coderouter-cli 2.9.3__tar.gz → 2.10.0__tar.gz

{coderouter_cli-2.9.3 → coderouter_cli-2.10.0}/.gitignore

@@ -7,6 +7,7 @@
 !.env.example
 secrets.env
 ~/.coderouter/
+CLAUDE.md
 
 # ============================================================
 # Python

{coderouter_cli-2.9.3 → coderouter_cli-2.10.0}/CHANGELOG.md

@@ -9,6 +9,193 @@ are kept verbatim where the Japanese text itself is the subject).
 
 ---
 
+## [v2.10.0] — 2026-07-20 (VSCode workspace scaffolder)
+
+### Added
+
+- **`coderouter vscode-init` — one-shot VSCode workspace scaffolder.**
+  Writes `.vscode/settings.json` with `terminal.integrated.env.osx` /
+  `.linux` / `.windows` populated so a Claude Code session launched
+  from VSCode's integrated terminal auto-points at CodeRouter with no
+  manual `ANTHROPIC_BASE_URL` juggling. Optionally emits a direnv
+  `.envrc` (`--with-envrc`). The command is **idempotent** (re-running
+  with the same arguments is a no-op) and **conflict-aware**: if an
+  existing `settings.json` has different values for one of our managed
+  keys, the file is left untouched and a unified diff is printed —
+  operators re-run with `--force` to overwrite. Unrelated top-level
+  keys (`editor.fontSize`, `python.testing.pytestEnabled`, etc.) and
+  unrelated env keys inside the terminal blocks (a user's `PATH`
+  tweak) are preserved verbatim on merge. All writes are atomic
+  (temp file + `os.replace`) so a partial write cannot corrupt an
+  existing `settings.json`. Non-Claude-Code extensions (Cline / Roo
+  Code / Kilo Code / Continue.dev) are handled by a cheat sheet
+  printed at end of run plus the fuller `docs/guides/vscode.md` —
+  `vscode-init` deliberately does NOT reach into those extensions'
+  own config schemas (each ships its own release cadence, so any
+  automation would need continuous maintenance to follow). The
+  Continue.dev cheat-sheet snippet is built via `json.dumps` rather
+  than hand-formatted concatenation, so it is guaranteed to be
+  syntactically valid JSON (an early hand-formatted version leaked
+  an extra `}` from mismatched f-string / plain-string brace
+  escaping; a test round-trips the snippet through `json.loads` to
+  pin the invariant). Options: `--target PATH` (default cwd),
+  `--port PORT` (default **8088**, matching every docs / quickstart
+  example — note that `coderouter serve` alone still defaults to
+  4000, so if you serve on 4000 pass `--port 4000` here too),
+  `--profile NAME` (adds `CODEROUTER_MODE=<profile>` to the terminal
+  env for header-less profile routing), `--with-envrc` (also emit
+  `.envrc`), `--dry-run` (compute diffs but write nothing —
+  byte-identical to the write path minus the final `os.replace`),
+  and `--force` (overwrite conflicting values). Exit codes: 0 clean,
+  2 conflicts (nothing was written for the conflicting file), 1
+  hard error (unparseable JSON, missing target directory, write
+  failure). Implementation lives in `coderouter/vscode_init.py`
+  (stdlib only — no new runtime deps, keeping the strict-5-package
+  rule in `pyproject.toml` intact); the CLI wrapper
+  (`_run_vscode_init` in `coderouter/cli.py`) mirrors the existing
+  `stats` / `audit` / `replay` "thin argparse plumbing + logic in a
+  sibling module" pattern. See `tests/test_vscode_init.py` (33 cases
+  across happy paths, merge preservation, conflict detection,
+  `--force`, `--dry-run` byte parity, `.envrc`, malformed inputs,
+  JSON validity of the Continue.dev snippet, and CLI wrapper
+  plumbing) and the new `docs/guides/vscode.md` for the full Cline /
+  Roo / Continue.dev cheat sheet.
+
+### Documentation
+
+- **Launcher port ↔ `providers.yaml` mismatch, three ways to prevent
+  it.** New section
+  ["providers.yaml とのポート整合"](docs/backends/launcher.md#providersyaml-とのポート整合食い違いを構造的に防ぐ)
+  in the Launcher guide walks through the three coexisting patterns
+  — (A) hardcoded `providers.yaml` with matching Launcher port
+  (traditional), (B) v2.7.4 Launcher auto-sync
+  (`launcher-<backend>-<port>` provider names, no `providers.yaml`
+  edit needed), and (C) the recommended hybrid where
+  `option_profiles` pins the port so the Launcher UI and
+  `providers.yaml` cannot drift. Includes a copy-paste `cr-check`
+  shell function that runs `coderouter doctor --check-model <name>`
+  against every hand-declared provider before work begins so
+  operators catch a port mismatch in seconds rather than after a
+  session of quiet fallback churn. The symptom side (dashboard's
+  `unhealthy` badge + half-and-half `provider-failed` events in the
+  ring buffer while requests still succeed via fallback) is
+  documented as a new
+  [troubleshooting §1-7](docs/guides/troubleshooting.md#1-7-transport-error-all-connections-failed--ポート不一致の疑い)
+  entry, cross-linked back to the Launcher guide for the fix. Both
+  additions are Japanese-side only (English counterpart will follow
+  in a later docs-sync commit).
+
+### Fixed
+
+- **`ProviderConfig.timeout_s` upper bound raised from 600s (10 min) to
+  86400s (24h)** (#76, contributed by @firelzrd). The previous 10-minute
+  ceiling was too restrictive for long-running local model inference —
+  configurations that legitimately need multi-hour timeouts (heavy
+  reasoning models, large-context sweeps, batch operations) were being
+  rejected by the Pydantic `le=600` validator before the request could
+  even leave the router. The default (30s) and lower bound (1s) are
+  unchanged, so existing configs are unaffected; only users who were
+  hitting the ceiling gain the new headroom.
+
+---
+
+## [v2.9.4] — 2026-07-19 (launcher device selection & bench sweep, cache-stable system prompt)
+
+### Added
+
+- **llama.cpp device selection (`--device` / `--tensor-split`) in the
+  Launcher.** Both the desktop GUI and the Web edition can now detect
+  available `llama-server` devices (via `--list-devices`), display their
+  VRAM, and let the operator pick which one(s) to offload to instead of
+  always relying on llama.cpp's own default placement. New shared module
+  `coderouter/launcher_devices.py` (dataclasses + pure functions only, no
+  pydantic, so the standalone GUI can import it without the `coderouter`
+  package) owns detection/caching (`detect_llama_devices`,
+  `parse_list_devices`, a 60s TTL cache keyed by binary path),
+  CLI-fragment building (`DeviceSelection.to_cli_args` — empty selection
+  emits nothing, preserving today's launch command byte-for-byte),
+  VRAM-ratio tensor-split suggestion (`suggest_tensor_split`), and backend
+  grouping (`backend_of` / `group_by_backend` / `selectable_devices`) so a
+  GPU listed twice under different backends (e.g. `CUDA0` and `Vulkan1`
+  for the same physical card) is never double-counted and cross-backend
+  configs are never auto-generated. Devices reporting `0 MiB` (e.g. macOS
+  `BLAS: Accelerate`) are still shown for information but excluded from
+  selection/suggestion. `StartRequest` gained `device_ids` / `tensor_split`
+  (both default empty — existing Web clients are unaffected), and
+  `GET /api/launcher/devices` (`?backend=`, `?refresh=1`) returns the
+  probe plus a per-backend `suggested_tensor_split` and `auto_configs`.
+  `launcher_gui.py`'s LAUNCH form gained a device checklist (🔍 Detect,
+  falling back to a manual comma-separated entry on detection failure)
+  and a tensor-split field that auto-hides when one or fewer devices are
+  detected. See `tests/test_launcher_devices.py`,
+  `tests/test_launcher_devices_detect.py`,
+  `tests/test_launcher_devices_routes.py`,
+  `tests/test_launcher_gui_devices.py`, and the new "Device selection"
+  section in `docs/backends/launcher.md` / `launcher.en.md`.
+- **Bench sweep: automated device-configuration benchmarking against an
+  external `llmbench`.** Both Launcher editions can now drive a list of
+  device configurations (e.g. `CUDA0` alone / `CUDA1` alone / a
+  multi-GPU split) through start → wait for readiness → run an external
+  bench command → stop → advance to the next configuration, then compare
+  the results. `coderouter/launcher_devices.py` supplies the shared
+  `SweepPlan` / `SweepStep` / `SweepState` state machine,
+  `build_auto_sweep_configs` (one config per selectable device plus one
+  per multi-device backend group), `render_bench_command` (template
+  placeholders `{port}` `{config}` `{base_url}` `{results_dir}` `{runs}`,
+  expanded via plain string substitution rather than `str.format` so JSON
+  braces in the command aren't misinterpreted, then `shlex`-split with
+  `posix=False` on Windows to keep backslash paths intact), and
+  `load_latest_results` / `summarize_results` (best-effort, alias-tolerant
+  extraction of `tokens_per_sec` / `ttft_ms` / `latency_ms` from the
+  newest `llmbench` results JSON). New `LauncherBenchConfig`
+  (`command_template` / `runs` / `results_dir` / `readiness_timeout_s`,
+  read from an optional `launcher.bench:` block — omitted entirely by
+  default, so existing `providers.yaml` files are unaffected) supplies
+  the sweep's defaults. The Web edition gained
+  `POST /api/launcher/sweep/start` (token-gated; 409 if a sweep is
+  already running, 400 for empty `configs` or a busy/unavailable port),
+  `GET /api/launcher/sweep/status`, `POST /api/launcher/sweep/abort`
+  (token-gated), and `GET /api/launcher/sweep/logs`, backed by a new
+  `_SweepRunner` that reuses the existing `spawn_process` /
+  `stop_process` / `proc.ready` readiness primitives. `launcher_gui.py`
+  gained a "📊 Bench sweep" button opening a separate `SweepWindow` with
+  its own `_SweepWorker` thread, reusing `poll_until_ready` for readiness
+  and driving the same external bench subprocess model. See
+  `tests/test_launcher_config_bench.py` and the new "Bench sweep" section
+  in `docs/backends/launcher.md` / `launcher.en.md`.
+
+### Fixed
+
+- **Mid-conversation `role:"system"` messages are no longer hoisted into
+  the top-level `system` field.** (#75, contributed by @firelzrd.)
+  Claude Code CLI ≥ 2.1.154 emits volatile system reminders as
+  `role:"system"` messages *inside* `messages[]`; hoisting them meant the
+  serialized prompt's head changed every turn, so llama.cpp / LM Studio
+  style prefix-matching KV caches were invalidated wholesale (a captured
+  real session measured only 11.8 % reusable prefix between consecutive
+  requests). `normalize_message_roles()` now splits by position: leading
+  system messages (before any emitted turn) are hoisted as before — the
+  legitimate OpenAI-style system prompt — while mid-conversation ones are
+  coerced to `user` in place, keeping `messages` append-only and the
+  prompt head stable as the conversation grows. See
+  `tests/test_role_normalization.py`
+  (`test_leading_system_is_hoisted_but_mid_conversation_is_not`,
+  `test_system_prompt_is_stable_as_conversation_grows`).
+
+### Security
+
+- **Dependency bumps for two fresh advisories** (lockfile-only; both
+  verified with a clean `uv sync --frozen` + full test suite, 1905
+  passed): `click` 8.3.2 → 8.4.2 (PYSEC-2026-2132, fix ≥ 8.3.3;
+  transitive via uvicorn) and `json-repair` 0.59.10 → 0.61.5
+  (GHSA-xf7x-x43h-rpqh, CVSS 7.5, fix ≥ 0.60.1). The `json-repair`
+  floor in `pyproject.toml` was raised from `>=0.30` to `>=0.60.1` so
+  fresh installs can never resolve a vulnerable version; `uv.lock` also
+  records the new lockfile revision 3 format (upload-time metadata)
+  emitted by current uv.
+
+---
+
 ## [v2.9.3] — 2026-07-12 (GUI/Web parity, swap polish, launcher UI fixes)
 
 ### Added

{coderouter_cli-2.9.3 → coderouter_cli-2.10.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: coderouter-cli
-Version: 2.9.3
+Version: 2.10.0
 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
@@ -40,7 +40,7 @@ Requires-Dist: types-pyyaml>=6.0.12; extra == 'dev'
 Provides-Extra: doctor
 Requires-Dist: ruamel-yaml>=0.18.6; extra == 'doctor'
 Provides-Extra: repair
-Requires-Dist: json-repair>=0.30; extra == 'repair'
+Requires-Dist: json-repair>=0.60.1; extra == 'repair'
 Description-Content-Type: text/markdown
 
 <h1 align="center">CodeRouter</h1>
@@ -132,6 +132,8 @@ ANTHROPIC_BASE_URL=http://localhost:8088 ANTHROPIC_AUTH_TOKEN=dummy claude
 
 これだけ。Claude Code はいつも通り動きますが、裏ではローカルの Ollama が答えています。
 
+**VSCode の統合ターミナルから使う場合** は、環境変数の書き忘れを避けるため `coderouter vscode-init` が便利です。プロジェクトルートで 1 回叩くと `.vscode/settings.json` に `terminal.integrated.env.*` をマージ書き込みするので、以後 VSCode のターミナルで `claude` と打つだけで通ります。Cline / Roo Code / Continue.dev の設定コピペも含めて → [VSCode 連携ガイド](./docs/guides/vscode.md)
+
 ---
 
 ## 自分に必要？
@@ -281,6 +283,7 @@ providers:
 | 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) |
 | 言語税を計測・回避する | [言語税ガイド](./docs/guides/language-tax.md) |
+| VSCode / Cline / Continue から使う | [VSCode 連携ガイド](./docs/guides/vscode.md) |
 | 別の PC から安全に繋ぐ | [リモートアクセスガイド](./docs/guides/remote-access.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) |

{coderouter_cli-2.9.3 → coderouter_cli-2.10.0}/README.md

@@ -87,6 +87,8 @@ ANTHROPIC_BASE_URL=http://localhost:8088 ANTHROPIC_AUTH_TOKEN=dummy claude
 
 これだけ。Claude Code はいつも通り動きますが、裏ではローカルの Ollama が答えています。
 
+**VSCode の統合ターミナルから使う場合** は、環境変数の書き忘れを避けるため `coderouter vscode-init` が便利です。プロジェクトルートで 1 回叩くと `.vscode/settings.json` に `terminal.integrated.env.*` をマージ書き込みするので、以後 VSCode のターミナルで `claude` と打つだけで通ります。Cline / Roo Code / Continue.dev の設定コピペも含めて → [VSCode 連携ガイド](./docs/guides/vscode.md)
+
 ---
 
 ## 自分に必要？
@@ -236,6 +238,7 @@ providers:
 | 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) |
 | 言語税を計測・回避する | [言語税ガイド](./docs/guides/language-tax.md) |
+| VSCode / Cline / Continue から使う | [VSCode 連携ガイド](./docs/guides/vscode.md) |
 | 別の PC から安全に繋ぐ | [リモートアクセスガイド](./docs/guides/remote-access.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) |

{coderouter_cli-2.9.3 → coderouter_cli-2.10.0}/coderouter/cli.py

@@ -336,6 +336,78 @@ def _build_parser() -> argparse.ArgumentParser:
         ),
     )
 
+    # v2.10-A: `coderouter vscode-init` — scaffold a VSCode workspace so
+    # Claude Code launched from the integrated terminal auto-points at
+    # CodeRouter (no manual env-var juggling). Optionally writes .envrc
+    # for direnv users. Cline / Roo / Continue are covered by the cheat
+    # sheet printed on completion and docs/guides/vscode.md — this
+    # command deliberately does NOT touch those extensions' settings
+    # (their schemas change with their own release cadence).
+    vscode_init = sub.add_parser(
+        "vscode-init",
+        help="Scaffold VSCode workspace settings for CodeRouter (v2.10-A).",
+        description=(
+            "Write .vscode/settings.json (terminal.integrated.env.*) so a "
+            "Claude Code session launched from VSCode's integrated terminal "
+            "auto-points at CodeRouter. Optionally emit a direnv .envrc. "
+            "Idempotent: safe to re-run. Conflict-aware: refuses to overwrite "
+            "differing existing values without --force."
+        ),
+    )
+    vscode_init.add_argument(
+        "--target",
+        default=".",
+        metavar="PATH",
+        help="Workspace root (default: current directory).",
+    )
+    vscode_init.add_argument(
+        "--port",
+        type=int,
+        default=None,
+        metavar="PORT",
+        help=(
+            "CodeRouter port for ANTHROPIC_BASE_URL. Defaults to 8088 "
+            "(matches every docs / quickstart example). Note that "
+            "`coderouter serve` alone defaults to 4000; if you serve on "
+            "4000, pass --port 4000 here too."
+        ),
+    )
+    vscode_init.add_argument(
+        "--profile",
+        default=None,
+        metavar="NAME",
+        help=(
+            "Preload CODEROUTER_MODE=<profile> so the VSCode terminal "
+            "routes to a non-default profile without a per-request header."
+        ),
+    )
+    vscode_init.add_argument(
+        "--with-envrc",
+        action="store_true",
+        help=(
+            "Also write .envrc (direnv). Run `direnv allow` once after "
+            "generation. Tip: keep secrets in a separate .envrc.local."
+        ),
+    )
+    vscode_init.add_argument(
+        "--dry-run",
+        action="store_true",
+        help=(
+            "Compute all changes and print unified diffs, but do not write "
+            "any files. Byte-identical to the write path minus the final "
+            "os.replace."
+        ),
+    )
+    vscode_init.add_argument(
+        "--force",
+        action="store_true",
+        help=(
+            "Overwrite existing conflicting values. Without this flag, a "
+            "conflict is reported with a diff and the file is left "
+            "untouched (exit 2)."
+        ),
+    )
+
     return parser
 
 
@@ -432,6 +504,9 @@ def main(argv: list[str] | None = None) -> int:
     if args.command == "replay":
         return _run_replay(args)
 
+    if args.command == "vscode-init":
+        return _run_vscode_init(args)
+
     print(f"unknown command: {args.command}", file=sys.stderr)
     return 2
 
@@ -765,6 +840,56 @@ def _run_replay(args: argparse.Namespace) -> int:
     return 0
 
 
+def _run_vscode_init(args: argparse.Namespace) -> int:
+    """v2.10-A: drive :func:`coderouter.vscode_init.run_vscode_init`.
+
+    Kept small and testable — the actual scaffolding logic lives in
+    :mod:`coderouter.vscode_init` so tests can exercise it without
+    argparse in the way. This wrapper's job is:
+
+    * resolve the ``--target`` argument to a real directory
+    * map ``--port`` (default None) to the module's ``DEFAULT_PORT``
+    * translate exceptions (missing target directory) to a friendly
+      stderr message + exit 1
+    * print the formatted result and propagate the module's exit code
+    """
+    from coderouter.vscode_init import (
+        DEFAULT_PORT,
+        exit_code_for,
+        format_result,
+        run_vscode_init,
+    )
+
+    target = Path(args.target).expanduser()
+    if not target.is_dir():
+        print(
+            f"vscode-init: target directory does not exist: {target}",
+            file=sys.stderr,
+        )
+        return 1
+
+    port = args.port if args.port is not None else DEFAULT_PORT
+
+    try:
+        result = run_vscode_init(
+            target,
+            port=port,
+            profile=args.profile,
+            with_envrc=args.with_envrc,
+            dry_run=args.dry_run,
+            force=args.force,
+        )
+    except FileNotFoundError as exc:
+        # Raised only when target vanishes between the pre-check above
+        # and the module's own check — unlikely in practice, but map
+        # it to the same shape as the pre-check for consistency.
+        print(f"vscode-init: {exc}", file=sys.stderr)
+        return 1
+
+    print(format_result(result, dry_run=args.dry_run, port=port))
+    return exit_code_for(result)
+
+
 def _run_check_env(arg_value: str) -> int:
     """v1.6.3: filesystem / git security checks for `.env`.
 

{coderouter_cli-2.9.3 → coderouter_cli-2.10.0}/coderouter/config/schemas.py

@@ -394,7 +394,7 @@ class ProviderConfig(BaseModel):
         default=False,
         description="If true, only used when ALLOW_PAID=true (plan.md §2.3).",
     )
-    timeout_s: float = Field(default=30.0, ge=1.0, le=600.0)
+    timeout_s: float = Field(default=30.0, ge=1.0, le=86400.0)
 
     # v2.6 language-tax track: path to a LOCAL ``tokenizer.json`` for this
     # provider's model, used to measure the CJK over-count vs the char/4
@@ -1460,6 +1460,49 @@ class LauncherOptionProfile(BaseModel):
     )
 
 
+class LauncherBenchConfig(BaseModel):
+    """The ``launcher.bench:`` block — defaults for the bench sweep feature.
+
+    設計 §4.1。デバイス構成スイープ(起動→readiness→外部ベンチ→停止→次)の
+    既定値を providers.yaml から供給する。未設定(``LauncherConfig.bench`` が
+    None)ならスイープ UI はハードコード既定を使う(完全後方互換)。すべての
+    フィールドに ``default`` があるので、既存 YAML は無改変で通過する。
+    """
+
+    model_config = ConfigDict(extra="forbid")
+
+    command_template: str = Field(
+        default="llmbench run --model local-openai --runs {runs}",
+        description=(
+            "外部ベンチコマンドのテンプレ。``{port}`` ``{config}`` "
+            "``{base_url}`` ``{results_dir}`` ``{runs}`` を単純置換で展開して "
+            "argv 化する(``str.format`` ではないので JSON 波括弧で誤爆しない)。"
+        ),
+    )
+    runs: int = Field(
+        default=5,
+        ge=1,
+        le=1000,
+        description="1 構成あたりのベンチ実行回数。``{runs}`` に展開される。",
+    )
+    results_dir: str | None = Field(
+        default=None,
+        description=(
+            "llmbench の results/ ディレクトリ。相対はサーバ CWD 基準。"
+            "設定時のみスイープ結果 JSON を読み比較サマリを付与する。"
+        ),
+    )
+    readiness_timeout_s: float = Field(
+        default=300.0,
+        ge=5.0,
+        le=3600.0,
+        description=(
+            "スイープの各構成でサーバが ready になるのを待つ最大秒数。"
+            "大きな GGUF ロードを考慮した既定 5 分。"
+        ),
+    )
+
+
 class SwapModelSpec(BaseModel):
     """One entry in ``launcher.swap.models`` (docs/designs/launcher-model-swap.md).
 
@@ -1863,6 +1906,13 @@ class LauncherConfig(BaseModel):
             "behavior."
         ),
     )
+    bench: LauncherBenchConfig | None = Field(
+        default=None,
+        description=(
+            "デバイス構成ベンチスイープ(設計 §4)の既定値。None(既定)なら "
+            "スイープ UI はハードコード既定を使う=完全後方互換。"
+        ),
+    )
 
     @model_validator(mode="after")
     def _check_auto_restart_backoff_ordered(self) -> LauncherConfig: