feat(url): 支持docreader不上传替换白名单url的图片

.env.example

@@ -493,6 +493,11 @@ DOCREADER_TRANSPORT=grpc
 # Weaviate gRPC 地址（Docker 内：weaviate:50051；宿主机访问：localhost:50052）
 # WEAVIATE_GRPC_ADDRESS=weaviate:50051
 
+# 保留原始 URL 的图片域名白名单（可选，逗号分隔）
+# 配置后，这些域名的图片仍会被下载和分析（OCR/字幕），但 markdown 中保留原始 URL，
+# 不会被替换为对象存储的 provider:// URL。适用于内部稳定服务（如自建 MinerU）。
+# IMAGE_HOST_KEEP_URL=mineru.internal.example.com
+
 # Weaviate 架构模式
 # WEAVIATE_SCHEME=http
 

.env.lite.example

@@ -43,6 +43,8 @@ NEO4J_ENABLE=false
 WEKNORA_SANDBOX_MODE=disabled
 ENABLE_GRAPH_RAG=false
 DISABLE_REGISTRATION=false
+# 保留原始 URL 的图片域名白名单（可选）
+# IMAGE_HOST_KEEP_URL=mineru.internal.example.com
 
 # === Langfuse 可观测性（可选） ===
 # 追踪 chat / embedding / rerank / VLM / ASR 的 prompt、响应与 token 消耗。

docker-compose.yml

@@ -149,6 +149,8 @@ services:
       - TENANT_AES_KEY=${TENANT_AES_KEY:-}
       - SYSTEM_AES_KEY=${SYSTEM_AES_KEY:-}
       - SSRF_WHITELIST=${SSRF_WHITELIST:-}
+      # 保留原始 URL 的图片域名白名单（逗号分隔，不替换为 provider://）
+      - IMAGE_HOST_KEEP_URL=${IMAGE_HOST_KEEP_URL:-}
       # Always allow the optional searxng sidecar (compose service hostname);
       # merged on top of SSRF_WHITELIST so user overrides don't clobber it.
       - SSRF_WHITELIST_EXTRA=${SSRF_WHITELIST_EXTRA:-searxng}

internal/infrastructure/docparser/image_resolver.go

@@ -13,6 +13,8 @@ import (
 	"log"
 	"mime"
 	"net/http"
+	"net/url"
+	"os"
 	"path"
 	"path/filepath"
 	"regexp"
@@ -195,6 +197,35 @@ func isProviderScheme(p string) bool {
 	return false
 }
 
+// isWhitelistedImageHost checks if the image URL's host is in the whitelist.
+// Whitelisted hosts are trusted (e.g. internal MinerU service) — images are
+// still downloaded for validation and OCR/caption analysis, but not uploaded
+// to object storage. The markdown keeps the original URL.
+// Configure via IMAGE_HOST_KEEP_URL env var (comma-separated hosts).
+func isWhitelistedImageHost(rawURL string) bool {
+	whitelist := strings.TrimSpace(os.Getenv("IMAGE_HOST_KEEP_URL"))
+	if whitelist == "" {
+		return false
+	}
+	u, err := url.Parse(rawURL)
+	if err != nil || u.Host == "" {
+		return false
+	}
+	host := strings.ToLower(u.Host)
+	hostname := strings.ToLower(u.Hostname())
+	for _, h := range strings.Split(whitelist, ",") {
+		h = strings.ToLower(strings.TrimSpace(h))
+		if h == "" {
+			continue
+		}
+		// Exact host match (includes port) or hostname match (any port)
+		if host == h || hostname == h {
+			return true
+		}
+	}
+	return false
+}
+
 // ---------------------------------------------------------------------------
 // Helper functions for base64 image handling
 // ---------------------------------------------------------------------------
@@ -670,10 +701,18 @@ func (r *ImageResolver) ResolveRemoteImages(
 			continue
 		}
 
-		// --- SSRF check (centralised entry-point with whitelist support) ---
-		if err := secutils.ValidateURLForSSRF(imgURL); err != nil {
-			log.Printf("WARN: remote image blocked by SSRF check (%v): %s", err, imgURL)
-			continue
+		// For whitelisted hosts: download to validate (mime type, icon check),
+		// create StoredImage for downstream OCR/caption analysis, but do NOT
+		// upload to storage and keep the original URL in markdown.
+		// The multimodal service will download from the original URL later.
+		whitelisted := isWhitelistedImageHost(imgURL)
+
+		// --- SSRF check (skip for whitelisted) ---
+		if !whitelisted {
+			if err := secutils.ValidateURLForSSRF(imgURL); err != nil {
+				log.Printf("WARN: remote image blocked by SSRF check (%v): %s", err, imgURL)
+				continue
+			}
 		}
 
 		// --- Download ---
@@ -697,12 +736,20 @@ func (r *ImageResolver) ResolveRemoteImages(
 			ext = ".png" // safe default
 		}
 
-		// --- Upload to storage ---
-		fileName := uuid.New().String() + ext
-		servingURL, saveErr := fileSvc.SaveBytes(ctx, data, tenantID, fileName, false)
-		if saveErr != nil {
-			log.Printf("WARN: failed to save remote image %s: %v", imgURL, saveErr)
-			continue
+		var servingURL string
+		if whitelisted {
+			// Keep the original URL — ImageMultimodalService will download it
+			// directly for OCR/caption analysis.
+			servingURL = imgURL
+		} else {
+			// --- Upload to storage ---
+			fileName := uuid.New().String() + ext
+			var saveErr error
+			servingURL, saveErr = fileSvc.SaveBytes(ctx, data, tenantID, fileName, false)
+			if saveErr != nil {
+				log.Printf("WARN: failed to save remote image %s: %v", imgURL, saveErr)
+				continue
+			}
 		}
 
 		images = append(images, StoredImage{
@@ -711,8 +758,10 @@ func (r *ImageResolver) ResolveRemoteImages(
 			MimeType:    mimeType,
 		})
 
-		// Replace URL in markdown.
-		markdown = markdown[:m[4]] + servingURL + markdown[m[5]:]
+		if !whitelisted {
+			// Replace URL in markdown.
+			markdown = markdown[:m[4]] + servingURL + markdown[m[5]:]
+		}
 		processed++
 	}