@customize-agent/knowledge 4.0.14 → 4.0.15

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@customize-agent/knowledge",
-  "version": "4.0.14",
+  "version": "4.0.15",
   "description": "Local knowledge base infrastructure for customize-agent",
   "type": "module",
   "main": "./dist/index.js",
@@ -39,6 +39,7 @@
   "dependencies": {
     "@huggingface/transformers": "^3.8.0",
     "@napi-rs/canvas": "^0.1.82",
+    "@tesseract.js-data/chi_sim": "^1.0.0",
     "better-sqlite3": "^12.10.0",
     "dwgdxf": "^2.0.1",
     "dxf-parser": "^1.1.2",
@@ -47,6 +48,8 @@
     "jszip": "^3.10.1",
     "mammoth": "^1.12.0",
     "node-gyp": "^12.1.0",
+    "onnxruntime-node": "^1.27.0",
+    "paddleocr": "1.2.0",
     "pdf-parse": "^2.4.5",
     "pdfjs-dist": "^5.4.394",
     "sharp": "0.34.5",

package/scripts/download-paddleocr-models.sh

@@ -0,0 +1,63 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# PaddleOCR ONNX 模型下载脚本
+# 从 Hugging Face 下载 PP-OCRv6_small 模型文件
+#
+# 用法：
+#   bash scripts/download-paddleocr-models.sh
+#
+# 环境变量：
+#   PADDLEOCR_MODEL_PATH  - 模型存放目录（默认 packages/knowledge/models/paddleocr）
+#   HF_ENDPOINT           - Hugging Face 镜像（默认 https://huggingface.co）
+#                           国内可用 https://hf-mirror.com
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+MODEL_DIR="${PADDLEOCR_MODEL_PATH:-$SCRIPT_DIR/../models/paddleocr}"
+HF_BASE="${HF_ENDPOINT:-https://huggingface.co}"
+REPO="x3zvawq/paddleocr-js-onnx"
+
+# 模型文件列表
+FILES=(
+  "ppocr_v6_small/PP-OCRv6_small_det_infer.onnx"
+  "ppocr_v6_small/PP-OCRv6_small_rec_infer.onnx"
+  "ppocr_v6_small/ppocrv6_dict.txt"
+  "pp_lcnet_x0_25_textline_ori/PP-LCNet_x0_25_textline_ori_infer.onnx"
+)
+
+echo "==> 模型存放目录: $MODEL_DIR"
+mkdir -p "$MODEL_DIR"
+
+download_file() {
+  local remote_path="$1"
+  local local_name="$2"
+  local url="${HF_BASE}/${REPO}/resolve/main/${remote_path}"
+  local output="${MODEL_DIR}/${local_name}"
+
+  if [ -f "$output" ]; then
+    echo "  [跳过] $local_name 已存在 ($(wc -c < "$output" | tr -d ' ') bytes)"
+    return 0
+  fi
+
+  echo "  [下载] $local_name <- $url"
+  if command -v wget &>/dev/null; then
+    wget -q --show-progress --timeout=30 -O "$output" "$url"
+  elif command -v curl &>/dev/null; then
+    curl -L --connect-timeout 30 --max-time 300 -o "$output" "$url"
+  else
+    echo "错误: 需要 wget 或 curl" >&2
+    exit 1
+  fi
+
+  echo "        完成 ($(wc -c < "$output" | tr -d ' ') bytes)"
+}
+
+echo "==> 下载 PP-OCRv6_small 模型文件..."
+download_file "${FILES[0]}" "PP-OCRv6_small_det_infer.onnx"
+download_file "${FILES[1]}" "PP-OCRv6_small_rec_infer.onnx"
+download_file "${FILES[2]}" "ppocrv6_dict.txt"
+download_file "${FILES[3]}" "PP-LCNet_x0_25_textline_ori_infer.onnx"
+
+echo ""
+echo "==> 下载完成！模型文件位于: $MODEL_DIR"
+echo "    PaddleOCR.js PP-OCRv6_small 模型已就绪，重启知识库索引即可使用。"

package/scripts/render_pdf_pages.py

@@ -0,0 +1,29 @@
+"""PDF 页面渲染——用 PyMuPDF 提取高保真页面图片。
+用法: python3 render_pdf_pages.py <pdf_path> <output_dir> [dpi=300]
+输出: <output_dir>/page-<n>.png"""
+import sys, os
+import fitz
+
+def main():
+    if len(sys.argv) < 3:
+        sys.exit(1)
+
+    pdf_path = sys.argv[1]
+    output_dir = sys.argv[2]
+    dpi = int(sys.argv[3]) if len(sys.argv) > 3 else 300
+
+    os.makedirs(output_dir, exist_ok=True)
+    doc = fitz.open(pdf_path)
+
+    num_pages = doc.page_count
+    for i in range(num_pages):
+        pix = doc[i].get_pixmap(dpi=dpi)
+        out = os.path.join(output_dir, f"page-{i + 1}.png")
+        pix.save(out)
+
+    doc.close()
+    print(f"OK:{num_pages}")
+
+
+if __name__ == '__main__':
+    main()