alibaba · Devwallem · Jun 16, 2026 · Jun 16, 2026 · Jun 16, 2026 · Jun 17, 2026
diff --git a/README.ja-JP.md b/README.ja-JP.md
@@ -152,6 +152,31 @@ export OCR_USE_ANTHROPIC=true
 
 また、Claude Codeの環境変数（`ANTHROPIC_BASE_URL`、`ANTHROPIC_AUTH_TOKEN`、`ANTHROPIC_MODEL`）とも互換性があり、`~/.zshrc` / `~/.bashrc`からこれらのexportをパースします。
 
+### Google Vertex AIでClaudeを使用
+
+OCRはGoogle Application Default Credentials（ADC）を使ってVertex AI上のClaudeを呼び出せます。Anthropic APIキーやカスタムLLM URLは不要です：
+
+```bash
+export OCR_USE_VERTEX=true
+export OCR_LLM_MODEL=claude-sonnet-4-6
+export ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project
+export CLOUD_ML_REGION=global
+
+gcloud auth application-default login
+ocr llm test
+```
+
+OCR専用の別名として`OCR_VERTEX_PROJECT_ID`と`OCR_VERTEX_REGION`もサポートしています。プロジェクトIDは`GOOGLE_CLOUD_PROJECT`にもフォールバックします。選択したGCPプロジェクトではVertex AI APIが有効で、対象Claudeモデルへのアクセス権が必要です。
+
+同じ設定は既存の`llm`設定にも保存できます：
+
+```bash
+ocr config set llm.use_anthropic_vertex true
+ocr config set llm.vertex_project_id your-gcp-project
+ocr config set llm.vertex_region global
+ocr config set llm.model claude-sonnet-4-6
+```
+
 > **CC-Switchユーザー向けの注意**: [CC-Switch](https://github.com/farion1231/cc-switch)を[ルーティングサービス](https://www.ccswitch.io/en/docs?section=proxy&item=service)有効で使用している場合、追加設定なしで`llm.url`をCC-Switchのプロキシアドレスに向けることができます：
 > - **Claude**プロバイダーの場合: `llm.url`を`http://127.0.0.1:15721`に設定
 > - **Codex**プロバイダーの場合: `llm.url`を`http://127.0.0.1:15721/v1`に設定
@@ -462,6 +487,9 @@ OCRは4層の優先度チェーンを使ってレビュールールを解決し
 | `llm.auth_header` | string | Anthropicのみ：`x-api-key` \| `authorization` |
 | `llm.model` | string | `claude-opus-4-6` |
 | `llm.use_anthropic` | boolean | `true` \| `false` |
+| `llm.use_anthropic_vertex` | boolean | Google Vertex AIでClaudeを使用 |
+| `llm.vertex_project_id` | string | Google CloudプロジェクトID |
+| `llm.vertex_region` | string | Vertex AIリージョン |
 | `language` | string | 任意の言語名、例：`English`、`Chinese`（デフォルト：`English`） |
 | `telemetry.enabled` | boolean | `true` \| `false` |
 | `telemetry.exporter` | string | `console` \| `otlp` |
@@ -479,6 +507,9 @@ OCRは4層の優先度チェーンを使ってレビュールールを解決し
 | `OCR_LLM_AUTH_HEADER` | Anthropic認証ヘッダー（`x-api-key`または`authorization`） |
 | `OCR_LLM_MODEL` | モデル名 |
 | `OCR_USE_ANTHROPIC` | `true` = Anthropic、`false` = OpenAI |
+| `OCR_USE_VERTEX` | Vertex AIモードを有効化 |
+| `OCR_VERTEX_PROJECT_ID` | Google CloudプロジェクトID。`ANTHROPIC_VERTEX_PROJECT_ID`と`GOOGLE_CLOUD_PROJECT`もサポート |
+| `OCR_VERTEX_REGION` | Vertex AIリージョン。`CLOUD_ML_REGION`もサポート |
 
 
 ## テレメトリー

diff --git a/README.ko-KR.md b/README.ko-KR.md
@@ -152,6 +152,31 @@ export OCR_USE_ANTHROPIC=true
 
 Claude Code 환경 변수(`ANTHROPIC_BASE_URL`, `ANTHROPIC_AUTH_TOKEN`, `ANTHROPIC_MODEL`)와도 호환되며, `~/.zshrc` / `~/.bashrc`의 export도 파싱합니다.
 
+### Google Vertex AI에서 Claude 사용
+
+OCR은 Google Application Default Credentials(ADC)를 사용해 Vertex AI의 Claude를 호출할 수 있습니다. Anthropic API key나 custom LLM URL은 필요하지 않습니다.
+
+```bash
+export OCR_USE_VERTEX=true
+export OCR_LLM_MODEL=claude-sonnet-4-6
+export ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project
+export CLOUD_ML_REGION=global
+
+gcloud auth application-default login
+ocr llm test
+```
+
+OCR 전용 alias인 `OCR_VERTEX_PROJECT_ID`와 `OCR_VERTEX_REGION`도 지원합니다. project ID는 `GOOGLE_CLOUD_PROJECT`로도 fallback됩니다. 선택한 GCP project는 Vertex AI API가 활성화되어 있어야 하며 대상 Claude model에 접근할 수 있어야 합니다.
+
+같은 설정은 기존 `llm` 설정에도 저장할 수 있습니다.
+
+```bash
+ocr config set llm.use_anthropic_vertex true
+ocr config set llm.vertex_project_id your-gcp-project
+ocr config set llm.vertex_region global
+ocr config set llm.model claude-sonnet-4-6
+```
+
 > **CC-Switch 사용자 참고**: [CC-Switch](https://github.com/farion1231/cc-switch)를 [routing service](https://www.ccswitch.io/en/docs?section=proxy&item=service)와 함께 사용한다면, 추가 설정 없이 `llm.url`을 CC-Switch proxy 주소로 지정할 수 있습니다.
 > - **Claude** provider: `llm.url`을 `http://127.0.0.1:15721`로 설정
 > - **Codex** provider: `llm.url`을 `http://127.0.0.1:15721/v1`로 설정
@@ -420,6 +445,9 @@ Config file: `~/.opencodereview/config.json`
 | `llm.auth_header` | string | Anthropic only: `x-api-key` \| `authorization` |
 | `llm.model` | string | `claude-opus-4-6` |
 | `llm.use_anthropic` | boolean | `true` \| `false` |
+| `llm.use_anthropic_vertex` | boolean | Google Vertex AI에서 Claude 사용 |
+| `llm.vertex_project_id` | string | Google Cloud project ID |
+| `llm.vertex_region` | string | Vertex AI region |
 | `language` | string | 임의의 언어 이름, 예: `English`, `Chinese` (기본값: `English`) |
 | `telemetry.enabled` | boolean | `true` \| `false` |
 | `telemetry.exporter` | string | `console` \| `otlp` |
@@ -437,6 +465,9 @@ Config file: `~/.opencodereview/config.json`
 | `OCR_LLM_AUTH_HEADER` | Anthropic auth header (`x-api-key` 또는 `authorization`) |
 | `OCR_LLM_MODEL` | Model name |
 | `OCR_USE_ANTHROPIC` | `true` = Anthropic, `false` = OpenAI |
+| `OCR_USE_VERTEX` | Vertex AI 모드 활성화 |
+| `OCR_VERTEX_PROJECT_ID` | Google Cloud project ID. `ANTHROPIC_VERTEX_PROJECT_ID`와 `GOOGLE_CLOUD_PROJECT`도 지원 |
+| `OCR_VERTEX_REGION` | Vertex AI region. `CLOUD_ML_REGION`도 지원 |
 
 ## Telemetry
 

diff --git a/README.md b/README.md
@@ -152,6 +152,31 @@ export OCR_USE_ANTHROPIC=true
 
 It is also compatible with Claude Code environment variables (`ANTHROPIC_BASE_URL`, `ANTHROPIC_AUTH_TOKEN`, `ANTHROPIC_MODEL`) and parses `~/.zshrc` / `~/.bashrc` for those exports.
 
+### Claude on Google Vertex AI
+
+OCR can use Claude through Vertex AI with Google Application Default Credentials (ADC). No Anthropic API key or custom LLM URL is required:
+
+```bash
+export OCR_USE_VERTEX=true
+export OCR_LLM_MODEL=claude-sonnet-4-6
+export ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project
+export CLOUD_ML_REGION=global
+
+gcloud auth application-default login
+ocr llm test
+```
+
+`OCR_VERTEX_PROJECT_ID` and `OCR_VERTEX_REGION` are supported as OCR-specific aliases. `GOOGLE_CLOUD_PROJECT` is also accepted as a project ID fallback. The selected GCP project must have the Vertex AI API enabled and access to the requested Claude model.
+
+The same setup can be stored in the legacy `llm` configuration:
+
+```bash
+ocr config set llm.use_anthropic_vertex true
+ocr config set llm.vertex_project_id your-gcp-project
+ocr config set llm.vertex_region global
+ocr config set llm.model claude-sonnet-4-6
+```
+
 > **Note for CC-Switch Users**: If you are using [CC-Switch](https://github.com/farion1231/cc-switch) with [routing service](https://www.ccswitch.io/en/docs?section=proxy&item=service) enabled, you can point `llm.url` to the CC-Switch proxy address without additional configuration:
 > - For **Claude** provider: set `llm.url` to `http://127.0.0.1:15721`
 > - For **Codex** provider: set `llm.url` to `http://127.0.0.1:15721/v1`
@@ -464,6 +489,9 @@ Config file: `~/.opencodereview/config.json`
 | `llm.auth_header` | string | Anthropic only: `x-api-key` \| `authorization` |
 | `llm.model` | string | `claude-opus-4-6` |
 | `llm.use_anthropic` | boolean | `true` \| `false` |
+| `llm.use_anthropic_vertex` | boolean | Use Claude through Google Vertex AI |
+| `llm.vertex_project_id` | string | Google Cloud project ID |
+| `llm.vertex_region` | string | Vertex AI region |
 | `language` | string | Any language name, e.g. `English`, `Chinese` (default: `English`) |
 | `telemetry.enabled` | boolean | `true` \| `false` |
 | `telemetry.exporter` | string | `console` \| `otlp` |
@@ -481,6 +509,9 @@ Environment variables take precedence over the config file.
 | `OCR_LLM_AUTH_HEADER` | Anthropic auth header (`x-api-key` or `authorization`) |
 | `OCR_LLM_MODEL` | Model name |
 | `OCR_USE_ANTHROPIC` | `true` = Anthropic, `false` = OpenAI |
+| `OCR_USE_VERTEX` | Enable Vertex AI mode |
+| `OCR_VERTEX_PROJECT_ID` | Google Cloud project ID; aliases `ANTHROPIC_VERTEX_PROJECT_ID` and `GOOGLE_CLOUD_PROJECT` are supported |
+| `OCR_VERTEX_REGION` | Vertex AI region; `CLOUD_ML_REGION` is also supported |
 
 
 ## Telemetry

diff --git a/README.ru-RU.md b/README.ru-RU.md
@@ -152,6 +152,31 @@ export OCR_USE_ANTHROPIC=true
 
 Инструмент также совместим с переменными окружения Claude Code (`ANTHROPIC_BASE_URL`, `ANTHROPIC_AUTH_TOKEN`, `ANTHROPIC_MODEL`) и разбирает `~/.zshrc` / `~/.bashrc` в поисках соответствующих export'ов.
 
+### Claude через Google Vertex AI
+
+OCR может вызывать Claude через Vertex AI с помощью Google Application Default Credentials (ADC). Ключ Anthropic API и пользовательский LLM URL не требуются:
+
+```bash
+export OCR_USE_VERTEX=true
+export OCR_LLM_MODEL=claude-sonnet-4-6
+export ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project
+export CLOUD_ML_REGION=global
+
+gcloud auth application-default login
+ocr llm test
+```
+
+Также поддерживаются OCR-алиасы `OCR_VERTEX_PROJECT_ID` и `OCR_VERTEX_REGION`. Для project ID также используется fallback на `GOOGLE_CLOUD_PROJECT`. В выбранном GCP-проекте должен быть включён Vertex AI API и доступ к нужной модели Claude.
+
+Эту же настройку можно сохранить в существующей секции `llm`:
+
+```bash
+ocr config set llm.use_anthropic_vertex true
+ocr config set llm.vertex_project_id your-gcp-project
+ocr config set llm.vertex_region global
+ocr config set llm.model claude-sonnet-4-6
+```
+
 > **Примечание для пользователей CC-Switch**: если вы используете [CC-Switch](https://github.com/farion1231/cc-switch) с включённым [routing service](https://www.ccswitch.io/en/docs?section=proxy&item=service), можно указать в `llm.url` адрес прокси CC-Switch без дополнительной настройки:
 > - для провайдера **Claude**: установите `llm.url` в `http://127.0.0.1:15721`
 > - для провайдера **Codex**: установите `llm.url` в `http://127.0.0.1:15721/v1`
@@ -464,6 +489,9 @@ OCR разрешает правила ревью по цепочке приор
 | `llm.auth_header` | string | Только для Anthropic: `x-api-key` \| `authorization` |
 | `llm.model` | string | `claude-opus-4-6` |
 | `llm.use_anthropic` | boolean | `true` \| `false` |
+| `llm.use_anthropic_vertex` | boolean | Использовать Claude через Google Vertex AI |
+| `llm.vertex_project_id` | string | ID проекта Google Cloud |
+| `llm.vertex_region` | string | Регион Vertex AI |
 | `language` | string | Любое название языка, например `English`, `Chinese` (по умолчанию: `English`) |
 | `telemetry.enabled` | boolean | `true` \| `false` |
 | `telemetry.exporter` | string | `console` \| `otlp` |
@@ -481,6 +509,9 @@ OCR разрешает правила ревью по цепочке приор
 | `OCR_LLM_AUTH_HEADER` | Заголовок авторизации Anthropic (`x-api-key` или `authorization`) |
 | `OCR_LLM_MODEL` | Имя модели |
 | `OCR_USE_ANTHROPIC` | `true` = Anthropic, `false` = OpenAI |
+| `OCR_USE_VERTEX` | Включить режим Vertex AI |
+| `OCR_VERTEX_PROJECT_ID` | ID проекта Google Cloud; также поддерживаются `ANTHROPIC_VERTEX_PROJECT_ID` и `GOOGLE_CLOUD_PROJECT` |
+| `OCR_VERTEX_REGION` | Регион Vertex AI; также поддерживается `CLOUD_ML_REGION` |
 
 
 ## Телеметрия

diff --git a/README.zh-CN.md b/README.zh-CN.md
@@ -152,6 +152,31 @@ export OCR_USE_ANTHROPIC=true
 
 同时兼容了 Claude Code 环境变量（`ANTHROPIC_BASE_URL`、`ANTHROPIC_AUTH_TOKEN`、`ANTHROPIC_MODEL`），并解析 `~/.zshrc` / `~/.bashrc` 中的相关导出。
 
+### 通过 Google Vertex AI 使用 Claude
+
+OCR 可以通过 Google 应用默认凭据（ADC）调用 Vertex AI 上的 Claude，不需要 Anthropic API Key 或自定义 LLM URL：
+
+```bash
+export OCR_USE_VERTEX=true
+export OCR_LLM_MODEL=claude-sonnet-4-6
+export ANTHROPIC_VERTEX_PROJECT_ID=your-gcp-project
+export CLOUD_ML_REGION=global
+
+gcloud auth application-default login
+ocr llm test
+```
+
+也可以使用 OCR 专用别名 `OCR_VERTEX_PROJECT_ID` 和 `OCR_VERTEX_REGION`。项目 ID 还会回退读取 `GOOGLE_CLOUD_PROJECT`。所选 GCP 项目必须启用 Vertex AI API，并具有目标 Claude 模型的访问权限。
+
+同样可以写入已有的 `llm` 配置：
+
+```bash
+ocr config set llm.use_anthropic_vertex true
+ocr config set llm.vertex_project_id your-gcp-project
+ocr config set llm.vertex_region global
+ocr config set llm.model claude-sonnet-4-6
+```
+
 > **CC-Switch 用户特别提醒**：如果你使用 [CC-Switch](https://github.com/farion1231/cc-switch) 并开启了[路由服务](https://www.ccswitch.io/zh/docs?section=proxy&item=service)，可以将 `llm.url` 配置成 CC-Switch 启动的代理地址，无需额外配置：
 > - 如果路由的是 **Claude** 供应商：设置 `llm.url` 为 `http://127.0.0.1:15721`
 > - 如果路由的是 **Codex** 供应商：设置 `llm.url` 为 `http://127.0.0.1:15721/v1`
@@ -452,6 +477,9 @@ OCR 通过四层优先级链解析评审规则。每层采用首次匹配原则
 | `llm.auth_header` | string | 仅 Anthropic：`x-api-key` \| `authorization` |
 | `llm.model` | string | `claude-opus-4-6` |
 | `llm.use_anthropic` | boolean | `true` \| `false` |
+| `llm.use_anthropic_vertex` | boolean | 通过 Google Vertex AI 使用 Claude |
+| `llm.vertex_project_id` | string | Google Cloud 项目 ID |
+| `llm.vertex_region` | string | Vertex AI 区域 |
 | `language` | string | 任意语言名称，例如 `English`、`Chinese`（默认：`English`） |
 | `telemetry.enabled` | boolean | `true` \| `false` |
 | `telemetry.exporter` | string | `console` \| `otlp` |
@@ -469,6 +497,9 @@ OCR 通过四层优先级链解析评审规则。每层采用首次匹配原则
 | `OCR_LLM_AUTH_HEADER` | Anthropic 认证头（`x-api-key` 或 `authorization`） |
 | `OCR_LLM_MODEL` | 模型名称 |
 | `OCR_USE_ANTHROPIC` | `true` = Anthropic，`false` = OpenAI |
+| `OCR_USE_VERTEX` | 启用 Vertex AI 模式 |
+| `OCR_VERTEX_PROJECT_ID` | Google Cloud 项目 ID；也支持 `ANTHROPIC_VERTEX_PROJECT_ID` 和 `GOOGLE_CLOUD_PROJECT` |
+| `OCR_VERTEX_REGION` | Vertex AI 区域；也支持 `CLOUD_ML_REGION` |
 
 
 ## 遥测

diff --git a/cmd/opencodereview/config_cmd.go b/cmd/opencodereview/config_cmd.go
@@ -103,12 +103,15 @@ type Config struct {
 }
 
 type LlmConfig struct {
-	URL          string         `json:"url,omitempty"`
-	AuthToken    string         `json:"auth_token,omitempty"`
-	AuthHeader   string         `json:"auth_header,omitempty"`
-	Model        string         `json:"model,omitempty"`
-	UseAnthropic *bool          `json:"use_anthropic,omitempty"` // nil = default true; false = OpenAI protocol
-	ExtraBody    map[string]any `json:"extra_body,omitempty"`
+	URL                string         `json:"url,omitempty"`
+	AuthToken          string         `json:"auth_token,omitempty"`
+	AuthHeader         string         `json:"auth_header,omitempty"`
+	Model              string         `json:"model,omitempty"`
+	UseAnthropic       *bool          `json:"use_anthropic,omitempty"` // nil = default true; false = OpenAI protocol
+	UseAnthropicVertex bool           `json:"use_anthropic_vertex,omitempty"`
+	VertexProjectID    string         `json:"vertex_project_id,omitempty"`
+	VertexRegion       string         `json:"vertex_region,omitempty"`
+	ExtraBody          map[string]any `json:"extra_body,omitempty"`
 }
 
 // TelemetryConfig holds telemetry-specific settings.
@@ -218,6 +221,16 @@ func setConfigValue(cfg *Config, key, value string) error {
 			return fmt.Errorf("invalid boolean for llm.use_anthropic: %w", err)
 		}
 		cfg.Llm.UseAnthropic = &b
+	case "llm.use_anthropic_vertex", "llm.UseAnthropicVertex":
+		b, err := strconv.ParseBool(value)
+		if err != nil {
+			return fmt.Errorf("invalid boolean for llm.use_anthropic_vertex: %w", err)
+		}
+		cfg.Llm.UseAnthropicVertex = b
+	case "llm.vertex_project_id", "llm.VertexProjectID":
+		cfg.Llm.VertexProjectID = value
+	case "llm.vertex_region", "llm.VertexRegion":
+		cfg.Llm.VertexRegion = value
 	case "language", "Language":
 		cfg.Language = value
 	case "telemetry.enabled", "telemetry.Enabled":
@@ -247,7 +260,7 @@ func setConfigValue(cfg *Config, key, value string) error {
 		}
 		cfg.Llm.ExtraBody = m
 	default:
-		return fmt.Errorf("unknown config key: %s\nSupported keys: provider, model, providers.<name>.<field>, custom_providers.<name>.<field>, llm.url, llm.auth_token, llm.auth_header, llm.model, llm.use_anthropic, llm.extra_body, language, telemetry.enabled, telemetry.exporter, telemetry.otlp_endpoint, telemetry.content_logging\nProvider fields: api_key, url, protocol, model, models, auth_header, extra_body", key)
+		return fmt.Errorf("unknown config key: %s\nSupported keys: provider, model, providers.<name>.<field>, custom_providers.<name>.<field>, llm.url, llm.auth_token, llm.auth_header, llm.model, llm.use_anthropic, llm.use_anthropic_vertex, llm.vertex_project_id, llm.vertex_region, llm.extra_body, language, telemetry.enabled, telemetry.exporter, telemetry.otlp_endpoint, telemetry.content_logging\nProvider fields: api_key, url, protocol, model, models, auth_header, extra_body", key)
 	}
 	return nil
 }

diff --git a/cmd/opencodereview/config_cmd_test.go b/cmd/opencodereview/config_cmd_test.go
@@ -200,6 +200,24 @@ func TestSetConfigValueProviderEntryExtraBody(t *testing.T) {
 	}
 }
 
+func TestSetConfigValueLegacyVertexFields(t *testing.T) {
+	cfg := &Config{}
+
+	if err := setConfigValue(cfg, "llm.use_anthropic_vertex", "true"); err != nil {
+		t.Fatalf("set use_anthropic_vertex: %v", err)
+	}
+	if err := setConfigValue(cfg, "llm.vertex_project_id", "test-project"); err != nil {
+		t.Fatalf("set vertex_project_id: %v", err)
+	}
+	if err := setConfigValue(cfg, "llm.vertex_region", "global"); err != nil {
+		t.Fatalf("set vertex_region: %v", err)
+	}
+
+	if !cfg.Llm.UseAnthropicVertex || cfg.Llm.VertexProjectID != "test-project" || cfg.Llm.VertexRegion != "global" {
+		t.Errorf("Llm Vertex config = %#v", cfg.Llm)
+	}
+}
+
 func TestSetConfigValueModelWithCustomProvider(t *testing.T) {
 	cfg := &Config{
 		Provider: "my-gateway",

diff --git a/cmd/opencodereview/flags.go b/cmd/opencodereview/flags.go
@@ -296,6 +296,6 @@ Examples:
   ocr config set language English
   ocr config set telemetry.enabled true
 
-Supported keys: provider, model, providers.<name>.<field>, custom_providers.<name>.<field>, llm.url, llm.auth_token, llm.auth_header, llm.model, llm.use_anthropic, llm.extra_body, language, telemetry.enabled, telemetry.exporter, telemetry.otlp_endpoint, telemetry.content_logging
+Supported keys: provider, model, providers.<name>.<field>, custom_providers.<name>.<field>, llm.url, llm.auth_token, llm.auth_header, llm.model, llm.use_anthropic, llm.use_anthropic_vertex, llm.vertex_project_id, llm.vertex_region, llm.extra_body, language, telemetry.enabled, telemetry.exporter, telemetry.otlp_endpoint, telemetry.content_logging
 Provider fields: api_key, url, protocol, model, models, auth_header, extra_body`)
 }
diff --git a/cmd/opencodereview/llm_cmd.go b/cmd/opencodereview/llm_cmd.go
@@ -59,7 +59,10 @@ func runLLMTest() error {
 		timeout = time.Duration(task.Timeout) * time.Second
 	}
 
-	llmClient := llm.NewLLMClient(ep)
+	llmClient, err := llm.NewLLMClient(context.Background(), ep)
+	if err != nil {
+		return fmt.Errorf("create LLM client: %w", err)
+	}
 
 	messages := make([]llm.Message, 0, len(task.Messages))
 	for _, m := range task.Messages {

diff --git a/cmd/opencodereview/review_cmd.go b/cmd/opencodereview/review_cmd.go
@@ -95,7 +95,10 @@ func runReview(args []string) error {
 		return fmt.Errorf("resolve LLM endpoint: %w", err)
 	}
 
-	llmClient := llm.NewLLMClient(ep)
+	llmClient, err := llm.NewLLMClient(context.Background(), ep)
+	if err != nil {
+		return fmt.Errorf("create LLM client: %w", err)
+	}
 	model := ep.Model
 
 	gitRunner := gitcmd.New(opts.maxGitProcs)