docs(comments): 注释中文化 · internal/runtime/multimodal（21 文件，纯替换）

- 只翻注释/文档字符串，代码零变化：go/scanner 词法流比对 21 文件 token+字面量全等，增删行数 1:1（467/467）
- 刻意保留非散文英文：JSON Schema description、工具 Description() 返回值、adapter prompt、API 路径与命令示例、环境变量名
- 逐行确认代码前缀未变，未触碰其他模块

internal/runtime/multimodal/vision/adapter.go

@@ -13,28 +13,28 @@ import (
 	"git.yeij.top/AskaEth/Cyrene_v2/internal/config"
 )
 
-// visionMaxTokens caps the vision/OCR response length.
+// visionMaxTokens 限制 vision/OCR 响应的长度。
 const visionMaxTokens = 512
 
-// VisionAdapter understands an image and returns a description of it. It is the
-// backend behind the vision tool; real providers implement it without exposing
-// the image to the main-model context.
+// VisionAdapter 理解一张图像并返回对它的描述。它是
+// 视觉工具背后的后端；真实提供方实现它，且不把图像
+// 暴露给主模型上下文。
 type VisionAdapter interface {
-	// Describe returns a description of img. prompt is an optional instruction
-	// overriding the default "describe this image" ask.
+	// Describe 返回 img 的描述。prompt 是可选的指令，
+	// 用于覆盖默认的「描述这张图」提问。
 	Describe(ctx context.Context, img ImageRef, prompt string) (string, error)
 }
 
-// OCRAdapter extracts text from an image. It is the backend behind the OCR tool.
+// OCRAdapter 从图像中抽取文字。它是 OCR 工具背后的后端。
 type OCRAdapter interface {
-	// Extract returns the text found in img.
+	// Extract 返回 img 中发现的文字。
 	Extract(ctx context.Context, img ImageRef) (string, error)
 }
 
-// OpenAICompatibleAdapter implements both VisionAdapter and OCRAdapter by
-// calling an OpenAI-compatible /chat/completions endpoint with the image
-// embedded as a content part. It is the ready default implementation; real
-// cloud vision/OCR providers expose the same protocol.
+// OpenAICompatibleAdapter 通过调用 OpenAI 兼容的 /chat/completions
+// endpoint（图像作为内容部件内嵌）同时实现 VisionAdapter 与 OCRAdapter。
+// 它是开箱可用的默认实现；真实的
+// 云端 vision/OCR 提供方暴露相同协议。
 type OpenAICompatibleAdapter struct {
 	cfg    config.LLMConfig
 	model  string
@@ -42,10 +42,10 @@ type OpenAICompatibleAdapter struct {
 	log    *slog.Logger
 }
 
-// OpenAIOption customizes an OpenAICompatibleAdapter.
+// OpenAIOption 用于定制 OpenAICompatibleAdapter。
 type OpenAIOption func(*OpenAICompatibleAdapter)
 
-// WithModel overrides the adapter model (defaults to cfg.Model).
+// WithModel 覆盖适配器的模型（默认取 cfg.Model）。
 func WithModel(m string) OpenAIOption {
 	return func(a *OpenAICompatibleAdapter) {
 		if m != "" {
@@ -54,7 +54,7 @@ func WithModel(m string) OpenAIOption {
 	}
 }
 
-// WithHTTPClient sets the *http.Client used for the request; nil keeps default.
+// WithHTTPClient 设置请求所用的 *http.Client；nil 表示保持默认。
 func WithHTTPClient(c *http.Client) OpenAIOption {
 	return func(a *OpenAICompatibleAdapter) {
 		if c != nil {
@@ -63,7 +63,7 @@ func WithHTTPClient(c *http.Client) OpenAIOption {
 	}
 }
 
-// WithLogger sets the logger used on failures; nil keeps a discard logger.
+// WithLogger 设置失败时使用的日志器；nil 表示保持丢弃日志器。
 func WithLogger(l *slog.Logger) OpenAIOption {
 	return func(a *OpenAICompatibleAdapter) {
 		if l != nil {
@@ -72,8 +72,8 @@ func WithLogger(l *slog.Logger) OpenAIOption {
 	}
 }
 
-// NewOpenAICompatible builds an adapter from cfg, using cfg.Model unless
-// overridden.
+// NewOpenAICompatible 用 cfg 构造一个适配器，未覆盖时使用
+// cfg.Model。
 func NewOpenAICompatible(cfg config.LLMConfig, opts ...OpenAIOption) *OpenAICompatibleAdapter {
 	a := &OpenAICompatibleAdapter{
 		cfg:    cfg,
@@ -87,7 +87,7 @@ func NewOpenAICompatible(cfg config.LLMConfig, opts ...OpenAIOption) *OpenAIComp
 	return a
 }
 
-// Describe returns an image description from the underlying chat endpoint.
+// Describe 从底层 chat endpoint 返回图像描述。
 func (a *OpenAICompatibleAdapter) Describe(ctx context.Context, img ImageRef, prompt string) (string, error) {
 	if prompt == "" {
 		prompt = "Describe this image in detail."
@@ -95,12 +95,12 @@ func (a *OpenAICompatibleAdapter) Describe(ctx context.Context, img ImageRef, pr
 	return a.ask(ctx, img, prompt)
 }
 
-// Extract returns text found in the image via the same endpoint.
+// Extract 经同一 endpoint 返回图像中发现的文字。
 func (a *OpenAICompatibleAdapter) Extract(ctx context.Context, img ImageRef) (string, error) {
 	return a.ask(ctx, img, "Extract all text visible in this image. Return only the text.")
 }
 
-// ask resolves the image, builds a chat request and returns the assistant text.
+// ask 解析图像、构造 chat 请求并返回助手文本。
 func (a *OpenAICompatibleAdapter) ask(ctx context.Context, img ImageRef, prompt string) (string, error) {
 	if strings.TrimSpace(a.cfg.BaseURL) == "" {
 		return "", ErrNotConfigured
@@ -139,8 +139,8 @@ func (a *OpenAICompatibleAdapter) ask(ctx context.Context, img ImageRef, prompt
 	return a.decode(resp.Body)
 }
 
-// imageURL converts an ImageRef to the image_url value the wire expects: an
-// http(s) URL passes through, anything else becomes a base64 data: URI.
+// imageURL 把 ImageRef 转换为线上格式期望的 image_url 取值：http(s)
+// URL 原样透传，其余一律转为 base64 data: URI。
 func (a *OpenAICompatibleAdapter) imageURL(ctx context.Context, img ImageRef) (string, error) {
 	if img.Source == SourceURL {
 		return img.Data, nil
@@ -152,7 +152,7 @@ func (a *OpenAICompatibleAdapter) imageURL(ctx context.Context, img ImageRef) (s
 	return DataURI(mime, data), nil
 }
 
-// openAIVisionRequest is the JSON body sent to /chat/completions.
+// openAIVisionRequest 是发往 /chat/completions 的 JSON 请求体。
 type openAIVisionRequest struct {
 	Model     string          `json:"model"`
 	Messages  []openAIMessage `json:"messages"`
@@ -186,7 +186,7 @@ type openAIResponseMessage struct {
 	Content string `json:"content"`
 }
 
-// request builds the wire body for an image+prompt.
+// request 构造图像 + prompt 的线上请求体。
 func (a *OpenAICompatibleAdapter) request(imageURL, prompt string) openAIVisionRequest {
 	return openAIVisionRequest{
 		Model: a.model,
@@ -200,7 +200,7 @@ func (a *OpenAICompatibleAdapter) request(imageURL, prompt string) openAIVisionR
 	}
 }
 
-// decode extracts the first choice's text from a response body.
+// decode 从响应体中取出第一个 choice 的文本。
 func (a *OpenAICompatibleAdapter) decode(r io.Reader) (string, error) {
 	var wire openAIVisionResponse
 	if err := json.NewDecoder(r).Decode(&wire); err != nil {
@@ -212,6 +212,6 @@ func (a *OpenAICompatibleAdapter) decode(r io.Reader) (string, error) {
 	return wire.Choices[0].Message.Content, nil
 }
 
-// interface assertion: one concrete adapter serves both directions.
+// 接口断言：同一个具体适配器同时服务两个方向。
 var _ VisionAdapter = (*OpenAICompatibleAdapter)(nil)
 var _ OCRAdapter = (*OpenAICompatibleAdapter)(nil)