2-A](./troubleshooting.md#4-2-a-qwen36) — known Qwen3.6 + Ollama pitfalls.
+- [`troubleshooting.md` §4-2-A](../guides/troubleshooting.md#4-2-a-qwen36) — known Qwen3.6 + Ollama pitfalls.
 - [`hf-ollama-models.md`](./hf-ollama-models.md) — pulling HF models through Ollama.
-- [`examples/providers.yaml`](../examples/providers.yaml) — bundled `llamacpp-qwen3-6-35b-a3b` provider example.
-- The session that produced this guide: [v1.8.1 note](./articles/v1-saga/note-1-v1-8-1-reality-check.md) → [v1.8.2 note](./articles/v1-saga/note-2-v1-8-2-doctor-self-diagnosis.md) → [v1.8.3 note](./articles/v1-saga/note-3-v1-8-3-llamacpp-rescue.md).
+- [`examples/providers.yaml`](../../examples/providers.yaml) — bundled `llamacpp-qwen3-6-35b-a3b` provider example.
+- The session that produced this guide: [v1.8.1 note](../articles/v1-saga/note-1-v1-8-1-reality-check.md) → [v1.8.2 note](../articles/v1-saga/note-2-v1-8-2-doctor-self-diagnosis.md) → [v1.8.3 note](../articles/v1-saga/note-3-v1-8-3-llamacpp-rescue.md).
 
 ---
 

{coderouter_cli-2.5.0/docs → coderouter_cli-2.5.1/docs/backends}/llamacpp-direct.md

@@ -61,7 +61,7 @@ ls -lah ~/models/qwen3.6-35b-a3b-unsloth/
 # → Qwen3.6-35B-A3B-UD-Q4_K_M.gguf (~22GB) と tokenizer 関連ファイル
 ```
 
-> **旧 `huggingface-cli` からの移行**: コマンド名が `hf` に変わり、サブコマンド体系も更新されました (`huggingface-cli download` → `hf download`、`huggingface-cli login` → `hf auth login`)。本リポジトリの補助スクリプト [`gguf_dl.py`](../gguf_dl.py) を使えば、HF の URL を貼るだけで `hf download` 相当の処理を実行できます。
+> **旧 `huggingface-cli` からの移行**: コマンド名が `hf` に変わり、サブコマンド体系も更新されました (`huggingface-cli download` → `hf download`、`huggingface-cli login` → `hf auth login`)。本リポジトリの補助スクリプト [`gguf_dl.py`](../../gguf_dl.py) を使えば、HF の URL を貼るだけで `hf download` 相当の処理を実行できます。
 
 > **`UD-Q4_K_M` の意味**: Unsloth Dynamic Quantization。同 GGUF サイズで通常の Q4_K_M より精度が高い variant。コミュニティ報告でも de facto 標準として使われています。
 >
@@ -270,10 +270,10 @@ provider-ok provider=llamacpp-qwen3-6-35b-a3b
 
 ## 関連ドキュメント
 
-- [troubleshooting.md §4-2-A](./troubleshooting.md#4-2-a-qwen36) — Qwen3.6 + Ollama の既知問題
+- [troubleshooting.md §4-2-A](../guides/troubleshooting.md#4-2-a-qwen36) — Qwen3.6 + Ollama の既知問題
 - [hf-ollama-models.md](./hf-ollama-models.md) — Ollama 経由 HF モデル取得
-- [examples/providers.yaml](../examples/providers.yaml) — `llamacpp-qwen3-6-35b-a3b` provider 例同梱
-- 実機検証の経緯: [v1.8.1 note](./articles/v1-saga/note-1-v1-8-1-reality-check.md) → [v1.8.2 note](./articles/v1-saga/note-2-v1-8-2-doctor-self-diagnosis.md) → [v1.8.3 note](./articles/v1-saga/note-3-v1-8-3-llamacpp-rescue.md)
+- [examples/providers.yaml](../../examples/providers.yaml) — `llamacpp-qwen3-6-35b-a3b` provider 例同梱
+- 実機検証の経緯: [v1.8.1 note](../articles/v1-saga/note-1-v1-8-1-reality-check.md) → [v1.8.2 note](../articles/v1-saga/note-2-v1-8-2-doctor-self-diagnosis.md) → [v1.8.3 note](../articles/v1-saga/note-3-v1-8-3-llamacpp-rescue.md)
 
 ---
 

{coderouter_cli-2.5.0/docs → coderouter_cli-2.5.1/docs/backends}/lmstudio-direct.en.md

@@ -381,10 +381,10 @@ No `capability-degraded` log lines (Anthropic-native shape has nothing to strip)
 ## Related docs
 
 - [llamacpp-direct.en.md](./llamacpp-direct.en.md) — llama.cpp direct path (CLI build + Unsloth GGUF), Qwen3.6 35B-A3B verified.
-- [troubleshooting.md §4-2](./troubleshooting.md) — known issues across Ollama / llama.cpp / LM Studio.
+- [troubleshooting.md §4-2](../guides/troubleshooting.md) — known issues across Ollama / llama.cpp / LM Studio.
 - [hf-ollama-models.md](./hf-ollama-models.md) — pulling HF models through Ollama.
-- [examples/providers.yaml](../examples/providers.yaml) — bundled `lmstudio-*` provider stanzas + `test-lmstudio-openai` / `test-lmstudio-anthropic` profiles.
-- Real-machine narrative: [v1.8.4 note (JP)](./articles/v1-saga/note-4-v1-8-4-lmstudio-revival.md) — the day a "framework wait" verdict expired in 24 hours.
+- [examples/providers.yaml](../../examples/providers.yaml) — bundled `lmstudio-*` provider stanzas + `test-lmstudio-openai` / `test-lmstudio-anthropic` profiles.
+- Real-machine narrative: [v1.8.4 note (JP)](../articles/v1-saga/note-4-v1-8-4-lmstudio-revival.md) — the day a "framework wait" verdict expired in 24 hours.
 
 ---