From b2dff38f61d84472225b8c3796192fc794ac4c09 Mon Sep 17 00:00:00 2001
From: tangwang <tangwang@essa.top>
Date: Mon, 20 Apr 2026 22:45:07 +0800
Subject: [PATCH] embedding-image接口（POST /embed/image）支持SVG格式：先转 PNG 再走 CN-CLIP

---
 embeddings/server.py               | 96 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-------------
 requirements_embedding_service.txt |  3 +++
 2 files changed, 86 insertions(+), 13 deletions(-)

diff --git a/embeddings/server.py b/embeddings/server.py
index ef92bc7..dec063d 100644
--- a/embeddings/server.py
+++ b/embeddings/server.py
@@ -15,9 +15,11 @@ import time
 import uuid
 from collections import deque
 from dataclasses import dataclass
+import tempfile
 from typing import Any, Dict, List, Optional
 
 import numpy as np
+import requests
 from fastapi import FastAPI, HTTPException, Request, Response
 from fastapi.concurrency import run_in_threadpool
 
@@ -847,19 +849,36 @@ def _embed_image_lane_impl(
         )
 
     backend_t0 = time.perf_counter()
-    with _image_encode_lock:
-        if lane == "image":
-            vectors = _image_model.encode_image_urls(
-                missing_items,
-                batch_size=CONFIG.IMAGE_BATCH_SIZE,
-                normalize_embeddings=effective_normalize,
-            )
-        else:
-            vectors = _image_model.encode_clip_texts(
-                missing_items,
-                batch_size=CONFIG.IMAGE_BATCH_SIZE,
-                normalize_embeddings=effective_normalize,
-            )
+    tmp_png_paths: List[str] = []
+    encode_inputs: List[str] = list(missing_items)
+    if lane == "image":
+        # Best-effort: rasterize SVGs into temporary PNGs so CN-CLIP can encode them.
+        for i, item in enumerate(missing_items):
+            if _looks_like_svg_image_ref(item):
+                png_path = _rasterize_svg_to_temp_png(item)
+                tmp_png_paths.append(png_path)
+                encode_inputs[i] = png_path
+
+    try:
+        with _image_encode_lock:
+            if lane == "image":
+                vectors = _image_model.encode_image_urls(
+                    encode_inputs,
+                    batch_size=CONFIG.IMAGE_BATCH_SIZE,
+                    normalize_embeddings=effective_normalize,
+                )
+            else:
+                vectors = _image_model.encode_clip_texts(
+                    missing_items,
+                    batch_size=CONFIG.IMAGE_BATCH_SIZE,
+                    normalize_embeddings=effective_normalize,
+                )
+    finally:
+        for p in tmp_png_paths:
+            try:
+                os.remove(p)
+            except Exception:
+                pass
     if vectors is None or len(vectors) != len(missing_items):
         raise RuntimeError(
             f"{lane} lane length mismatch: expected {len(missing_items)}, "
@@ -1284,6 +1303,57 @@ def _parse_string_inputs(raw: List[Any], *, kind: str, empty_detail: str) -> Lis
     return out
 
 
+def _looks_like_svg_image_ref(value: str) -> bool:
+    """
+    CN-CLIP image embedding path expects raster images (jpg/png/webp/...) that PIL can decode.
+    SVG is a vector format and currently not supported by the image embedding backend.
+    """
+    v = (value or "").strip().lower()
+    if not v:
+        return False
+    # Drop query/fragment for URL suffix check.
+    for sep in ("?", "#"):
+        if sep in v:
+            v = v.split(sep, 1)[0]
+    return v.endswith(".svg") or v.startswith("data:image/svg+xml")
+
+
+def _rasterize_svg_to_temp_png(svg_ref: str, *, timeout_sec: int = 10) -> str:
+    """
+    Download/resolve an SVG ref (URL or local path) and rasterize it into a temporary PNG file.
+
+    Returns the PNG file path (caller is responsible for cleanup).
+    """
+    if svg_ref.startswith(("http://", "https://")):
+        resp = requests.get(svg_ref, timeout=timeout_sec)
+        if resp.status_code != 200:
+            raise ValueError(f"HTTP {resp.status_code} when downloading SVG")
+        svg_bytes = resp.content
+    else:
+        with open(svg_ref, "rb") as f:
+            svg_bytes = f.read()
+
+    try:
+        import cairosvg  # type: ignore
+    except Exception as exc:  # pragma: no cover
+        raise RuntimeError(
+            "SVG rasterization requires optional dependency 'cairosvg'. "
+            "Install it in the embedding-image service environment."
+        ) from exc
+
+    fd, out_path = tempfile.mkstemp(prefix="embed_svg_", suffix=".png")
+    os.close(fd)
+    try:
+        cairosvg.svg2png(bytestring=svg_bytes, write_to=out_path)
+    except Exception:
+        try:
+            os.remove(out_path)
+        except Exception:
+            pass
+        raise
+    return out_path
+
+
 async def _run_image_lane_embed(
     *,
     route: str,
diff --git a/requirements_embedding_service.txt b/requirements_embedding_service.txt
index 1086750..4dd4e6a 100644
--- a/requirements_embedding_service.txt
+++ b/requirements_embedding_service.txt
@@ -12,6 +12,9 @@ numpy>=1.24.0
 pyyaml>=6.0
 redis>=5.0.0
 
+# Optional: rasterize SVG to PNG for /embed/image
+cairosvg>=2.7.0
+
 # Image backend via clip-as-service client
 setuptools<82
 jina>=3.12.0
--
libgit2 0.21.2