PyPI - coderouter-cli - Versions diffs - 2.6.0__py3-none-any.whl → 2.6.1__py3-none-any.whl - Mend

coderouter-cli 2.6.0py3-none-any.whl → 2.6.1py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (7) hide show

coderouter/ingress/dashboard_routes.py CHANGED Viewed

@@ -184,6 +184,21 @@ _DASHBOARD_HTML = r"""<!doctype html>
         </div>
       </div>
       <div id="language-tax-by-provider" class="text-xs text-slate-400 tabnum mt-3"></div>
+      <!-- Token Savings: trim (core budget guard) + compress (plugin) -->
+      <div class="grid grid-cols-3 gap-3 mt-3">
+        <div class="rounded-md bg-slate-800/50 p-3">
+          <div class="text-xs text-slate-400">Tokens saved &middot; trim</div>
+          <div class="text-2xl font-semibold tabnum text-sky-400" data-bind="tokens_saved_trim">0</div>
+        </div>
+        <div class="rounded-md bg-slate-800/50 p-3">
+          <div class="text-xs text-slate-400">Tokens saved &middot; compress</div>
+          <div class="text-2xl font-semibold tabnum text-sky-400" data-bind="tokens_saved_compress">0</div>
+        </div>
+        <div class="rounded-md bg-slate-800/50 p-3">
+          <div class="text-xs text-slate-400">Tokens saved &middot; total</div>
+          <div class="text-2xl font-semibold tabnum text-green-400" data-bind="tokens_saved_total">0</div>
+        </div>
+      </div>
     </section>
     <section class="bg-slate-900/60 border border-slate-800 rounded-lg p-4">
       <h2 class="text-sm font-semibold uppercase tracking-wider text-slate-400 mb-3">Usage Mix</h2>
@@ -474,6 +489,14 @@ _DASHBOARD_HTML = r"""<!doctype html>
       rows.map(([n, v]) =>
         '<span class="mr-4"><span class="text-slate-500">' + escapeHTML(n) +
         '</span> ' + usd(v) + '</span>').join("");
+    // token-savings tiles. trim comes from the core budget guard;
+    // compress from the optional plugin. Collector zero-fills, so an
+    // install with neither still renders "0" cleanly.
+    const ts = c.tokens_saved_by_mechanism || {};
+    const intfmt = (x) => (Number(x) || 0).toLocaleString();
+    setBind("tokens_saved_trim", intfmt(ts.trim));
+    setBind("tokens_saved_compress", intfmt(ts.compress));
+    setBind("tokens_saved_total", intfmt(c.tokens_saved_total));
   };
   const renderSnapshot = (snap) => {

coderouter/metrics/collector.py CHANGED Viewed

@@ -197,6 +197,16 @@ class MetricsCollector(logging.Handler):
         self._language_tax_usd: dict[str, float] = {}
         self._language_tax_usd_aggregate: float = 0.0
+        # token-savings accounting. Tokens that never reached the backend
+        # because a reduction mechanism shrank or dropped them. Two
+        # mechanisms feed this: ``trim`` (core context-budget guard, derived
+        # from the existing ``context-budget-trimmed`` event) and
+        # ``compress`` (the optional compress plugin, via the neutral
+        # ``tokens-saved`` event). Owned by core so the figure exists even
+        # when no plugin is installed.
+        self._tokens_saved_total: int = 0
+        self._tokens_saved_by_mechanism: Counter[str] = Counter()
         # v2.0-F (L1): context budget guard counters. Per-profile counts
         # of warnings (over warn threshold) and trims (messages removed).
         # The ``latest_usage_ratio`` dict records the most recent ratio
@@ -425,6 +435,41 @@ class MetricsCollector(logging.Handler):
                 profile = _str(extras.get("profile"))
                 self._context_budget_trims_total += 1
                 self._context_budget_trims_by_profile[profile] += 1
+                # token-savings: the trim already carries the estimated
+                # before/after token counts, so fold the delta into the
+                # shared savings buckets (mechanism="trim").
+                before = extras.get("estimated_tokens_before")
+                after = extras.get("estimated_tokens_after")
+                if isinstance(before, int | float) and isinstance(
+                    after, int | float
+                ):
+                    saved = max(0, int(before) - int(after))
+                    if saved:
+                        self._tokens_saved_total += saved
+                        self._tokens_saved_by_mechanism["trim"] += saved
+            elif event == "tokens-saved":
+                # Neutral token-savings event emitted by non-core reduction
+                # mechanisms (e.g. the compress plugin). Carries a
+                # ``mechanism`` label plus either a precomputed
+                # ``tokens_saved`` or a ``tokens_before`` / ``tokens_after``
+                # pair. Negative deltas clamp to zero. No core import needed
+                # by the emitter — it just logs this record name.
+                mechanism = _str(extras.get("mechanism")) or "unknown"
+                saved_raw = extras.get("tokens_saved")
+                before = extras.get("tokens_before")
+                after = extras.get("tokens_after")
+                if isinstance(saved_raw, int | float):
+                    saved = max(0, int(saved_raw))
+                elif isinstance(before, int | float) and isinstance(
+                    after, int | float
+                ):
+                    saved = max(0, int(before) - int(after))
+                else:
+                    saved = 0
+                if saved:
+                    self._tokens_saved_total += saved
+                    self._tokens_saved_by_mechanism[mechanism] += saved
+                self._push_recent(event, extras, record)
             elif event == "drift-detected":
                 # v2.0-G (L4): drift detection fired.
                 provider = _str(extras.get("provider"))
@@ -624,6 +669,11 @@ class MetricsCollector(logging.Handler):
                     "language_tax_usd_aggregate": round(
                         self._language_tax_usd_aggregate, 6
                     ),
+                    # token-savings: total + per-mechanism (trim / compress).
+                    "tokens_saved_total": self._tokens_saved_total,
+                    "tokens_saved_by_mechanism": dict(
+                        self._tokens_saved_by_mechanism
+                    ),
                     # v2.0-F (L1): context budget guard aggregate counters.
                     "context_budget_warnings_total": self._context_budget_warnings_total,
                     "context_budget_trims_total": self._context_budget_trims_total,
@@ -684,6 +734,10 @@ class MetricsCollector(logging.Handler):
                 "chain_budget_exceeded_total": self._chain_budget_exceeded_total,
                 "chain_memory_pressure_blocked_total": self._chain_memory_pressure_blocked_total,
                 "chain_uniform_auth_failure_total": self._chain_uniform_auth_failure_total,
+                "tokens_saved_total": self._tokens_saved_total,
+                "tokens_saved_by_mechanism": dict(
+                    self._tokens_saved_by_mechanism
+                ),
                 "probe_rounds_total": self._probe_rounds_total,
             }
@@ -724,6 +778,9 @@ class MetricsCollector(logging.Handler):
             self._language_tax_usd_aggregate += float(
                 state.get("language_tax_usd_aggregate", 0.0)
             )
+            self._tokens_saved_total += int(state.get("tokens_saved_total", 0))
+            for k, v in (state.get("tokens_saved_by_mechanism") or {}).items():
+                self._tokens_saved_by_mechanism[k] += int(v)
             self._chain_paid_gate_blocked_total += int(
                 state.get("chain_paid_gate_blocked_total", 0)
             )
@@ -782,6 +839,9 @@ class MetricsCollector(logging.Handler):
             # v2.6
             self._language_tax_usd.clear()
             self._language_tax_usd_aggregate = 0.0
+            # token-savings
+            self._tokens_saved_total = 0
+            self._tokens_saved_by_mechanism.clear()
             # v2.0-H (L6)
             self._partial_stitch_surfaced_total = 0
             # v2.0-I

{coderouter_cli-2.6.0.dist-info → coderouter_cli-2.6.1.dist-info}/METADATA RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: coderouter-cli
-Version: 2.6.0
+Version: 2.6.1
 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
@@ -161,6 +161,36 @@ ANTHROPIC_BASE_URL=http://localhost:8088 ANTHROPIC_AUTH_TOKEN=dummy claude
 | **`coderouter replay`** | provider 切替の効果を統計比較 (A/B 分析) / `--suggest-rules` でルール最適化提案 |
 | **Continuous Probe** | idle 時も定期的に backend を監視 |
+### 言語税トラッキング — v2.6.0
+日本語などの CJK テキストは、クラウドのトークナイザだと「同じ意味の英語」より多くのトークンを消費します（**実測: GPT-4o 系 o200k で平均 1.6 倍、GPT-4 系 cl100k で平均 2.0 倍**）。ローカル LLM は課金されないので、この「言語税」はクラウド利用時だけ効いてきます。CodeRouter v2.6.0 はこれを **計測・ルーティング回避・可視化** します。
+| 機能 | 何をしてくれるか |
+|---|---|
+| **言語税の計測** | プロバイダに `tokenizer_path`（ローカルの `tokenizer.json`）を指定すると、char/4 ヒューリスティック比の実トークン倍率と割増 USD を算出（ネットワーク不要・未設定なら無効） |
+| **`cjk_ratio_min` ルーティング** | CJK 比率が高いリクエストを自動でローカル LLM（課金ゼロ）へ。コードや英語はクラウドへ |
+| **ダッシュボード可視化** | `/dashboard` の「Cost & Language Tax」パネルで総支出・キャッシュ節約・言語税をリアルタイム表示 |
+```yaml
+# providers.yaml — CJK 多めのターンはローカルへ自動回避
+auto_router:
+  rules:
+    - match: { cjk_ratio_min: 0.3 }   # 日本語が3割以上 → ローカル
+      profile: local
+    - match: { has_tools: true }      # ツール使用 → クラウド
+      profile: cloud
+  default_rule_profile: cloud
+providers:
+  - name: cloud-sonnet
+    kind: anthropic
+    base_url: https://api.anthropic.com
+    model: claude-sonnet-4-6
+    tokenizer_path: ~/.coderouter/tokenizers/sonnet.json   # 言語税の正確計測（任意）
+```
+詳細 → [言語税ガイド](./docs/guides/language-tax.md)
 ### Launcher — llama.cpp / vllm 起動 UI
 `http://localhost:8088/launcher` で開けるブラウザ UI。llama.cpp や vllm を GUI で起動・管理できます。
@@ -229,6 +259,7 @@ providers:
 | 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) |
+| 言語税を計測・回避する | [言語税ガイド](./docs/guides/language-tax.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) |

{coderouter_cli-2.6.0.dist-info → coderouter_cli-2.6.1.dist-info}/RECORD RENAMED Viewed

@@ -40,12 +40,12 @@ coderouter/guards/tool_loop.py,sha256=EzeMcmU7BLeTW2jsRVevU81l5rhWcn1oUr7EpzgXjV
 coderouter/ingress/__init__.py,sha256=WQsCH2CGJCAhy0mS6GSEdeYZRkkQu2OHDsP4CJWTLug,155
 coderouter/ingress/anthropic_routes.py,sha256=It2f7XGe3fgKQX01J2F5JOCoZr96t_Tx_kY2om99MVo,16894
 coderouter/ingress/app.py,sha256=PcuTvUFNjr04EbsUOu8qdyKTdBzxkIJYB4xpz8dFfMo,12635
-coderouter/ingress/dashboard_routes.py,sha256=tEIayMHxCzlmpnLyKHgpqrE4W24DTJM97ewTlYvkKqI,24238
+coderouter/ingress/dashboard_routes.py,sha256=1SqNNrVTfQiZcx5n2MvjAzWirittX0e61tCt-MKp4Ag,25617
 coderouter/ingress/launcher_routes.py,sha256=Jh-E6qFmHnr7ON4W6QanafxQIoojT4F034mybLvhTyQ,47548
 coderouter/ingress/metrics_routes.py,sha256=M22dwOGn24P05Ge4W3c7d7mYytSGWjIR-pPSPOAiHJY,3965
 coderouter/ingress/openai_routes.py,sha256=Zw1efPw9DI6GgV8ZcLrzS6Cda0KLrFkKn2GBZWSe6Vo,6322
 coderouter/metrics/__init__.py,sha256=7Es351DPS7yLM0yVF_F0eesmiD83n7Zzhie44chht38,1465
-coderouter/metrics/collector.py,sha256=9lKnaFpdlu8R9mRUeyAeJWXR1urRCKt_6sUFn_9ybss,49657
+coderouter/metrics/collector.py,sha256=F6C19DgZDqEKq8Hidlo99cfev5vRQO63p4WMOrvUeqQ,52975
 coderouter/metrics/prometheus.py,sha256=YRqyT931s40zVkIj07D-M2UNfDhIEElVFRz3izdJcnQ,24419
 coderouter/plugins/__init__.py,sha256=76hMLe5dV_ilripHXzWn3HSYoIALjzlw4EJVyI-GyIM,1974
 coderouter/plugins/base.py,sha256=n9hsck2NCSqi6oeHIumKC5zhQ8JGwCXUz7J5AZQCQss,5772
@@ -67,8 +67,8 @@ coderouter/translation/__init__.py,sha256=PYXN7XVEwpG1uC8RLy6fvnGbzEZhhrEuUapH8I
 coderouter/translation/anthropic.py,sha256=aZkcYH4x82b0x7efJgJb9RWn9Hbyc9pEOthXe4vjUdU,11113
 coderouter/translation/convert.py,sha256=-qyzFzmmr9hhQV6_Sg75kJnvCZvHe3n7vRdaZtk_JqQ,47269
 coderouter/translation/tool_repair.py,sha256=Ok2PF947Liegc5oaytfptv5MWMkpfJYQie-zdP1y3cY,9946
-coderouter_cli-2.6.0.dist-info/METADATA,sha256=us2o2_EtIlzd2EjQqAqtKIX1ocpAD3YcaDiZKOG6ktE,11674
-coderouter_cli-2.6.0.dist-info/WHEEL,sha256=mffPy8wBnZQn2VnJUU5jE99KsxaSfiyMHV9Yt0aLVxs,87
-coderouter_cli-2.6.0.dist-info/entry_points.txt,sha256=-dnLfD1YZ2WjH2zSdNCvlO65wYltM9bsHt9Fhg3yGss,51
-coderouter_cli-2.6.0.dist-info/licenses/LICENSE,sha256=wkEzoR86jFw33jvfOHjULqmkGEfxTFMgMaJnpR8mPRw,1065
-coderouter_cli-2.6.0.dist-info/RECORD,,
+coderouter_cli-2.6.1.dist-info/METADATA,sha256=3zw3ZZGT71GYD8F9qeYIyMQB-ZN4jmRRoYOB4XD8eIo,13534
+coderouter_cli-2.6.1.dist-info/WHEEL,sha256=mffPy8wBnZQn2VnJUU5jE99KsxaSfiyMHV9Yt0aLVxs,87
+coderouter_cli-2.6.1.dist-info/entry_points.txt,sha256=-dnLfD1YZ2WjH2zSdNCvlO65wYltM9bsHt9Fhg3yGss,51
+coderouter_cli-2.6.1.dist-info/licenses/LICENSE,sha256=wkEzoR86jFw33jvfOHjULqmkGEfxTFMgMaJnpR8mPRw,1065
+coderouter_cli-2.6.1.dist-info/RECORD,,

{coderouter_cli-2.6.0.dist-info → coderouter_cli-2.6.1.dist-info}/WHEEL RENAMED Viewed

File without changes

{coderouter_cli-2.6.0.dist-info → coderouter_cli-2.6.1.dist-info}/entry_points.txt RENAMED Viewed

File without changes

{coderouter_cli-2.6.0.dist-info → coderouter_cli-2.6.1.dist-info}/licenses/LICENSE RENAMED Viewed

File without changes

coderouter-cli 2.6.0__py3-none-any.whl → 2.6.1__py3-none-any.whl

coderouter-cli 2.6.0py3-none-any.whl → 2.6.1py3-none-any.whl