@aituber-onair/chat 0.50.0 → 0.52.0

package/README.ja.md

@@ -581,7 +581,7 @@ const claudeService = ChatServiceFactory.createChatService('claude', {
 対応するClaudeモデルでは `reasoning_effort` を指定でき、Anthropic APIの
 `output_config.effort` に変換されます。省略時のClaude API既定値は `high` です。
 
-- Claude Opus 5、Sonnet 5、Opus 4.8、Opus 4.7:
+- Claude Fable 5、Opus 5、Sonnet 5、Opus 4.8、Opus 4.7:
   `low`、`medium`、`high`、`xhigh`、`max`
 - Claude Opus 4.6、Sonnet 4.6:
   `low`、`medium`、`high`、`max`
@@ -591,6 +591,9 @@ effortを下げるとレイテンシーとトークン効率を優先できま
 短さは保証されません。短い回答が必要な場合は、回答長の指示も併用してください。
 複数ターンのツール呼び出しでは provider 固有の thinking / tool-use ブロックを
 保持するため、返された `completion.assistant_message` を会話履歴へ追加してください。
+Claudeのrefusalはtool errorではなく終端completionとして扱います。Anthropicから
+返された場合は、streamingの`processChat` callbackを含めて
+`finish_reason: 'refusal'`と`stop_details`を保持します。
 
 #### Google Gemini
 
@@ -603,9 +606,10 @@ const geminiService = ChatServiceFactory.createChatService('gemini', {
 ```
 
 `gemini-3.1-flash-lite` は既定の Flash-Lite モデルとして維持します。
-最新の安定版・低遅延 Flash-Lite として `gemini-3.5-flash-lite`、
-より高度なエージェント・マルチモーダル用途向けに `gemini-3.6-flash`、
-安定版 Flash として `gemini-3.5-flash` も利用できます。
+最新の安定版・低遅延 Flash-Lite として `gemini-3.5-flash-lite` を利用できます。
+より高度なコーディング・エージェント・マルチモーダル用途向けの安定版として
+`gemini-3.7-flash` を明示的に選択でき、`gemini-3.6-flash` と
+`gemini-3.5-flash` も引き続き利用できます。
 `gemini-3.1-flash-lite-preview`、`gemini-3-pro-preview`、
 `gemini-2.5-flash-lite-preview-06-17` などの preview / deprecated
 モデルは後方互換のため明示的な model string では利用できますが、本番用途では
@@ -614,13 +618,14 @@ const geminiService = ChatServiceFactory.createChatService('gemini', {
 Gemini 3 モデルでは `reasoning_effort` を指定できます。この値は Gemini の
 `thinkingConfig.thinkingLevel` に変換され、`includeThoughts: false` を維持します。
 
-- Gemini 3 Flash / Flash-Lite: `minimal`, `low`, `medium`, `high`。
-  低遅延なチャット向けの既定値は `minimal` です。
-- Gemini 3 Pro: `low`, `medium`, `high`。Pro は `minimal` 非対応のため、
-  既定値は `low` です。
+- Gemini 3 モデルは、低遅延なチャット向けに利用可能な最小の effort を既定値にします。
+- `minimal` 対応の Flash / Flash-Lite は `minimal`, `low`, `medium`, `high`
+  を公開し、既定値は `minimal` です。
+- `minimal` 非対応モデルは `low`, `medium`, `high` を公開し、既定値は `low`
+  です。現在該当するのは Gemini 3.7 Flash と Gemini 3 Pro です。
 
-これにより Gemini 3.6/3.5 Flash の medium thinking 既定値を上書きし、短い
-出力上限を hidden thinking が使い切るリスクを抑えます。Gemini 2.5 は
+これにより各モデルで利用可能な最小の thinking level を使い、短い出力上限を
+hidden thinking が使い切るリスクを抑えます。Gemini 2.5 は
 `thinkingBudget` を使うため、`reasoning_effort` は送信しません。
 
 #### OpenRouter
@@ -645,9 +650,10 @@ const openRouterService = ChatServiceFactory.createChatService('openrouter', {
 - `openrouter/fusion` は複数モデルのパネルとジャッジモデルを実行します。単一モデルの固定単価ではなく、内部で使われた各モデル呼び出しと web search/fetch 利用分の合算で課金されます。
 - `z-ai/glm-5.2` は OpenRouter reasoning も `none` にします。
 - OpenRouterのreasoning無効化には `reasoning.effort: 'none'` を送ります。`exclude: true` はreasoning内容を返すかどうかだけを制御し、reasoning自体は無効化しません。
-- DeepSeek V4 Flashのスナップショットは、再現可能なバージョンを選べるようOpenRouterのデフォルトではなく明示的なテキスト専用選択肢にしています。どちらも応答性重視で`none`がデフォルトです:
+- DeepSeek V4モデルは、再現可能なバージョンを選べるようOpenRouterのデフォルトではなく明示的なテキスト専用選択肢にしています。すべて応答性重視で`none`がデフォルトです:
   - `deepseek/deepseek-v4-flash-0731`: 現行の固定0731スナップショット。`none`, `low`, `high`, `max`に対応。
   - `deepseek/deepseek-v4-flash`: 旧無印0423スナップショット。`none`, `high`, `xhigh`に対応。
+  - `deepseek/deepseek-v4-pro-0813`: 固定Proスナップショット。`none`, `high`, `xhigh`に対応。
 - コーディング特化モデルはデフォルトではなく明示的な選択肢です。`kwaipilot/kat-coder-air-v2.5` と `kwaipilot/kat-coder-pro-v2.5` はテキスト専用です。
 - `moonshotai/kimi-k3` の可用性は上流キャパシティに依存し、混雑時はOpenRouterから429レスポンスが返る場合があります。
 - `x-ai/grok-4.5` には現在のEU制限を含む地域別の可用性制限があります。`~x-ai/grok-latest` がGrok 4.5へ解決される場合も同じ制限を受けます。
@@ -658,18 +664,18 @@ const openRouterService = ChatServiceFactory.createChatService('openrouter', {
   - `~openai/gpt-latest`, `~openai/gpt-mini-latest`
   - `openai/gpt-5.6-sol`, `openai/gpt-5.6-terra`, `openai/gpt-5.6-luna`
   - `openai/gpt-5.5-pro`, `openai/gpt-5.5`
-  - `openai/gpt-5.1-chat`, `openai/gpt-5.1-codex`, `openai/gpt-5-mini`, `openai/gpt-5-nano`
+  - `openai/gpt-5.1-codex`, `openai/gpt-5-mini`, `openai/gpt-5-nano`
   - `openai/gpt-4o`, `openai/gpt-4.1-mini`, `openai/gpt-4.1-nano`
   - `~anthropic/claude-sonnet-latest`, `~anthropic/claude-haiku-latest`
-  - `anthropic/claude-opus-5`
+  - `anthropic/claude-fable-5`, `anthropic/claude-opus-5`
   - `anthropic/claude-opus-4`, `anthropic/claude-sonnet-4`
-  - `anthropic/claude-3.7-sonnet`, `anthropic/claude-3.5-sonnet`, `anthropic/claude-haiku-4.5`
+  - `anthropic/claude-haiku-4.5`
   - `~google/gemini-pro-latest`, `~google/gemini-flash-latest`
-  - `google/gemini-3.6-flash`, `google/gemini-3.5-flash-lite`
+  - `google/gemini-3.7-flash`, `google/gemini-3.6-flash`, `google/gemini-3.5-flash-lite`
   - `google/gemini-2.5-pro`, `google/gemini-2.5-flash`, `google/gemini-2.5-flash-lite-preview-09-2025`
-  - `z-ai/glm-5.2`, `z-ai/glm-4.7-flash`, `z-ai/glm-4.5-air`, `z-ai/glm-4.5-air:free`
-  - `~x-ai/grok-latest`, `x-ai/grok-4.5`
-  - `deepseek/deepseek-v4-flash-0731`, `deepseek/deepseek-v4-flash`
+  - `z-ai/glm-5.2`, `z-ai/glm-4.7-flash`, `z-ai/glm-4.5-air`
+  - `~x-ai/grok-latest`, `x-ai/grok-4.6`, `x-ai/grok-4.5`
+  - `deepseek/deepseek-v4-flash-0731`, `deepseek/deepseek-v4-flash`, `deepseek/deepseek-v4-pro-0813`
   - `~moonshotai/kimi-latest`, `moonshotai/kimi-k3`, `moonshotai/kimi-k2.7-code`, `moonshotai/kimi-k2.5`
   - `kwaipilot/kat-coder-air-v2.5`, `kwaipilot/kat-coder-pro-v2.5`
 
@@ -707,6 +713,8 @@ console.log(result.fetchedAt); // Date.now() のタイムスタンプ
 const zaiService = ChatServiceFactory.createChatService('zai', {
   apiKey: process.env.ZAI_API_KEY,
   model: 'glm-5.2',
+  reasoning_effort: 'none', // package既定値: thinkingを省略して低遅延化
+  responseLength: 'short',
   visionModel: 'glm-4.6V-Flash', // 任意: ビジョン対応モデル
   responseFormat: { type: 'json_object' } // 任意: JSONモード
 });
@@ -716,24 +724,28 @@ const zaiService = ChatServiceFactory.createChatService('zai', {
 - Z.aiはOpenAI互換のChat Completionsを利用します。
 - テキスト対応モデル: `glm-5.2`, `glm-5.1`, `glm-5`, `glm-5-turbo`, `glm-4.7`, `glm-4.7-FlashX`, `glm-4.7-Flash`, `glm-4.6`
 - ビジョン対応モデル: `glm-5v-turbo`, `glm-4.6V`, `glm-4.6V-FlashX`, `glm-4.6V-Flash`
-- `thinking` はデフォルトで無効化しています。
+- GLM-5.2では`reasoning_effort: 'none' | 'minimal' | 'low' | 'medium' | 'high' | 'xhigh' | 'max'`を公開します。package既定値は低遅延チャット向けの`none`で、thinkingを無効化します。
+- Z.ai側の実効tierは3段階です。`none`/`minimal`は`none`、`low`/`medium`/`high`は`high`、`xhigh`/`max`は`max`へpackage側で正規化して送信します。
+- `reasoning_effort`を省略した場合は、従来の明示的な`thinking`設定も利用できます。両方を指定した場合は`reasoning_effort`を優先します。
+- GLM-5.3はZ.aiの一般APIが現在「coming soon」と案内されているため、まだ対応一覧に加えていません。Coding Plan限定の利用経路は、このproviderの設定経路とは異なります。
 
 #### xAI（Grok）
 
 ```typescript
 const xaiService = ChatServiceFactory.createChatService('xai', {
   apiKey: process.env.XAI_API_KEY,
-  model: 'grok-4.5',
-  reasoning_effort: 'low', // 任意（Grok 4.5）: low, medium, high
-  visionModel: 'grok-4.3', // 任意: ビジョン対応の xAI モデルを指定
+  model: 'grok-4.6',
+  reasoning_effort: 'low', // 任意: low, medium, high, xhigh
+  visionModel: 'grok-4.6', // 任意: ビジョン対応の xAI モデルを指定
 });
 ```
 
 注意:
 - xAIはOpenAI互換のChat Completionsを利用します。
-- 対応モデル: `grok-4.5`, `grok-4.3`, `grok-4.20-0309-reasoning`, `grok-4.20-0309-non-reasoning`, `grok-4-1-fast-reasoning`, `grok-4-1-fast-non-reasoning`
-- `reasoning_effort` は対応モデルにのみ送信されます。`grok-4.5` は `low`, `medium`, `high` に対応し、チャット用途向けにデフォルトは `low` です。`grok-4.3` は `none`, `low`, `medium`, `high` に対応し、デフォルトは `none` です。
-- 対応 xAI モデルではビジョンとツール・関数呼び出しを利用できます。Grok 4.5 も React basic サンプルで画像チャットを直接検証できるようにビジョン対応として有効化しています。
+- 対応モデル: `grok-4.6`, `grok-4.5`, `grok-4.3`, `grok-4.20-0309-reasoning`, `grok-4.20-0309-non-reasoning`
+- packageのデフォルトは`reasoning_effort: 'none'`の`grok-4.3`です。廃止済みGrok 4.1 Fastの定数はソース互換のためexportを残しますが、対応一覧には表示しません。
+- `reasoning_effort` は対応モデルにのみ送信されます。`grok-4.6` は `low`, `medium`, `high`, `xhigh`、`grok-4.5` は `low`, `medium`, `high` に対応し、いずれもチャット用途向けにデフォルトは `low` です。`grok-4.3` は `none`, `low`, `medium`, `high` に対応し、デフォルトは `none` です。
+- 対応 xAI モデルではビジョンとツール・関数呼び出しを利用できます。Grok 4.6 も React basic サンプルで画像チャットを直接検証できるようにビジョン対応として有効化しています。
 
 #### Kimi（Moonshot）
 
@@ -809,7 +821,8 @@ await mistralService.processChat(
 注意:
 - Mistralは`https://api.mistral.ai/v1/chat/completions`のChat Completionsを利用します。
 - デフォルトモデルは`mistral-small-latest`です。低コスト、汎用チャット品質、vision対応、adjustable reasoning対応のバランスがよいためサンプル向けの既定値にしています。
-- 対応モデル: `mistral-small-latest`, `ministral-3b-2512`, `ministral-8b-2512`, `ministral-14b-2512`, `mistral-medium-3-5`, `mistral-large-latest`, `mistral-large-2512`, `mistral-small-2603`, `mistral-medium-2508`
+- 対応モデル: `mistral-small-latest`, `ministral-3b-2512`, `ministral-8b-2512`, `ministral-14b-2512`, `mistral-medium-3-5`, `mistral-large-latest`, `mistral-large-2512`, `mistral-small-2603`
+- 廃止済みの`mistral-medium-2508`はソース互換用にのみexportを残します。
 - Ministral 3系は同じChat Completions endpointでtext、vision、streaming、function callingに対応します。
 - `reasoning_effort`は`'none' | 'high'`に対応し、Mistral公式docsに合わせて`mistral-small-latest`と`mistral-medium-3-5`にだけ送信します。それ以外のモデルでは省略します。
 
@@ -837,7 +850,7 @@ const sakanaService = ChatServiceFactory.createChatService('sakana', {
 
 注意:
 - Sakana AI Fuguは`https://api.sakana.ai/v1/chat/completions`のOpenAI互換Chat Completionsを利用します。
-- 対応モデル: `fugu`（デフォルト）, `fugu-ultra`, `fugu-ultra-20260615`
+- 対応モデル: `fugu`（デフォルト）と現行の正確なFugu Ultra ID `fugu-ultra-v1.1`。`fugu-ultra` aliasと旧日付IDは互換用にexportを残します。
 - Sakanaは新しいChat Completions連携では`max_completion_tokens`を推奨していますが、legacyな`max_tokens`も受け付けます。このproviderでは既存OpenAI互換providerの挙動に合わせて`max_tokens`を使います。
 - Sakanaは最高性能にはResponses APIを推奨していますが、このproviderはpackageのOpenAI互換chat経路に合わせてChat Completionsを利用します。
 - ブラウザから直接呼ぶ場合、Sakana側がoriginに対してCORSを許可していないと失敗します。ブラウザ単体ではなく、Node.js、backend/serverless proxy、または`examples/node-basic/sakana-example.js`を利用してください。
@@ -853,8 +866,8 @@ const plamoService = ChatServiceFactory.createChatService('plamo', {
 
 注意:
 - PLaMoは`https://api.platform.preferredai.jp/v1/chat/completions`のOpenAI互換Chat Completionsを利用します。
-- 対応モデル: `plamo-3.0-prime`（デフォルト）, `plamo-2.2-prime`
-- `plamo-2.2-prime` は明示的な互換用途として残していますが、PLaMo公式docsでは2026-09-30に提供終了し、`plamo-3.0-prime`へ統合予定とされています。
+- 対応モデル: `plamo-3.0-prime`（デフォルト）
+- `plamo-2.2-prime`はソース互換用にexportを残しますが、PLaMO公式docsで2026-09-30の提供終了と`plamo-3.0-prime`への移行が案内されているため、対応一覧には表示しません。
 - reasoning対応モデルでは`reasoning_effort`に`none`または`medium`を指定できます。
 - このproviderではvision対応はサポート対象として扱いません。
 - PLaMoは`openai-compatible`にendpointとmodelを直接指定して使うこともできます。
@@ -1006,6 +1019,21 @@ const tools: ToolDefinition[] = [{
 // サービス作成時にツールハンドラーを設定してください
 ```
 
+アプリケーション側でToolを実行する場合は、backend helperを使ってprovider互換の
+assistant messageとTool結果messageを追加できます。OpenAIやClaudeなどの
+provider固有の継続データも保持します。
+
+```typescript
+import { buildToolContinuationMessages } from '@aituber-onair/chat/backend';
+
+const nextMessages = buildToolContinuationMessages({
+  provider: chatService.provider,
+  messages,
+  completion,
+  toolResults,
+});
+```
+
 ### 応答長制御
 
 プリセットの基準トークン値は次の通りです。
@@ -1253,16 +1281,16 @@ vision、JSON mode、reasoning 設定を使うべきかを provider 固有ロジ
 
 - **OpenAI**: GPT-5.6（Sol/Terra/Luna）、GPT-5.5、GPT-5.4 Pro、GPT-5.4、GPT-5.4 Mini、GPT-5.4 Nano、GPT-5.1、GPT-5（Nano/Mini/Standard）、GPT-4.1(miniとnanoを含む), GPT-4, GPT-4o-mini, O3-mini, o1, o1-miniのモデルをサポート
 - **OpenAI-Compatible**: OpenAI互換 endpoint 経由で任意のローカル/セルフホスト model ID を利用できます。vision 対応可否は endpoint ごとに差があるため、原則 `unknown` 扱いです
-- **Gemini**: Gemini 3.6 Flash、Gemini 3.5 Flash、Gemini 3.5 Flash-Lite、Gemini 3.1 Flash-Lite、Gemini 3.1 Pro Preview、Gemini 3 Flash Preview、Gemini 2.5 Pro、Gemini 2.5 Flash、Gemini 2.5 Flash Lite、Gemma 4 31B IT、Gemma 4 26B A4B IT などの推奨モデルをサポート。チャット用途向けに Gemini 3 Flash は minimal thinking、Gemini 3 Pro は low thinking を既定値にします。Gemini 3.1 Flash-Lite Preview、Gemini 3 Pro Preview、Gemini 2.5 Flash Lite Preview などの lifecycle 上 deprecated なモデルは明示指定用に export を残しています
-- **Claude**: Claude Opus 5, Claude Sonnet 5, Claude Opus 4.8, Claude Opus 4.7, Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.6, Claude Sonnet 4.5, Claude Haiku 4.5 に加え、まだ利用可能だが非推奨の Claude 4 Opus, Claude 4 Sonnet, Claude 3 Haiku をサポート。調整可能な `reasoning_effort` は対応モデルに限り `output_config.effort` として送信します
+- **Gemini**: Gemini 3.7 Flash、Gemini 3.6 Flash、Gemini 3.5 Flash、Gemini 3.5 Flash-Lite、Gemini 3.1 Flash-Lite、Gemini 3.1 Pro Preview、Gemini 3 Flash Preview、Gemini 2.5 Pro、Gemini 2.5 Flash、Gemini 2.5 Flash Lite、Gemma 4 31B IT、Gemma 4 26B A4B IT などの推奨モデルをサポート。`minimal` 非対応の Gemini 3.7 Flash と Gemini 3 Pro は low thinking、その他の Gemini 3 Flash は minimal thinking をチャット用途向けの既定値にします。Gemini 3.1 Flash-Lite Preview、Gemini 3 Pro Preview、Gemini 2.5 Flash Lite Preview などの lifecycle 上 deprecated なモデルは明示指定用に export を残しています
+- **Claude**: Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Opus 4.8, Claude Opus 4.7, Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.6, Claude Sonnet 4.5, Claude Haiku 4.5 をサポート。調整可能な`reasoning_effort`は対応モデルに限り`output_config.effort`として送信し、refusal metadataは終端completionとして保持します
 - **OpenRouter**: OpenRouterのキュレーション済みモデル一覧（OpenAI/Claude/Gemini/Z.ai/xAI/Kimi/DeepSeek/Kwaipilot）をサポート。モデルIDはOpenRouter節を参照してください
-- **Z.ai**: GLM-5.2/GLM-5.1/GLM-5/GLM-5-Turbo（テキスト）、GLM-4.7/4.6（テキスト）、GLM-5V-Turbo/GLM-4.6V系（ビジョン）をサポート
-- **xAI**: Grok 4.5 をチャット用途向けに `reasoning_effort: 'low'` デフォルトでサポート。加えて Grok 4.3、Grok 4.20 の Reasoning/Non-Reasoning、Grok 4-1 Fast の Reasoning/Non-Reasoning をサポートし、全モデルでビジョン対応
+- **Z.ai**: GLM-5.2/GLM-5.1/GLM-5/GLM-5-Turbo（テキスト）、GLM-4.7/4.6（テキスト）、GLM-5V-Turbo/GLM-4.6V系（ビジョン）をサポート。GLM-5.2はモデル別`reasoning_effort`を公開し、低遅延チャット向けに`none`を既定値にします
+- **xAI**: Grok 4.6、Grok 4.5、Grok 4.3、Grok 4.20 Reasoning/Non-Reasoningをvision対応でサポート。低遅延のデフォルトは`reasoning_effort: 'none'`のGrok 4.3です
 - **Kimi**: Kimi K3（`kimi-k3`、`low` / `high` / `max` reasoning、デフォルトは `max`）、Kimi K2.7 Code（`kimi-k2.7-code`）、Kimi K2.7 Code HighSpeed（`kimi-k2.7-code-highspeed`）、Kimi K2.6（`kimi-k2.6`、デフォルト）、Kimi K2.5（`kimi-k2.5`、いずれもビジョン対応）をサポート
 - **DeepSeek**: DeepSeek V4 Flash（`deepseek-v4-flash`）と DeepSeek V4 Pro（`deepseek-v4-pro`）をOpenAI互換Chat Completions経由でサポート。低遅延チャット向けにthinkingはデフォルト無効で、モデル別の`reasoning_effort`から有効化できます。legacy alias の`deepseek-chat`と`deepseek-reasoner`はDeepSeek側で非推奨です
 - **Mistral**: Ministral 3系（`ministral-3b-2512`, `ministral-8b-2512`, `ministral-14b-2512`）と現行generalist modelをサポートし、streamingとvisionにも対応。adjustable `reasoning_effort`は対応モデルにだけ送信します
-- **Sakana AI**: Fugu（`fugu`）と Fugu Ultra（`fugu-ultra`, `fugu-ultra-20260615`）をOpenAI互換Chat Completions経由でサポート
-- **PLaMo**: PLaMo 3.0 Prime（`plamo-3.0-prime`, デフォルト）と PLaMo 2.2 Prime（`plamo-2.2-prime`）をOpenAI互換Chat Completions経由でサポート
+- **Sakana AI**: Fugu（`fugu`）と Fugu Ultra（`fugu-ultra-v1.1`）をOpenAI互換Chat Completions経由でサポート
+- **PLaMo**: PLaMo 3.0 Prime（`plamo-3.0-prime`, デフォルト）をOpenAI互換Chat Completions経由でサポートし、廃止予定の2.2定数は互換用にexportを残します
 - **Gemini Nano**: Chromeブラウザ内蔵AI（LanguageModel API）。デバイス上で動作し、APIキー不要。Webページでは対応するデスクトップ端末上のChrome 148以降が必要で、Chromeフラグの設定は不要。非ストリーミング、ビジョン非対応
 
 ## ライセンス

package/README.md

@@ -595,7 +595,7 @@ the package default. Claude Opus 5 has adaptive thinking enabled by default.
 Supported Claude models accept `reasoning_effort`, which maps to Anthropic's
 `output_config.effort`. The Claude API defaults to `high` when it is omitted:
 
-- Claude Opus 5, Sonnet 5, Opus 4.8, and Opus 4.7:
+- Claude Fable 5, Opus 5, Sonnet 5, Opus 4.8, and Opus 4.7:
   `low`, `medium`, `high`, `xhigh`, or `max`.
 - Claude Opus 4.6 and Sonnet 4.6:
   `low`, `medium`, `high`, or `max`.
@@ -607,6 +607,9 @@ output is required.
 For multi-turn tool calls, append the returned
 `completion.assistant_message` to history so its provider-native thinking and
 tool-use blocks are preserved.
+Claude refusals are terminal completions rather than tool errors. The returned
+completion preserves `finish_reason: 'refusal'` and Anthropic `stop_details`
+when supplied, including in streaming `processChat` callbacks.
 
 #### Google Gemini
 
@@ -620,8 +623,9 @@ const geminiService = ChatServiceFactory.createChatService('gemini', {
 
 `gemini-3.1-flash-lite` remains the default Flash-Lite model.
 `gemini-3.5-flash-lite` is available as the latest stable, low-latency
-Flash-Lite option, while `gemini-3.6-flash` is available for stronger agentic
-and multimodal tasks. Deprecated
+Flash-Lite option. `gemini-3.7-flash` is available as an explicit stable option
+for stronger coding, agentic, and multimodal tasks, while `gemini-3.6-flash`
+remains available for compatibility. Deprecated
 preview and shutdown-scheduled models such as `gemini-3.1-flash-lite-preview`,
 `gemini-3-pro-preview`, and `gemini-2.5-flash-lite-preview-06-17` remain usable
 by explicit model string for backward compatibility, but are no longer
@@ -631,13 +635,14 @@ advertised in the standard supported-model list for production use.
 Gemini 3 models accept `reasoning_effort`, which maps to Gemini
 `thinkingConfig.thinkingLevel` while keeping `includeThoughts: false`:
 
-- Gemini 3 Flash / Flash-Lite: `minimal`, `low`, `medium`, or `high`;
-  defaults to `minimal` for low-latency chat.
-- Gemini 3 Pro: `low`, `medium`, or `high`; defaults to `low` because Pro does
-  not support `minimal`.
+- Gemini 3 models default to their lowest supported effort for low-latency chat.
+- Flash / Flash-Lite models that support `minimal` expose `minimal`, `low`,
+  `medium`, and `high`, and default to `minimal`.
+- Models that do not support `minimal` expose `low`, `medium`, and `high`, and
+  default to `low`. Current instances are Gemini 3.7 Flash and Gemini 3 Pro.
 
-This overrides the medium thinking default of Gemini 3.6/3.5 Flash and reduces
-the risk of hidden thinking exhausting short output limits. Gemini 2.5 uses
+This uses the lowest supported thinking level for chat and reduces the risk of
+hidden thinking exhausting short output limits. Gemini 2.5 uses
 `thinkingBudget` instead, so `reasoning_effort` is intentionally not sent for
 those models.
 
@@ -663,9 +668,10 @@ const openRouterService = ChatServiceFactory.createChatService('openrouter', {
 - `openrouter/fusion` runs a multi-model panel plus a judge model; OpenRouter bills the sum of the underlying model calls and any enabled web search/fetch usage, not a single fixed model rate.
 - For `z-ai/glm-5.2`, OpenRouter reasoning also defaults to `none`.
 - OpenRouter reasoning uses `reasoning.effort: 'none'` to disable reasoning. `exclude: true` only controls whether reasoning content is returned; it does not disable reasoning by itself.
-- DeepSeek V4 Flash snapshots are explicit text-only options rather than the OpenRouter default so applications can choose a reproducible version. Both default to `none` for responsive chat:
+- DeepSeek V4 models are explicit text-only options rather than the OpenRouter default so applications can choose a reproducible version. All default to `none` for responsive chat:
   - `deepseek/deepseek-v4-flash-0731`: current fixed 0731 snapshot; supports `none`, `low`, `high`, and `max`.
   - `deepseek/deepseek-v4-flash`: older unversioned 0423 snapshot; supports `none`, `high`, and `xhigh`.
+  - `deepseek/deepseek-v4-pro-0813`: fixed Pro snapshot; supports `none`, `high`, and `xhigh`.
 - Specialized coding models are explicit options rather than defaults. This includes `kwaipilot/kat-coder-air-v2.5` and `kwaipilot/kat-coder-pro-v2.5`, which are text-only.
 - `moonshotai/kimi-k3` availability depends on upstream capacity; OpenRouter may return 429 responses when capacity is constrained.
 - `x-ai/grok-4.5` has region-specific availability, including a current EU limitation. `~x-ai/grok-latest` inherits those limits when it resolves to Grok 4.5.
@@ -676,18 +682,18 @@ const openRouterService = ChatServiceFactory.createChatService('openrouter', {
   - `~openai/gpt-latest`, `~openai/gpt-mini-latest`
   - `openai/gpt-5.6-sol`, `openai/gpt-5.6-terra`, `openai/gpt-5.6-luna`
   - `openai/gpt-5.5-pro`, `openai/gpt-5.5`
-  - `openai/gpt-5.1-chat`, `openai/gpt-5.1-codex`, `openai/gpt-5-mini`, `openai/gpt-5-nano`
+  - `openai/gpt-5.1-codex`, `openai/gpt-5-mini`, `openai/gpt-5-nano`
   - `openai/gpt-4o`, `openai/gpt-4.1-mini`, `openai/gpt-4.1-nano`
   - `~anthropic/claude-sonnet-latest`, `~anthropic/claude-haiku-latest`
-  - `anthropic/claude-opus-5`
+  - `anthropic/claude-fable-5`, `anthropic/claude-opus-5`
   - `anthropic/claude-opus-4`, `anthropic/claude-sonnet-4`
-  - `anthropic/claude-3.7-sonnet`, `anthropic/claude-3.5-sonnet`, `anthropic/claude-haiku-4.5`
+  - `anthropic/claude-haiku-4.5`
   - `~google/gemini-pro-latest`, `~google/gemini-flash-latest`
-  - `google/gemini-3.6-flash`, `google/gemini-3.5-flash-lite`
+  - `google/gemini-3.7-flash`, `google/gemini-3.6-flash`, `google/gemini-3.5-flash-lite`
   - `google/gemini-2.5-pro`, `google/gemini-2.5-flash`, `google/gemini-2.5-flash-lite-preview-09-2025`
-  - `z-ai/glm-5.2`, `z-ai/glm-4.7-flash`, `z-ai/glm-4.5-air`, `z-ai/glm-4.5-air:free`
-  - `~x-ai/grok-latest`, `x-ai/grok-4.5`
-  - `deepseek/deepseek-v4-flash-0731`, `deepseek/deepseek-v4-flash`
+  - `z-ai/glm-5.2`, `z-ai/glm-4.7-flash`, `z-ai/glm-4.5-air`
+  - `~x-ai/grok-latest`, `x-ai/grok-4.6`, `x-ai/grok-4.5`
+  - `deepseek/deepseek-v4-flash-0731`, `deepseek/deepseek-v4-flash`, `deepseek/deepseek-v4-pro-0813`
   - `~moonshotai/kimi-latest`, `moonshotai/kimi-k3`, `moonshotai/kimi-k2.7-code`, `moonshotai/kimi-k2.5`
   - `kwaipilot/kat-coder-air-v2.5`, `kwaipilot/kat-coder-pro-v2.5`
 
@@ -724,6 +730,8 @@ Notes:
 const zaiService = ChatServiceFactory.createChatService('zai', {
   apiKey: process.env.ZAI_API_KEY,
   model: 'glm-5.2',
+  reasoning_effort: 'none', // Package default: skip thinking for low latency
+  responseLength: 'short',
   visionModel: 'glm-4.6V-Flash', // Optional: vision-capable model
   responseFormat: { type: 'json_object' } // Optional JSON mode
 });
@@ -733,24 +741,28 @@ Notes:
 - Z.ai uses OpenAI-compatible Chat Completions.
 - Supported text models: `glm-5.2`, `glm-5.1`, `glm-5`, `glm-5-turbo`, `glm-4.7`, `glm-4.7-FlashX`, `glm-4.7-Flash`, `glm-4.6`
 - Supported vision models: `glm-5v-turbo`, `glm-4.6V`, `glm-4.6V-FlashX`, `glm-4.6V-Flash`
-- `thinking` is disabled by default to match fast response behavior.
+- GLM-5.2 exposes `reasoning_effort: 'none' | 'minimal' | 'low' | 'medium' | 'high' | 'xhigh' | 'max'`. The package defaults to `none` and disables thinking for low-latency chat.
+- Z.ai currently applies only three effective tiers: `none`/`minimal` become `none`, `low`/`medium`/`high` become `high`, and `xhigh`/`max` become `max`. The package normalizes these values before sending them.
+- Explicit legacy `thinking` remains available when `reasoning_effort` is omitted. `reasoning_effort` takes precedence when both are supplied.
+- GLM-5.3 is not advertised yet because Z.ai currently documents its general API as coming soon; Coding Plan-only availability is not the same configuration path as this provider.
 
 #### xAI (Grok)
 
 ```typescript
 const xaiService = ChatServiceFactory.createChatService('xai', {
   apiKey: process.env.XAI_API_KEY,
-  model: 'grok-4.5',
-  reasoning_effort: 'low', // Optional for Grok 4.5: low, medium, high
-  visionModel: 'grok-4.3', // Optional: use a vision-capable xAI model
+  model: 'grok-4.6',
+  reasoning_effort: 'low', // Optional: low, medium, high, xhigh
+  visionModel: 'grok-4.6', // Optional: use a vision-capable xAI model
 });
 ```
 
 Notes:
 - xAI uses OpenAI-compatible Chat Completions.
-- Supported models: `grok-4.5`, `grok-4.3`, `grok-4.20-0309-reasoning`, `grok-4.20-0309-non-reasoning`, `grok-4-1-fast-reasoning`, `grok-4-1-fast-non-reasoning`
-- `reasoning_effort` is sent only for models that support it. `grok-4.5` supports `low`, `medium`, and `high` and defaults to `low` for chat-style responses. `grok-4.3` supports `none`, `low`, `medium`, and `high` and defaults to `none`.
-- Supported xAI models can be used with vision and tool/function calling. Grok 4.5 vision support is enabled so image chat can be validated directly in the React basic sample.
+- Supported models: `grok-4.6`, `grok-4.5`, `grok-4.3`, `grok-4.20-0309-reasoning`, `grok-4.20-0309-non-reasoning`.
+- `grok-4.3` is the package default with `reasoning_effort: 'none'`. Retired Grok 4.1 Fast constants remain exported for source compatibility but are no longer advertised.
+- `reasoning_effort` is sent only for models that support it. `grok-4.6` supports `low`, `medium`, `high`, and `xhigh`; `grok-4.5` supports `low`, `medium`, and `high`. Both default to `low` for chat-style responses. `grok-4.3` supports `none`, `low`, `medium`, and `high` and defaults to `none`.
+- Supported xAI models can be used with vision and tool/function calling. Grok 4.6 vision support is enabled so image chat can be validated directly in the React basic sample.
 
 #### Kimi (Moonshot)
 
@@ -826,7 +838,8 @@ await mistralService.processChat(
 Notes:
 - Mistral uses Chat Completions at `https://api.mistral.ai/v1/chat/completions`.
 - Default model: `mistral-small-latest`, chosen for the sample-friendly balance of low cost, strong general chat quality, vision support, and adjustable reasoning support.
-- Supported models: `mistral-small-latest`, `ministral-3b-2512`, `ministral-8b-2512`, `ministral-14b-2512`, `mistral-medium-3-5`, `mistral-large-latest`, `mistral-large-2512`, `mistral-small-2603`, `mistral-medium-2508`.
+- Supported models: `mistral-small-latest`, `ministral-3b-2512`, `ministral-8b-2512`, `ministral-14b-2512`, `mistral-medium-3-5`, `mistral-large-latest`, `mistral-large-2512`, `mistral-small-2603`.
+- Retired `mistral-medium-2508` remains exported only for source compatibility.
 - Ministral 3 models support text, vision, streaming, and function calling through the same Chat Completions endpoint.
 - `reasoning_effort` is supported as `'none' | 'high'` and is only sent for `mistral-small-latest` and `mistral-medium-3-5`, matching Mistral's adjustable reasoning docs. It is omitted for other models.
 
@@ -854,7 +867,7 @@ const sakanaService = ChatServiceFactory.createChatService('sakana', {
 
 Notes:
 - Sakana AI Fugu uses OpenAI-compatible Chat Completions at `https://api.sakana.ai/v1/chat/completions`.
-- Supported models: `fugu` (default), `fugu-ultra`, and `fugu-ultra-20260615`.
+- Supported models: `fugu` (default) and the exact current Fugu Ultra ID `fugu-ultra-v1.1`. The `fugu-ultra` alias and former dated ID remain exported for compatibility.
 - Sakana recommends `max_completion_tokens` for new Chat Completions integrations, but also accepts legacy `max_tokens`. This provider keeps `max_tokens` to match existing OpenAI-compatible provider behavior.
 - Sakana recommends the Responses API for best performance, but this provider uses Chat Completions because it matches the package's OpenAI-compatible chat path.
 - Direct browser usage may fail with CORS unless Sakana enables CORS for your origin. Use Node.js, a backend/serverless proxy, or `examples/node-basic/sakana-example.js` instead of calling Sakana directly from browser-only apps.
@@ -870,10 +883,8 @@ const plamoService = ChatServiceFactory.createChatService('plamo', {
 
 Notes:
 - PLaMo uses OpenAI-compatible Chat Completions at `https://api.platform.preferredai.jp/v1/chat/completions`.
-- Supported models: `plamo-3.0-prime` (default) and `plamo-2.2-prime`.
-- `plamo-2.2-prime` is kept for explicit compatibility, but PLaMo docs state
-  it is scheduled to be discontinued on 2026-09-30 and consolidated into
-  `plamo-3.0-prime`.
+- Supported model: `plamo-3.0-prime` (default).
+- `plamo-2.2-prime` remains exported for source compatibility, but is no longer advertised because PLaMo documents its 2026-09-30 retirement and replacement by `plamo-3.0-prime`.
 - `reasoning_effort` can be set to `none` or `medium` for reasoning-capable PLaMo models.
 - Vision is not advertised as supported by this provider.
 - PLaMo can also be used through `openai-compatible` by manually providing the full endpoint and model.
@@ -1031,6 +1042,21 @@ const tools: ToolDefinition[] = [{
 // Configure tool handlers when creating the service
 ```
 
+If your application executes Tools outside ChatService, use the backend helper
+to append provider-compatible assistant and Tool-result messages. It preserves
+provider-native continuation data for providers such as OpenAI and Claude.
+
+```typescript
+import { buildToolContinuationMessages } from '@aituber-onair/chat/backend';
+
+const nextMessages = buildToolContinuationMessages({
+  provider: chatService.provider,
+  messages,
+  completion,
+  toolResults,
+});
+```
+
 ### Response Length Control
 
 Base preset token targets are:
@@ -1278,16 +1304,16 @@ Currently, the following AI providers are built-in:
 
 - **OpenAI**: Supports models like GPT-5.6 (Sol/Terra/Luna), GPT-5.5, GPT-5.4 Pro, GPT-5.4, GPT-5.4 Mini, GPT-5.4 Nano, GPT-5.1, GPT-5 (Nano/Mini/Standard), GPT-4.1 (including mini and nano), GPT-4, GPT-4o-mini, O3-mini, o1, o1-mini
 - **OpenAI-Compatible**: Supports arbitrary local/self-hosted model IDs via OpenAI-compatible endpoints. Vision capability is treated as `unknown` unless your app knows the endpoint-specific model catalog.
-- **Gemini**: Supports recommended models like Gemini 3.6 Flash, Gemini 3.5 Flash, Gemini 3.5 Flash-Lite, Gemini 3.1 Flash-Lite, Gemini 3.1 Pro Preview, Gemini 3 Flash Preview, Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash Lite, Gemma 4 31B IT, and Gemma 4 26B A4B IT. Gemini 3 Flash models default to minimal thinking for chat-style responses, while Gemini 3 Pro models default to low. Deprecated lifecycle models such as Gemini 3.1 Flash-Lite Preview, Gemini 3 Pro Preview, and Gemini 2.5 Flash Lite Preview remain exported for explicit use.
-- **Claude**: Supports current Claude API model IDs including Claude Opus 5, Claude Sonnet 5, Claude Opus 4.8, Claude Opus 4.7, Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.6, Claude Sonnet 4.5, Claude Haiku 4.5, plus deprecated-but-still-available Claude 4 Opus, Claude 4 Sonnet, and Claude 3 Haiku. Adjustable `reasoning_effort` is sent as `output_config.effort` only for models that support it
+- **Gemini**: Supports recommended models like Gemini 3.7 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash, Gemini 3.5 Flash-Lite, Gemini 3.1 Flash-Lite, Gemini 3.1 Pro Preview, Gemini 3 Flash Preview, Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 2.5 Flash Lite, Gemma 4 31B IT, and Gemma 4 26B A4B IT. Gemini 3.7 Flash and Gemini 3 Pro default to low thinking because they do not support minimal; other Gemini 3 Flash models default to minimal for chat-style responses. Deprecated lifecycle models such as Gemini 3.1 Flash-Lite Preview, Gemini 3 Pro Preview, and Gemini 2.5 Flash Lite Preview remain exported for explicit use.
+- **Claude**: Supports current Claude API model IDs including Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Opus 4.8, Claude Opus 4.7, Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.6, Claude Sonnet 4.5, and Claude Haiku 4.5. Adjustable `reasoning_effort` is sent as `output_config.effort` only for models that support it; refusal metadata is preserved as a terminal completion.
 - **OpenRouter**: Supports a curated OpenRouter model list (OpenAI/Claude/Gemini/Z.ai/xAI/Kimi/DeepSeek/Kwaipilot). See the OpenRouter section for model IDs.
-- **Z.ai**: Supports GLM-5.2/GLM-5.1/GLM-5/GLM-5-Turbo (text), GLM-4.7/4.6 (text), and GLM-5V-Turbo/GLM-4.6V family (vision)
-- **xAI**: Supports Grok 4.5 with `reasoning_effort: 'low'` by default for chat-style responses, plus Grok 4.3, Grok 4.20 Reasoning/Non-Reasoning, and Grok 4-1 Fast Reasoning/Non-Reasoning, all with vision support.
+- **Z.ai**: Supports GLM-5.2/GLM-5.1/GLM-5/GLM-5-Turbo (text), GLM-4.7/4.6 (text), and GLM-5V-Turbo/GLM-4.6V family (vision). GLM-5.2 exposes model-aware `reasoning_effort` and defaults to `none` for low-latency chat.
+- **xAI**: Supports Grok 4.6, Grok 4.5, Grok 4.3, and Grok 4.20 Reasoning/Non-Reasoning with vision. Grok 4.3 is the low-latency default with `reasoning_effort: 'none'`.
 - **Kimi**: Supports Kimi K3 (`kimi-k3`, `low` / `high` / `max` reasoning with `max` as the default), Kimi K2.7 Code (`kimi-k2.7-code`), Kimi K2.7 Code HighSpeed (`kimi-k2.7-code-highspeed`), Kimi K2.6 (`kimi-k2.6`, default), and Kimi K2.5 (`kimi-k2.5`) with vision support
 - **DeepSeek**: Supports DeepSeek V4 Flash (`deepseek-v4-flash`) and DeepSeek V4 Pro (`deepseek-v4-pro`) via OpenAI-compatible Chat Completions. Thinking defaults to disabled for low-latency chat and can be enabled with model-aware `reasoning_effort`. Legacy aliases `deepseek-chat` and `deepseek-reasoner` are deprecated by DeepSeek.
 - **Mistral**: Supports the Ministral 3 family (`ministral-3b-2512`, `ministral-8b-2512`, `ministral-14b-2512`) and current Mistral generalist models, with streaming and vision support. Adjustable `reasoning_effort` is only sent for supported models.
-- **Sakana AI**: Supports Fugu (`fugu`) and Fugu Ultra (`fugu-ultra`, `fugu-ultra-20260615`) via OpenAI-compatible Chat Completions.
-- **PLaMo**: Supports PLaMo 3.0 Prime (`plamo-3.0-prime`, default) and PLaMo 2.2 Prime (`plamo-2.2-prime`) via OpenAI-compatible Chat Completions.
+- **Sakana AI**: Supports Fugu (`fugu`) and Fugu Ultra (`fugu-ultra-v1.1`) via OpenAI-compatible Chat Completions.
+- **PLaMo**: Supports PLaMo 3.0 Prime (`plamo-3.0-prime`, default) via OpenAI-compatible Chat Completions; the retiring 2.2 constant remains exported for compatibility.
 - **Gemini Nano**: Chrome built-in AI (LanguageModel API). Runs on-device with no API key required. Web pages require Chrome 148+ on a supported desktop device; no Chrome flags are required. Non-streaming, no vision support.
 
 ## License

package/dist/cjs/backend.d.ts

@@ -0,0 +1,20 @@
+import type { Message, ToolChatCompletion, ToolResultBlock } from './types';
+export interface ChatBackendProviderCapabilities {
+    readonly text: true;
+    readonly streaming: boolean;
+    readonly tools: boolean;
+}
+/** Returns the execution capabilities known for a built-in Chat provider. */
+export declare function getChatBackendProviderCapabilities(provider: string): ChatBackendProviderCapabilities | undefined;
+export interface BuildToolContinuationMessagesInput {
+    readonly provider: string;
+    readonly messages: readonly Message[];
+    readonly completion: Pick<ToolChatCompletion, 'assistant_message' | 'blocks'>;
+    readonly toolResults: readonly ToolResultBlock[];
+}
+/**
+ * Builds the provider-compatible history required after a Tool call.
+ * The input arrays and messages are never mutated.
+ */
+export declare function buildToolContinuationMessages(input: BuildToolContinuationMessagesInput): Message[];
+//# sourceMappingURL=backend.d.ts.map

package/dist/cjs/backend.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"backend.d.ts","sourceRoot":"","sources":["../../src/backend.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EACV,OAAO,EACP,kBAAkB,EAClB,eAAe,EAEhB,MAAM,SAAS,CAAC;AAsCjB,MAAM,WAAW,+BAA+B;IAC9C,QAAQ,CAAC,IAAI,EAAE,IAAI,CAAC;IACpB,QAAQ,CAAC,SAAS,EAAE,OAAO,CAAC;IAC5B,QAAQ,CAAC,KAAK,EAAE,OAAO,CAAC;CACzB;AAED,6EAA6E;AAC7E,wBAAgB,kCAAkC,CAChD,QAAQ,EAAE,MAAM,GACf,+BAA+B,GAAG,SAAS,CAO7C;AAED,MAAM,WAAW,kCAAkC;IACjD,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;IAC1B,QAAQ,CAAC,QAAQ,EAAE,SAAS,OAAO,EAAE,CAAC;IACtC,QAAQ,CAAC,UAAU,EAAE,IAAI,CAAC,kBAAkB,EAAE,mBAAmB,GAAG,QAAQ,CAAC,CAAC;IAC9E,QAAQ,CAAC,WAAW,EAAE,SAAS,eAAe,EAAE,CAAC;CAClD;AAED;;;GAGG;AACH,wBAAgB,6BAA6B,CAC3C,KAAK,EAAE,kCAAkC,GACxC,OAAO,EAAE,CAkBX"}

package/dist/cjs/backend.js

@@ -0,0 +1,115 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.getChatBackendProviderCapabilities = getChatBackendProviderCapabilities;
+exports.buildToolContinuationMessages = buildToolContinuationMessages;
+const KNOWN_CHAT_PROVIDERS = new Set([
+    'claude',
+    'claude-agent-sdk',
+    'codex-sdk',
+    'copilot-sdk',
+    'deepseek',
+    'gemini',
+    'gemini-nano',
+    'kimi',
+    'mistral',
+    'openai',
+    'openai-compatible',
+    'openrouter',
+    'plamo',
+    'sakana',
+    'xai',
+    'zai',
+]);
+const TOOL_SUPPORTED_PROVIDERS = new Set([
+    'claude',
+    'deepseek',
+    'gemini',
+    'kimi',
+    'mistral',
+    'openai',
+    'openai-compatible',
+    'openrouter',
+    'plamo',
+    'sakana',
+    'xai',
+    'zai',
+]);
+const NON_STREAMING_PROVIDERS = new Set(['codex-sdk']);
+/** Returns the execution capabilities known for a built-in Chat provider. */
+function getChatBackendProviderCapabilities(provider) {
+    if (!KNOWN_CHAT_PROVIDERS.has(provider))
+        return undefined;
+    return {
+        text: true,
+        streaming: !NON_STREAMING_PROVIDERS.has(provider),
+        tools: TOOL_SUPPORTED_PROVIDERS.has(provider),
+    };
+}
+/**
+ * Builds the provider-compatible history required after a Tool call.
+ * The input arrays and messages are never mutated.
+ */
+function buildToolContinuationMessages(input) {
+    const isClaude = input.provider === 'claude';
+    const messages = input.messages.filter((message) => !isEmptyClaudeAssistantMessage(isClaude, message));
+    const toolUses = input.completion.blocks.filter((block) => block.type === 'tool_use');
+    const assistantMessage = input.completion.assistant_message ??
+        buildAssistantToolCall(isClaude, toolUses);
+    if (!isEmptyClaudeAssistantMessage(isClaude, assistantMessage)) {
+        messages.push({ ...assistantMessage });
+    }
+    messages.push(...buildToolResultMessages(isClaude, input.toolResults));
+    return messages;
+}
+function isEmptyClaudeAssistantMessage(isClaude, message) {
+    if (!isClaude || message.role !== 'assistant') {
+        return false;
+    }
+    const content = message.content;
+    return Array.isArray(content) && content.length === 0;
+}
+function buildAssistantToolCall(isClaude, toolUses) {
+    if (isClaude) {
+        return {
+            role: 'assistant',
+            content: toolUses.map((toolUse) => ({
+                type: 'tool_use',
+                id: toolUse.id,
+                name: toolUse.name,
+                input: toolUse.input ?? {},
+            })),
+        };
+    }
+    return {
+        role: 'assistant',
+        content: [],
+        tool_calls: toolUses.map((toolUse) => ({
+            id: toolUse.id,
+            type: 'function',
+            function: {
+                name: toolUse.name,
+                arguments: JSON.stringify(toolUse.input ?? {}),
+            },
+        })),
+    };
+}
+function buildToolResultMessages(isClaude, toolResults) {
+    if (isClaude) {
+        return toolResults.map((result) => ({
+            role: 'user',
+            content: [
+                {
+                    type: 'tool_result',
+                    tool_use_id: result.tool_use_id,
+                    content: result.content,
+                },
+            ],
+        }));
+    }
+    return toolResults.map((result) => ({
+        role: 'tool',
+        tool_call_id: result.tool_use_id,
+        content: result.content,
+    }));
+}
+//# sourceMappingURL=backend.js.map

package/dist/cjs/backend.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"backend.js","sourceRoot":"","sources":["../../src/backend.ts"],"names":[],"mappings":";;AAkDA,gFASC;AAaD,sEAoBC;AArFD,MAAM,oBAAoB,GAAG,IAAI,GAAG,CAAC;IACnC,QAAQ;IACR,kBAAkB;IAClB,WAAW;IACX,aAAa;IACb,UAAU;IACV,QAAQ;IACR,aAAa;IACb,MAAM;IACN,SAAS;IACT,QAAQ;IACR,mBAAmB;IACnB,YAAY;IACZ,OAAO;IACP,QAAQ;IACR,KAAK;IACL,KAAK;CACN,CAAC,CAAC;AAEH,MAAM,wBAAwB,GAAG,IAAI,GAAG,CAAC;IACvC,QAAQ;IACR,UAAU;IACV,QAAQ;IACR,MAAM;IACN,SAAS;IACT,QAAQ;IACR,mBAAmB;IACnB,YAAY;IACZ,OAAO;IACP,QAAQ;IACR,KAAK;IACL,KAAK;CACN,CAAC,CAAC;AAEH,MAAM,uBAAuB,GAAG,IAAI,GAAG,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC;AAQvD,6EAA6E;AAC7E,SAAgB,kCAAkC,CAChD,QAAgB;IAEhB,IAAI,CAAC,oBAAoB,CAAC,GAAG,CAAC,QAAQ,CAAC;QAAE,OAAO,SAAS,CAAC;IAC1D,OAAO;QACL,IAAI,EAAE,IAAI;QACV,SAAS,EAAE,CAAC,uBAAuB,CAAC,GAAG,CAAC,QAAQ,CAAC;QACjD,KAAK,EAAE,wBAAwB,CAAC,GAAG,CAAC,QAAQ,CAAC;KAC9C,CAAC;AACJ,CAAC;AASD;;;GAGG;AACH,SAAgB,6BAA6B,CAC3C,KAAyC;IAEzC,MAAM,QAAQ,GAAG,KAAK,CAAC,QAAQ,KAAK,QAAQ,CAAC;IAC7C,MAAM,QAAQ,GAAG,KAAK,CAAC,QAAQ,CAAC,MAAM,CACpC,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,6BAA6B,CAAC,QAAQ,EAAE,OAAO,CAAC,CAC/D,CAAC;IACF,MAAM,QAAQ,GAAG,KAAK,CAAC,UAAU,CAAC,MAAM,CAAC,MAAM,CAC7C,CAAC,KAAK,EAAyB,EAAE,CAAC,KAAK,CAAC,IAAI,KAAK,UAAU,CAC5D,CAAC;IACF,MAAM,gBAAgB,GACpB,KAAK,CAAC,UAAU,CAAC,iBAAiB;QAClC,sBAAsB,CAAC,QAAQ,EAAE,QAAQ,CAAC,CAAC;IAE7C,IAAI,CAAC,6BAA6B,CAAC,QAAQ,EAAE,gBAA2B,CAAC,EAAE,CAAC;QAC1E,QAAQ,CAAC,IAAI,CAAC,EAAE,GAAG,gBAAgB,EAAa,CAAC,CAAC;IACpD,CAAC;IAED,QAAQ,CAAC,IAAI,CAAC,GAAG,uBAAuB,CAAC,QAAQ,EAAE,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC;IACvE,OAAO,QAAQ,CAAC;AAClB,CAAC;AAED,SAAS,6BAA6B,CACpC,QAAiB,EACjB,OAAgB;IAEhB,IAAI,CAAC,QAAQ,IAAI,OAAO,CAAC,IAAI,KAAK,WAAW,EAAE,CAAC;QAC9C,OAAO,KAAK,CAAC;IACf,CAAC;IACD,MAAM,OAAO,GAAI,OAAiC,CAAC,OAAO,CAAC;IAC3D,OAAO,KAAK,CAAC,OAAO,CAAC,OAAO,CAAC,IAAI,OAAO,CAAC,MAAM,KAAK,CAAC,CAAC;AACxD,CAAC;AAED,SAAS,sBAAsB,CAC7B,QAAiB,EACjB,QAAiC;IAEjC,IAAI,QAAQ,EAAE,CAAC;QACb,OAAO;YACL,IAAI,EAAE,WAAW;YACjB,OAAO,EAAE,QAAQ,CAAC,GAAG,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;gBAClC,IAAI,EAAE,UAAU;gBAChB,EAAE,EAAE,OAAO,CAAC,EAAE;gBACd,IAAI,EAAE,OAAO,CAAC,IAAI;gBAClB,KAAK,EAAE,OAAO,CAAC,KAAK,IAAI,EAAE;aAC3B,CAAC,CAAC;SACkB,CAAC;IAC1B,CAAC;IAED,OAAO;QACL,IAAI,EAAE,WAAW;QACjB,OAAO,EAAE,EAAE;QACX,UAAU,EAAE,QAAQ,CAAC,GAAG,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;YACrC,EAAE,EAAE,OAAO,CAAC,EAAE;YACd,IAAI,EAAE,UAAU;YAChB,QAAQ,EAAE;gBACR,IAAI,EAAE,OAAO,CAAC,IAAI;gBAClB,SAAS,EAAE,IAAI,CAAC,SAAS,CAAC,OAAO,CAAC,KAAK,IAAI,EAAE,CAAC;aAC/C;SACF,CAAC,CAAC;KACkB,CAAC;AAC1B,CAAC;AAED,SAAS,uBAAuB,CAC9B,QAAiB,EACjB,WAAuC;IAEvC,IAAI,QAAQ,EAAE,CAAC;QACb,OAAO,WAAW,CAAC,GAAG,CACpB,CAAC,MAAM,EAAE,EAAE,CACT,CAAC;YACC,IAAI,EAAE,MAAM;YACZ,OAAO,EAAE;gBACP;oBACE,IAAI,EAAE,aAAa;oBACnB,WAAW,EAAE,MAAM,CAAC,WAAW;oBAC/B,OAAO,EAAE,MAAM,CAAC,OAAO;iBACxB;aACF;SACF,CAAuB,CAC3B,CAAC;IACJ,CAAC;IAED,OAAO,WAAW,CAAC,GAAG,CACpB,CAAC,MAAM,EAAE,EAAE,CACT,CAAC;QACC,IAAI,EAAE,MAAM;QACZ,YAAY,EAAE,MAAM,CAAC,WAAW;QAChC,OAAO,EAAE,MAAM,CAAC,OAAO;KACxB,CAAY,CAChB,CAAC;AACJ,CAAC"}

package/dist/cjs/constants/claude.d.ts

@@ -12,8 +12,10 @@ export declare const MODEL_CLAUDE_4_6_SONNET = "claude-sonnet-4-6";
 export declare const MODEL_CLAUDE_4_6_OPUS = "claude-opus-4-6";
 export declare const MODEL_CLAUDE_4_7_OPUS = "claude-opus-4-7";
 export declare const MODEL_CLAUDE_4_8_OPUS = "claude-opus-4-8";
+export declare const MODEL_CLAUDE_5_FABLE = "claude-fable-5";
 export declare const MODEL_CLAUDE_5_SONNET = "claude-sonnet-5";
 export declare const MODEL_CLAUDE_5_OPUS = "claude-opus-5";
+export declare const CLAUDE_DEPRECATED_MODELS: string[];
 export type ClaudeReasoningEffort = 'low' | 'medium' | 'high' | 'xhigh' | 'max';
 export declare const CLAUDE_VISION_SUPPORTED_MODELS: string[];
 /**

package/dist/cjs/constants/claude.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"claude.d.ts","sourceRoot":"","sources":["../../../src/constants/claude.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,mBAAmB,0CAA0C,CAAC;AAG3E,eAAO,MAAM,oBAAoB,4BAA4B,CAAC;AAC9D,eAAO,MAAM,sBAAsB,8BAA8B,CAAC;AAClE,eAAO,MAAM,uBAAuB,+BAA+B,CAAC;AACpE,eAAO,MAAM,uBAAuB,+BAA+B,CAAC;AACpE,eAAO,MAAM,qBAAqB,6BAA6B,CAAC;AAChE,eAAO,MAAM,mBAAmB,2BAA2B,CAAC;AAC5D,eAAO,MAAM,uBAAuB,+BAA+B,CAAC;AACpE,eAAO,MAAM,sBAAsB,8BAA8B,CAAC;AAClE,eAAO,MAAM,qBAAqB,6BAA6B,CAAC;AAChE,eAAO,MAAM,uBAAuB,sBAAsB,CAAC;AAC3D,eAAO,MAAM,qBAAqB,oBAAoB,CAAC;AACvD,eAAO,MAAM,qBAAqB,oBAAoB,CAAC;AACvD,eAAO,MAAM,qBAAqB,oBAAoB,CAAC;AACvD,eAAO,MAAM,qBAAqB,oBAAoB,CAAC;AACvD,eAAO,MAAM,mBAAmB,kBAAkB,CAAC;AAEnD,MAAM,MAAM,qBAAqB,GAAG,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,OAAO,GAAG,KAAK,CAAC;AAgChF,eAAO,MAAM,8BAA8B,UAa1C,CAAC;AAEF;;GAEG;AACH,wBAAgB,kCAAkC,CAChD,KAAK,EAAE,MAAM,GACZ,SAAS,qBAAqB,EAAE,CAElC;AAED,wBAAgB,4BAA4B,CAAC,KAAK,EAAE,MAAM,GAAG,OAAO,CAEnE;AAED;;GAEG;AACH,wBAAgB,+BAA+B,CAC7C,KAAK,EAAE,MAAM,GACZ,qBAAqB,GAAG,SAAS,CAEnC;AAED;;GAEG;AACH,wBAAgB,8BAA8B,CAC5C,KAAK,EAAE,MAAM,EACb,MAAM,CAAC,EAAE,qBAAqB,GAC7B,qBAAqB,GAAG,SAAS,CAenC"}
+{"version":3,"file":"claude.d.ts","sourceRoot":"","sources":["../../../src/constants/claude.ts"],"names":[],"mappings":"AAAA,eAAO,MAAM,mBAAmB,0CAA0C,CAAC;AAG3E,eAAO,MAAM,oBAAoB,4BAA4B,CAAC;AAC9D,eAAO,MAAM,sBAAsB,8BAA8B,CAAC;AAClE,eAAO,MAAM,uBAAuB,+BAA+B,CAAC;AACpE,eAAO,MAAM,uBAAuB,+BAA+B,CAAC;AACpE,eAAO,MAAM,qBAAqB,6BAA6B,CAAC;AAChE,eAAO,MAAM,mBAAmB,2BAA2B,CAAC;AAC5D,eAAO,MAAM,uBAAuB,+BAA+B,CAAC;AACpE,eAAO,MAAM,sBAAsB,8BAA8B,CAAC;AAClE,eAAO,MAAM,qBAAqB,6BAA6B,CAAC;AAChE,eAAO,MAAM,uBAAuB,sBAAsB,CAAC;AAC3D,eAAO,MAAM,qBAAqB,oBAAoB,CAAC;AACvD,eAAO,MAAM,qBAAqB,oBAAoB,CAAC;AACvD,eAAO,MAAM,qBAAqB,oBAAoB,CAAC;AACvD,eAAO,MAAM,oBAAoB,mBAAmB,CAAC;AACrD,eAAO,MAAM,qBAAqB,oBAAoB,CAAC;AACvD,eAAO,MAAM,mBAAmB,kBAAkB,CAAC;AAGnD,eAAO,MAAM,wBAAwB,UAOpC,CAAC;AAEF,MAAM,MAAM,qBAAqB,GAAG,KAAK,GAAG,QAAQ,GAAG,MAAM,GAAG,OAAO,GAAG,KAAK,CAAC;AAiChF,eAAO,MAAM,8BAA8B,UAW1C,CAAC;AAEF;;GAEG;AACH,wBAAgB,kCAAkC,CAChD,KAAK,EAAE,MAAM,GACZ,SAAS,qBAAqB,EAAE,CAElC;AAED,wBAAgB,4BAA4B,CAAC,KAAK,EAAE,MAAM,GAAG,OAAO,CAEnE;AAED;;GAEG;AACH,wBAAgB,+BAA+B,CAC7C,KAAK,EAAE,MAAM,GACZ,qBAAqB,GAAG,SAAS,CAEnC;AAED;;GAEG;AACH,wBAAgB,8BAA8B,CAC5C,KAAK,EAAE,MAAM,EACb,MAAM,CAAC,EAAE,qBAAqB,GAC7B,qBAAqB,GAAG,SAAS,CAenC"}