{"id":55859,"date":"2025-02-19T13:17:38","date_gmt":"2025-02-19T05:17:38","guid":{"rendered":"https:\/\/fwq.ai\/blog\/55859\/"},"modified":"2025-02-19T13:17:38","modified_gmt":"2025-02-19T05:17:38","slug":"%e5%ae%9e%e8%b7%b5%ef%bc%9a%e4%bd%bf%e7%94%a8-voyager-3-%e5%92%8c-langgraph-%e6%9e%84%e5%bb%ba%e5%bc%ba%e5%a4%a7%e7%9a%84%e5%a4%9a%e6%a8%a1%e6%80%81%e6%90%9c%e7%b4%a2","status":"publish","type":"post","link":"https:\/\/fwq.ai\/blog\/55859\/","title":{"rendered":"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22"},"content":{"rendered":"<p>Voyage AI \u7684 Voyager 3 \u662f\u4e00\u79cd\u65b0\u7684\u6700\u5148\u8fdb\u7684\u6a21\u578b\uff0c\u5b83\u5141\u8bb8\u60a8\u5c06\u6587\u672c\u548c\u56fe\u50cf\u5d4c\u5165\u5230\u540c\u4e00\u7a7a\u95f4\u4e2d\u3002\u5728\u8fd9\u7bc7\u6587\u7ae0\u4e2d\uff0c\u6211\u5c06\u89e3\u91ca\u5982\u4f55\u4ece\u6742\u5fd7\u4e2d\u63d0\u53d6\u8fd9\u4e9b\u591a\u6a21\u6001\u5d4c\u5165\uff0c\u5c06\u5b83\u4eec\u5b58\u50a8\u5728\u5411\u91cf\u6570\u636e\u5e93\uff08Weaviate\uff09\u4e2d\uff0c\u5e76\u4f7f\u7528\u76f8\u540c\u7684\u5d4c\u5165\u5411\u91cf\u5bf9\u6587\u672c\u548c\u56fe\u50cf\u6267\u884c\u76f8\u4f3c\u6027\u641c\u7d22\u3002<\/p>\n<p>&nbsp;<\/p>\n<p> <\/p>\n<p>\u5c06\u56fe\u50cf\u548c\u6587\u672c\u5d4c\u5165\u5230\u540c\u4e00\u7a7a\u95f4\u4e2d\uff0c\u5c06\u4f7f\u6211\u4eec\u80fd\u591f\u5bf9\u591a\u6a21\u6001\u5185\u5bb9\uff08\u5982\u7f51\u9875\u3001PDF \u6587\u4ef6\u3001\u6742\u5fd7\u3001\u4e66\u7c4d\u3001\u5ba3\u4f20\u518c\u548c\u5404\u79cd\u8bba\u6587\uff09\u6267\u884c\u9ad8\u5ea6\u7cbe\u786e\u7684\u641c\u7d22\u3002\u4e3a\u4ec0\u4e48\u8fd9\u79cd\u6280\u672f\u5982\u6b64\u6709\u8da3\uff1f\u5c06\u6587\u672c\u548c\u56fe\u50cf\u5d4c\u5165\u5230\u540c\u4e00\u7a7a\u95f4\u7684\u4e3b\u8981\u4ee4\u4eba\u5174\u594b\u4e4b\u5904\u5728\u4e8e\uff0c\u60a8\u53ef\u4ee5\u641c\u7d22\u548c\u68c0\u7d22\u4e0e\u7279\u5b9a\u56fe\u50cf\u76f8\u5173\u7684\u6587\u672c\uff0c\u53cd\u4e4b\u4ea6\u7136\u3002\u4f8b\u5982\uff0c\u5982\u679c\u60a8\u6b63\u5728\u641c\u7d22\u732b\uff0c\u60a8\u5c06\u627e\u5230\u663e\u793a\u732b\u7684\u56fe\u7247\uff0c\u4f46\u60a8\u4e5f\u4f1a\u5f97\u5230\u5f15\u7528\u8fd9\u4e9b\u56fe\u50cf\u7684\u6587\u672c\uff0c\u5373\u4f7f\u6587\u672c\u6ca1\u6709\u660e\u786e\u5730\u8bf4\u51fa\u201c\u732b\u201d\u8fd9\u4e2a\u8bcd\u3002<\/p>\n<p>\u8ba9\u6211\u5c55\u793a\u4e00\u4e0b\u4f20\u7edf\u7684\u6587\u672c\u5d4c\u5165\u76f8\u4f3c\u6027\u641c\u7d22\u548c\u591a\u6a21\u6001\u5d4c\u5165\u7a7a\u95f4\u4e4b\u95f4\u7684\u533a\u522b\uff1a<\/p>\n<h2>\u793a\u4f8b\u95ee\u9898\uff1a\u6742\u5fd7\u4e0a\u5173\u4e8e\u732b\u8bf4\u4e86\u4ec0\u4e48\uff1f<\/h2>\n<p><img decoding=\"async\" class=\"aligncenter\" src=\"\/\/www.w3.org\/2000\/svg'%20viewBox='0%200%20720%20542'%3E%3C\/svg%3E\" title=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe\" alt=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe\" \/><br \/>\n<img decoding=\"async\" class=\"aligncenter\" src=\"https:\/\/www.aisharenet.com\/wp-content\/uploads\/2025\/01\/3651742ec3c7e8e.jpg\" title=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe1\" alt=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe1\" \/> <\/p>\n<p>\u4e00\u5f20\u6765\u81ea\u6444\u5f71\u6742\u5fd7\u7684\u622a\u56fe \u2014\u2014 OUTDOOR<\/p>\n<p>&nbsp;<\/p>\n<p><strong>\u5e38\u89c4\u76f8\u4f3c\u6027\u641c\u7d22\u7b54\u6848<\/strong><\/p>\n<p>\u63d0\u4f9b\u7684\u641c\u7d22\u7ed3\u679c\u4e0d\u5305\u542b\u5173\u4e8e\u732b\u7684\u5177\u4f53\u4fe1\u606f\u3002\u5b83\u4eec\u63d0\u5230\u4e86\u52a8\u7269\u8096\u50cf\u548c\u6444\u5f71\u6280\u5de7\uff0c\u4f46\u6ca1\u6709\u660e\u786e\u63d0\u53ca\u732b\u6216\u4e0e\u5176\u76f8\u5173\u7684\u7ec6\u8282\u3002<\/p>\n<p>\u5982\u4e0a\u56fe\u6240\u793a\uff0c\u201c\u732b\u201d\u8fd9\u4e2a\u8bcd\u6ca1\u6709\u88ab\u63d0\u53ca\uff1b\u53ea\u6709\u4e00\u5f20\u56fe\u7247\u548c\u5173\u4e8e\u5982\u4f55\u62cd\u6444\u52a8\u7269\u7167\u7247\u7684\u89e3\u91ca\u3002\u7531\u4e8e\u6ca1\u6709\u63d0\u5230\u201c\u732b\u201d\u8fd9\u4e2a\u8bcd\uff0c\u5e38\u89c4\u7684\u76f8\u4f3c\u6027\u641c\u7d22\u6ca1\u6709\u4ea7\u751f\u4efb\u4f55\u7ed3\u679c\u3002<\/p>\n<p><strong>\u591a\u6a21\u6001\u641c\u7d22\u7b54\u6848<\/strong><\/p>\n<p>\u8fd9\u672c\u6742\u5fd7\u520a\u767b\u4e86\u4e00\u5f20\u732b\u7684\u8096\u50cf\uff0c\u7a81\u51fa\u4e86\u5176\u9762\u90e8\u7279\u5f81\u548c\u6027\u683c\u7684\u7cbe\u7ec6\u6355\u6349\u3002\u6587\u5b57\u5f3a\u8c03\u4e86\u5236\u4f5c\u7cbe\u826f\u7684\u52a8\u7269\u8096\u50cf\u5982\u4f55\u6df1\u5165\u5230\u4e3b\u9898\u7684\u7075\u9b42\uff0c\u5e76\u901a\u8fc7\u5f15\u4eba\u6ce8\u76ee\u7684\u773c\u795e\u4ea4\u6d41\u4e0e\u89c2\u770b\u8005\u5efa\u7acb\u60c5\u611f\u8054\u7cfb\u3002<\/p>\n<p>\u4f7f\u7528\u591a\u6a21\u6001\u641c\u7d22\uff0c\u6211\u4eec\u5c06\u627e\u5230\u4e00\u5f20\u732b\u7684\u56fe\u7247\uff0c\u7136\u540e\u5c06\u76f8\u5173\u7684\u6587\u5b57\u94fe\u63a5\u5230\u5b83\u3002\u5c06\u8fd9\u4e9b\u6570\u636e\u63d0\u4f9b\u7ed9\u6a21\u578b\u5c06\u4f7f\u5176\u80fd\u591f\u66f4\u597d\u5730\u56de\u7b54\u548c\u7406\u89e3\u4e0a\u4e0b\u6587\u3002<\/p>\n<p>&nbsp;<\/p>\n<h2>\u5982\u4f55\u6784\u5efa\u591a\u6a21\u6001\u5d4c\u5165\u548c\u68c0\u7d22\u7ba1\u9053<\/h2>\n<p>\u73b0\u5728\uff0c\u6211\u5c06\u5206\u51e0\u4e2a\u6b65\u9aa4\u63cf\u8ff0\u8fd9\u6837\u4e00\u4e2a\u7ba1\u9053\u7684\u5de5\u4f5c\u539f\u7406\uff1a<\/p>\n<ol>\n<li>\u6211\u4eec\u5c06\u4f7f\u7528&nbsp;\uff08\u4e00\u4e2a\u7528\u4e8e\u6570\u636e\u63d0\u53d6\u7684\u5f3a\u5927 Python \u5e93\uff09\u4ece PDF \u6587\u4ef6\u4e2d\u63d0\u53d6\u6587\u672c\u548c\u56fe\u50cf\u3002<\/li>\n<li>\u6211\u4eec\u5c06\u4f7f\u7528&nbsp;&nbsp;\u6a21\u578b\u4e3a\u540c\u4e00\u5411\u91cf\u7a7a\u95f4\u5185\u7684\u6587\u672c\u548c\u56fe\u50cf\u521b\u5efa\u591a\u6a21\u6001\u5411\u91cf\u3002<\/li>\n<li>\u6211\u4eec\u5c06\u628a\u5b83\u63d2\u5165\u5230\u5411\u91cf\u5b58\u50a8 () \u4e2d\u3002<\/li>\n<li>\u6700\u540e\uff0c\u6211\u4eec\u5c06\u6267\u884c\u76f8\u4f3c\u6027\u641c\u7d22\u5e76\u6bd4\u8f83\u6587\u672c\u548c\u56fe\u50cf\u7684\u7ed3\u679c\u3002<\/li>\n<\/ol>\n<h3>\u7b2c 1 \u6b65\uff1a\u8bbe\u7f6e\u5411\u91cf\u5b58\u50a8\u5e76\u4ece\u6587\u4ef6 (PDF) \u4e2d\u63d0\u53d6\u56fe\u50cf\u548c\u6587\u672c<\/h3>\n<p>\u5728\u8fd9\u91cc\uff0c\u6211\u4eec\u5fc5\u987b\u505a\u4e00\u4e9b\u624b\u52a8\u5de5\u4f5c\u3002\u901a\u5e38\uff0cWeaviate \u662f\u4e00\u4e2a\u975e\u5e38\u6613\u4e8e\u4f7f\u7528\u7684\u5411\u91cf\u5b58\u50a8\uff0c\u5b83\u4f1a\u5728\u63d2\u5165\u65f6\u81ea\u52a8\u8f6c\u6362\u6570\u636e\u5e76\u6dfb\u52a0\u5d4c\u5165\u3002\u4f46\u662f\uff0c\u6ca1\u6709\u7528\u4e8e Voyager Multimodal v3 \u7684\u63d2\u4ef6\uff0c\u56e0\u6b64\u6211\u4eec\u5fc5\u987b\u624b\u52a8\u8ba1\u7b97\u5d4c\u5165\u3002<strong>\u5728\u8fd9\u79cd\u60c5\u51b5\u4e0b\uff0c\u6211\u4eec\u5fc5\u987b\u521b\u5efa\u4e00\u4e2a\u96c6\u5408\u800c\u4e0d\u5b9a\u4e49\u5411\u91cf\u5316\u5668\u6a21\u5757\u3002<\/strong><\/p>\n<pre><code>import weaviate\r\nfrom weaviate.classes.config import Configure\r\nclient = weaviate.connect_to_local()\r\ncollection_name = \"multimodal_demo\"\r\nclient.collections.delete(collection_name)\r\ntry:\r\nclient.collections.create(\r\nname=collection_name,\r\nvectorizer_config=Configure.Vectorizer.none() # \u4e0d\u4e3a\u6b64\u96c6\u5408\u8bbe\u7f6e\u5411\u91cf\u5316\u5668\r\n)\r\ncollection = client.collections.get(collection_name)\r\nexcept Exception:\r\ncollection = client.collections.get(collection_name)pyt\r\n<\/code><\/pre>\n<p>\u5728\u8fd9\u91cc\uff0c\u6211\u5728 Docker \u5bb9\u5668\u4e2d\u8fd0\u884c\u4e00\u4e2a\u672c\u5730 Weaviate \u5b9e\u4f8b\u3002<\/p>\n<h3>\u7b2c 2 \u6b65\uff1a<strong>\u4ece PDF \u4e2d\u63d0\u53d6\u6587\u6863\u548c\u56fe\u50cf<\/strong><\/h3>\n<p>\u8fd9\u662f\u6d41\u7a0b\u5de5\u4f5c\u7684\u5173\u952e\u6b65\u9aa4\u3002\u5728\u8fd9\u91cc\uff0c\u6211\u4eec\u5c06\u83b7\u53d6\u4e00\u4e2a\u5305\u542b\u6587\u672c\u548c\u56fe\u7247\u7684 PDF\u3002\u7136\u540e\uff0c\u6211\u4eec\u5c06\u63d0\u53d6\u5185\u5bb9\uff08\u56fe\u50cf\u548c\u6587\u672c\uff09\u5e76\u5c06\u5176\u5b58\u50a8\u5728\u76f8\u5173\u7684\u5757\u4e2d\u3002\u56e0\u6b64\uff0c\u6bcf\u4e2a\u5757\u5c06\u662f\u4e00\u4e2a\u5305\u542b\u5b57\u7b26\u4e32\uff08\u5b9e\u9645\u6587\u672c\uff09\u548c&nbsp;&nbsp;\u7684\u5143\u7d20\u5217\u8868\u3002<\/p>\n<p>\u6211\u4eec\u5c06\u4f7f\u7528  \u5e93\u6765\u5b8c\u6210\u4e00\u4e9b\u7e41\u91cd\u7684\u5de5\u4f5c\uff0c\u4f46\u6211\u4eec\u4ecd\u7136\u9700\u8981\u7f16\u5199\u4e00\u4e9b\u903b\u8f91\u5e76\u914d\u7f6e\u5e93\u53c2\u6570\u3002<\/p>\n<pre><code>from unstructured.partition.auto import partition\r\nfrom unstructured.chunking.title import chunk_by_title\r\nelements = partition(\r\nfilename=\".\/files\/magazine_sample.pdf\",\r\nstrategy=\"hi_res\",\r\nextract_image_block_types=[\"Image\", \"Table\"],\r\nextract_image_block_to_payload=True)\r\nchunks = chunk_by_title(elements)\r\n<\/code><\/pre>\n<p>\u5728\u8fd9\u91cc\uff0c\u6211\u4eec\u5fc5\u987b\u4f7f\u7528&nbsp;<strong>hi_res<\/strong>&nbsp;\u7b56\u7565\uff0c\u5e76\u4f7f\u7528&nbsp;<strong>extract_image_block_to_payload<\/strong>&nbsp;\u5c06\u56fe\u50cf\u5bfc\u51fa\u5230\u6709\u6548\u8f7d\u8377\uff0c\u56e0\u4e3a\u6211\u4eec\u7a0d\u540e\u9700\u8981\u6b64\u4fe1\u606f\u7528\u4e8e\u5b9e\u9645\u7684\u5d4c\u5165\u3002\u4e00\u65e6\u6211\u4eec\u63d0\u53d6\u4e86\u6240\u6709\u5143\u7d20\uff0c\u6211\u4eec\u5c06\u6839\u636e\u6587\u6863\u4e2d\u7684\u6807\u9898\u5c06\u5b83\u4eec\u5206\u7ec4\u5230\u5757\u4e2d\u3002<\/p>\n<p><em>\u6709\u5173\u66f4\u591a\u4fe1\u606f\uff0c\u8bf7\u67e5\u770b&nbsp;\u3002<\/em><\/p>\n<p>\u5728\u4e0b\u9762\u7684\u811a\u672c\u4e2d\uff0c\u6211\u4eec\u5c06\u4f7f\u7528\u8fd9\u4e9b\u5757\u8f93\u51fa\u4e24\u4e2a\u5217\u8868\uff1a<\/p>\n<ol>\n<li>\u4e00\u4e2a\u5305\u542b\u6211\u4eec\u5c06\u53d1\u9001\u5230 Voyager 3 \u4ee5\u521b\u5efa\u5411\u91cf\u7684\u5bf9\u8c61\u7684\u5217\u8868<\/li>\n<li>\u4e00\u4e2a\u5305\u542b Unstructured \u63d0\u53d6\u7684\u5143\u6570\u636e\u7684\u5217\u8868\u3002\u6b64\u5143\u6570\u636e\u662f\u5fc5\u9700\u7684\uff0c\u56e0\u4e3a\u6211\u4eec\u5fc5\u987b\u5c06\u5176\u6dfb\u52a0\u5230\u5411\u91cf\u5b58\u50a8\u4e2d\u3002\u5b83\u5c06\u4e3a\u6211\u4eec\u63d0\u4f9b\u989d\u5916\u7684\u5c5e\u6027\u8fdb\u884c\u8fc7\u6ee4\uff0c\u5e76\u544a\u8bc9\u6211\u4eec\u4e00\u4e9b\u5173\u4e8e\u68c0\u7d22\u5230\u7684\u6570\u636e\u7684\u4fe1\u606f\u3002<\/li>\n<\/ol>\n<pre><code>from unstructured.staging.base import elements_from_base64_gzipped_json\r\nimport PIL.Image\r\nimport io\r\nimport base64\r\nembedding_objects = []\r\nembedding_metadatas = []\r\nfor chunk in chunks:\r\nembedding_object = []\r\nmetedata_dict = {\r\n\"text\": chunk.to_dict()[\"text\"],\r\n\"filename\": chunk.to_dict()[\"metadata\"][\"filename\"],\r\n\"page_number\": chunk.to_dict()[\"metadata\"][\"page_number\"],\r\n\"last_modified\": chunk.to_dict()[\"metadata\"][\"last_modified\"],\r\n\"languages\": chunk.to_dict()[\"metadata\"][\"languages\"],\r\n\"filetype\": chunk.to_dict()[\"metadata\"][\"filetype\"]\r\n}\r\nembedding_object.append(chunk.to_dict()[\"text\"])\r\n# \u5c06\u56fe\u50cf\u6dfb\u52a0\u5230\u5d4c\u5165\u5bf9\u8c61\r\nif \"orig_elements\" in chunk.to_dict()[\"metadata\"]:\r\nbase64_elements_str = chunk.to_dict()[\"metadata\"][\"orig_elements\"]\r\neles = elements_from_base64_gzipped_json(base64_elements_str)\r\nimage_data = []\r\nfor ele in eles:\r\nif ele.to_dict()[\"type\"] == \"Image\":\r\nbase64_image = ele.to_dict()[\"metadata\"][\"image_base64\"]\r\nimage_data.append(base64_image)\r\npil_image = PIL.Image.open(io.BytesIO(base64.b64decode(base64_image)))\r\n# \u5982\u679c\u56fe\u50cf\u5927\u4e8e 1000x1000\uff0c\u5219\u5728\u4fdd\u6301\u7eb5\u6a2a\u6bd4\u7684\u540c\u65f6\u8c03\u6574\u56fe\u50cf\u5927\u5c0f\r\nif pil_image.size[0] &gt; 1000 or pil_image.size[1] &gt; 1000:\r\nratio = min(1000\/pil_image.size[0], 1000\/pil_image.size[1])\r\nnew_size = (int(pil_image.size[0] * ratio), int(pil_image.size[1] * ratio))\r\npil_image = pil_image.resize(new_size, PIL.Image.Resampling.LANCZOS)\r\nembedding_object.append(pil_image)\r\nmetedata_dict[\"image_data\"] = image_data\r\nembedding_objects.append(embedding_object)\r\nembedding_metadatas.append(metedata_dict)\r\n<\/code><\/pre>\n<p>\u6b64\u811a\u672c\u7684\u7ed3\u679c\u5c06\u662f\u4e00\u4e2a\u5217\u8868\u7684\u5217\u8868\uff0c\u5176\u5185\u5bb9\u5982\u4e0b\u6240\u793a\uff1a<\/p>\n<pre><code>[['\u6765\u81ea\\n\\n\u51b0\u5c9b KIRKJUFELL \u7684\u4f4d\u7f6e',\r\n&lt;PIL.Image.Image image mode=RGB size=1000x381&gt;,\r\n&lt;PIL.Image.Image image mode=RGB size=526x1000&gt;],\r\n['\u8fd9\u5ea7\u6807\u5fd7\u6027\u7684\u5c71\u5cf0\u662f\u6211\u4eec\u51b0\u5c9b\u62cd\u6444\u5730\u70b9\u7684\u9996\u9009\uff0c\u800c\u4e14\u5728\u6211\u4eec\u53bb\u90a3\u91cc\u4e4b\u524d\uff0c\u6211\u4eec\u5c31\u770b\u8fc7\u8bb8\u591a\u4ece\u9644\u8fd1\u7011\u5e03\u62cd\u6444\u7684\u7167\u7247\u3002\u56e0\u6b64\uff0c\u8fd9\u662f\u6211\u4eec\u5728\u65e5\u51fa\u65f6\u524d\u5f80\u7684\u7b2c\u4e00\u4e2a\u5730\u65b9 - \u6211\u4eec\u6ca1\u6709\u5931\u671b\u3002\u8fd9\u4e9b\u7011\u5e03\u4e3a\u8fd9\u5f20\u7167\u7247\uff08\u9876\u90e8\uff09\u63d0\u4f9b\u4e86\u5b8c\u7f8e\u7684\u8fd1\u666f\u8da3\u5473\uff0c\u800c\u4ece\u8fd9\u4e2a\u89d2\u5ea6\u6765\u770b\uff0cKirkjufell \u662f\u4e00\u5ea7\u5b8c\u7f8e\u7684\u5c16\u5c71\u3002\u6211\u4eec\u82b1\u4e86\u4e00\u4e24\u4e2a\u5c0f\u65f6\u7b80\u5355\u5730\u63a2\u7d22\u8fd9\u4e9b\u7011\u5e03\uff0c\u627e\u5230\u4e86\u51e0\u4e2a\u4e0d\u540c\u7684\u89d2\u5ea6\u3002']]\r\n<\/code><\/pre>\n<h3>\u7b2c 3 \u6b65\uff1a\u5411\u91cf\u5316\u63d0\u53d6\u7684\u6570\u636e<\/h3>\n<p>\u5728\u8fd9\u4e00\u6b65\u4e2d\uff0c\u6211\u4eec\u5c06\u4f7f\u7528\u4e0a\u4e00\u6b65\u4e2d\u521b\u5efa\u7684\u5757\uff0c\u5e76\u4f7f\u7528&nbsp;&nbsp;\u5c06\u5b83\u4eec\u53d1\u9001\u5230 Voyager\u3002\u5b83\u5c06\u8fd4\u56de\u7ed9\u6211\u4eec\u6240\u6709\u5d4c\u5165\u5bf9\u8c61\u7684\u5217\u8868\u3002\u7136\u540e\uff0c\u6211\u4eec\u53ef\u4ee5\u4f7f\u7528\u6b64\u7ed3\u679c\uff0c\u5e76\u6700\u7ec8\u5c06\u5176\u5b58\u50a8\u5728 Weaviate \u4e2d\u3002<\/p>\n<pre><code>from dotenv import load_dotenv\r\nimport voyageai\r\nload_dotenv()\r\nvo = voyageai.Client()\r\n# \u8fd9\u5c06\u81ea\u52a8\u4f7f\u7528\u73af\u5883\u53d8\u91cf VOYAGE_API_KEY\u3002\r\n# \u6216\u8005\uff0c\u60a8\u53ef\u4ee5\u4f7f\u7528 vo = voyageai.Client(api_key=\"&lt;\u60a8\u7684\u5bc6\u94a5&gt;\")\r\n# \u5305\u542b\u6587\u672c\u5b57\u7b26\u4e32\u548c PIL \u56fe\u50cf\u5bf9\u8c61\u7684\u793a\u4f8b\u8f93\u5165\r\ninputs = embedding_objects\r\n# \u5411\u91cf\u5316\u8f93\u5165\r\nresult = vo.multimodal_embed(\r\ninputs,\r\nmodel=\"voyage-multimodal-3\",\r\ntruncation=False\r\n)\r\n<\/code><\/pre>\n<p>\u5982\u679c\u6211\u4eec\u8bbf\u95ee result.embeddings\uff0c\u6211\u4eec\u5c06\u83b7\u5f97\u4e00\u4e2a\u5305\u542b\u6240\u6709\u8ba1\u7b97\u51fa\u7684\u5d4c\u5165\u5411\u91cf\u7684\u5217\u8868\u7684\u5217\u8868\uff1a<\/p>\n<p> [[-0.052734375, -0.0164794921875, 0.050048828125, 0.01348876953125, -0.048095703125, \u2026]] <\/p>\n<p>\u6211\u4eec\u73b0\u5728\u53ef\u4ee5\u4f7f\u7528&nbsp;<code>batch.add_object<\/code>&nbsp;\u65b9\u6cd5\u5c06\u6b64\u5d4c\u5165\u6570\u636e\u4ee5\u5355\u4e2a\u6279\u6b21\u5b58\u50a8\u5728 Weaviate \u4e2d\u3002\u8bf7\u6ce8\u610f\uff0c\u6211\u4eec\u8fd8\u5728 properties \u53c2\u6570\u4e2d\u6dfb\u52a0\u4e86\u5143\u6570\u636e\u3002<\/p>\n<pre><code>with collection.batch.dynamic() as batch:\r\nfor i, data_row in enumerate(embedding_objects):\r\nbatch.add_object(\r\nproperties=embedding_metadatas[i],\r\nvector=result.embeddings[i]\r\n)\r\n<\/code><\/pre>\n<h3>\u7b2c 4 \u6b65\uff1a\u67e5\u8be2\u6570\u636e<\/h3>\n<p>\u6211\u4eec\u73b0\u5728\u53ef\u4ee5\u6267\u884c\u76f8\u4f3c\u6027\u641c\u7d22\u5e76\u67e5\u8be2\u6570\u636e\u3002\u8fd9\u5f88\u5bb9\u6613\uff0c\u56e0\u4e3a\u6b64\u6d41\u7a0b\u7c7b\u4f3c\u4e8e\u5bf9\u6587\u672c\u5d4c\u5165\u6267\u884c\u7684\u5e38\u89c4\u76f8\u4f3c\u6027\u641c\u7d22\u3002\u7531\u4e8e Weaviate \u6ca1\u6709\u7528\u4e8e Voyager \u591a\u6a21\u6001\u7684\u6a21\u5757\uff0c\u56e0\u6b64\u6211\u4eec\u5fc5\u987b\u5728\u5c06\u641c\u7d22\u5411\u91cf\u4f20\u9012\u7ed9 Weaviate \u4ee5\u6267\u884c\u76f8\u4f3c\u6027\u641c\u7d22\u4e4b\u524d\uff0c\u81ea\u5df1\u8ba1\u7b97\u641c\u7d22\u67e5\u8be2\u7684\u5411\u91cf\u3002<\/p>\n<pre><code>from weaviate.classes.query import MetadataQuery\r\nquestion = \"\u6742\u5fd7\u4e0a\u5173\u4e8e\u7011\u5e03\u8bf4\u4e86\u4ec0\u4e48\uff1f\"\r\nvector = vo.multimodal_embed([[question]], model=\"voyage-multimodal-3\")\r\nvector.embeddings[0]\r\nresponse = collection.query.near_vector(\r\nnear_vector=vector.embeddings[0], # \u60a8\u7684\u67e5\u8be2\u5411\u91cf\u5728\u6b64\u5904\r\nlimit=2,\r\nreturn_metadata=MetadataQuery(distance=True)\r\n)\r\n# \u663e\u793a\u7ed3\u679c\r\nfor o in response.objects:\r\nprint(o.properties['text'])\r\nfor image_data in o.properties['image_data']:\r\n# \u4f7f\u7528 PIL \u663e\u793a\u56fe\u50cf\r\nimg = PIL.Image.open(io.BytesIO(base64.b64decode(image_data)))\r\nwidth, height = img.size\r\nif width &gt; 500 or height &gt; 500:\r\nratio = min(500\/width, 500\/height)\r\nnew_size = (int(width * ratio), int(height * ratio))\r\nimg = img.resize(new_size)\r\ndisplay(img)\r\nprint(o.metadata.distance)\r\n<\/code><\/pre>\n<p>\u4e0b\u56fe\u663e\u793a\uff0c\u641c\u7d22\u7011\u5e03\u5c06\u8fd4\u56de\u4e0e\u6b64\u641c\u7d22\u67e5\u8be2\u76f8\u5173\u7684\u6587\u672c\u548c\u56fe\u50cf\u3002\u5982\u60a8\u6240\u89c1\uff0c\u8fd9\u4e9b\u7167\u7247\u53cd\u6620\u4e86\u7011\u5e03\uff0c\u4f46\u6587\u672c\u672c\u8eab\u5e76\u6ca1\u6709\u63d0\u53ca\u5b83\u4eec\u3002\u8fd9\u6bb5\u6587\u5b57\u662f\u5173\u4e8e\u4e00\u5f20\u91cc\u9762\u6709\u7011\u5e03\u7684\u56fe\u7247\uff0c\u8fd9\u5c31\u662f\u5b83\u4e5f\u88ab\u68c0\u7d22\u7684\u539f\u56e0\u3002\u8fd9\u5bf9\u4e8e\u5e38\u89c4\u6587\u672c\u5d4c\u5165\u641c\u7d22\u662f\u4e0d\u53ef\u80fd\u7684\u3002<\/p>\n<p><img decoding=\"async\" class=\"aligncenter\" src=\"\/\/www.w3.org\/2000\/svg'%20viewBox='0%200%20720%20671'%3E%3C\/svg%3E\" title=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe2\" alt=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe2\" \/><br \/>\n<img decoding=\"async\" class=\"aligncenter\" src=\"https:\/\/www.aisharenet.com\/wp-content\/uploads\/2025\/01\/df4dfdea4cd2e81.jpg\" title=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe3\" alt=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe3\" \/> <\/p>\n<p>\u4e00\u5f20\u663e\u793a\u76f8\u4f3c\u6027\u641c\u7d22\u7ed3\u679c\u7684\u56fe\u7247<\/p>\n<p>&nbsp;<\/p>\n<h3>\u7b2c 5 \u6b65\uff1a\u5c06\u5176\u6dfb\u52a0\u5230\u6574\u4e2a\u68c0\u7d22\u7ba1\u9053\u4e2d<\/h3>\n<p>\u73b0\u5728\u6211\u4eec\u5df2\u7ecf\u4ece\u6742\u5fd7\u4e2d\u63d0\u53d6\u4e86\u6587\u672c\u548c\u56fe\u50cf\uff0c\u4e3a\u5b83\u4eec\u521b\u5efa\u4e86\u5d4c\u5165\uff0c\u5c06\u5b83\u4eec\u6dfb\u52a0\u5230 Weaviate \u4e2d\uff0c\u5e76\u8bbe\u7f6e\u4e86\u6211\u4eec\u7684\u76f8\u4f3c\u6027\u641c\u7d22\uff0c\u6211\u5c06\u628a\u5b83\u6dfb\u52a0\u5230\u6574\u4e2a\u68c0\u7d22\u7ba1\u9053\u4e2d\u3002\u5728\u6b64\u793a\u4f8b\u4e2d\uff0c\u6211\u5c06\u4f7f\u7528 LangGraph\u3002\u7528\u6237\u5c06\u63d0\u51fa\u5173\u4e8e\u8fd9\u672c\u6742\u5fd7\u7684\u95ee\u9898\uff0c\u7ba1\u9053\u5c06\u56de\u7b54\u8fd9\u4e2a\u95ee\u9898\u3002\u65e2\u7136\u6240\u6709\u5de5\u4f5c\u90fd\u5df2\u5b8c\u6210\uff0c\u8fd9\u90e8\u5206\u5c31\u50cf\u4f7f\u7528\u5e38\u89c4\u6587\u672c\u8bbe\u7f6e\u5178\u578b\u7684\u68c0\u7d22\u7ba1\u9053\u4e00\u6837\u7b80\u5355\u3002<\/p>\n<p>\u6211\u5df2\u7ecf\u5c06\u6211\u4eec\u5728\u524d\u9762\u90e8\u5206\u8ba8\u8bba\u7684\u4e00\u4e9b\u903b\u8f91\u62bd\u8c61\u5230\u5176\u4ed6\u6a21\u5757\u4e2d\u3002\u8fd9\u662f\u4e00\u4e2a\u6211\u5982\u4f55\u5c06\u5176\u96c6\u6210\u5230  \u7ba1\u9053\u4e2d\u7684\u793a\u4f8b\u3002<\/p>\n<pre><code>class MultiModalRetrievalState(TypedDict):\r\nmessages: Annotated[Sequence[BaseMessage], add_messages]\r\nresults: List[Document]\r\nbase_64_images: List[str]\r\nclass RAGNodes(BaseNodes):\r\ndef __init__(self, logger, mode=\"online\", document_handler=None):\r\nsuper().__init__(logger, mode)\r\nself.weaviate = Weaviate()\r\nself.mode = mode\r\nasync def multi_modal_retrieval(self, state: MultiModalRetrievalState, config):\r\ncollection_name = config.get(\"configurable\", {}).get(\"collection_name\")\r\nself.weaviate.set_collection(collection_name)\r\nprint(\"\u6b63\u5728\u8fd0\u884c\u591a\u6a21\u6001\u68c0\u7d22\")\r\nprint(f\"\u6b63\u5728\u641c\u7d22 {state['messages'][-1].content}\")\r\nresults = self.weaviate.similarity_search(\r\nquery=state[\"messages\"][-1].content, k=3, type=\"multimodal\"\r\n)\r\nreturn {\"results\": results}\r\nasync def answer_question(self, state: MultiModalRetrievalState, config):\r\nprint(\"\u6b63\u5728\u56de\u7b54\u95ee\u9898\")\r\nllm = self.llm_factory.create_llm(mode=self.mode, model_type=\"default\")\r\ninclude_images = config.get(\"configurable\", {}).get(\"include_images\", False)\r\nchain = self.chain_factory.create_multi_modal_chain(\r\nllm,\r\nstate[\"messages\"][-1].content,\r\nstate[\"results\"],\r\ninclude_images=include_images,\r\n)\r\nresponse = await chain.ainvoke({})\r\nmessage = AIMessage(content=response)\r\nreturn {\"messages\": message}\r\n# \u5b9a\u4e49\u914d\u7f6e\r\nclass GraphConfig(TypedDict):\r\nmode: str = \"online\"\r\ncollection_name: str\r\ninclude_images: bool = False\r\ngraph_nodes = RAGNodes(logger)\r\ngraph = StateGraph(MultiModalRetrievalState, config_schema=GraphConfig)\r\ngraph.add_node(\"multi_modal_retrieval\", graph_nodes.multi_modal_retrieval)\r\ngraph.add_node(\"answer_question\", graph_nodes.answer_question)\r\ngraph.add_edge(START, \"multi_modal_retrieval\")\r\ngraph.add_edge(\"multi_modal_retrieval\", \"answer_question\")\r\ngraph.add_edge(\"answer_question\", END)\r\nmulti_modal_graph = graph.compile()\r\n__all__ = [\"multi_modal_graph\"]\r\n<\/code><\/pre>\n<p>\u4e0a\u9762\u7684\u4ee3\u7801\u5c06\u751f\u6210\u4ee5\u4e0b\u56fe\u8868<\/p>\n<p><img decoding=\"async\" class=\"aligncenter\" src=\"\/\/www.w3.org\/2000\/svg'%20viewBox='0%200%20448%20543'%3E%3C\/svg%3E\" title=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe4\" alt=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe4\" \/><br \/>\n<img decoding=\"async\" class=\"aligncenter\" src=\"https:\/\/www.aisharenet.com\/wp-content\/uploads\/2025\/01\/afb95d54915c6e5.jpg\" title=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe5\" alt=\"\u5b9e\u8df5\uff1a\u4f7f\u7528 Voyager-3 \u548c LangGraph \u6784\u5efa\u5f3a\u5927\u7684\u591a\u6a21\u6001\u641c\u7d22\u63d2\u56fe5\" \/> <\/p>\n<p>\u521b\u5efa\u7684\u56fe\u8868\u7684\u76f4\u89c2\u8868\u793a<\/p>\n<p>&nbsp;<\/p>\n<\/p>\n<p>\u60a8\u53ef\u4ee5\u770b\u5230\u5185\u5bb9\u548c\u56fe\u50cf\u90fd\u88ab\u53d1\u9001\u5230 OpenAI \u4ee5\u56de\u7b54\u95ee\u9898\u3002<\/p>\n<p>&nbsp;<\/p>\n<h2>\u7ed3\u8bba<\/h2>\n<p>\u591a\u6a21\u6001\u5d4c\u5165\u4e3a\u5728\u540c\u4e00\u5d4c\u5165\u7a7a\u95f4\u5185\u96c6\u6210\u548c\u68c0\u7d22\u4e0d\u540c\u6570\u636e\u7c7b\u578b\uff08\u5982\u6587\u672c\u548c\u56fe\u50cf\uff09\u7684\u4fe1\u606f\u5f00\u8f9f\u4e86\u53ef\u80fd\u6027\u3002\u901a\u8fc7\u7ed3\u5408 Voyager Multimodal 3 \u6a21\u578b\u3001Weaviate \u548c LangGraph \u7b49\u5c16\u7aef\u5de5\u5177\uff0c\u6211\u4eec\u5c55\u793a\u4e86\u5982\u4f55\u6784\u5efa\u4e00\u4e2a\u5f3a\u5927\u7684\u68c0\u7d22\u7ba1\u9053\uff0c\u8be5\u7ba1\u9053\u80fd\u591f\u6bd4\u4f20\u7edf\u7684\u7eaf\u6587\u672c\u65b9\u6cd5\u66f4\u76f4\u89c2\u5730\u7406\u89e3\u548c\u94fe\u63a5\u5185\u5bb9\u3002<\/p>\n<p>\u8fd9\u79cd\u65b9\u6cd5\u663e\u7740\u63d0\u9ad8\u4e86\u5bf9\u6742\u5fd7\u3001\u5ba3\u4f20\u518c\u548c PDF \u7b49\u5404\u79cd\u6570\u636e\u6e90\u7684\u641c\u7d22\u548c\u68c0\u7d22\u51c6\u786e\u6027\u3002\u5b83\u8fd8\u6f14\u793a\u4e86\u591a\u6a21\u6001\u5d4c\u5165\u5982\u4f55\u63d0\u4f9b\u66f4\u4e30\u5bcc\u3001\u4e0a\u4e0b\u6587\u611f\u77e5\u7684\u89c1\u89e3\uff0c\u5373\u4f7f\u5728\u6ca1\u6709\u660e\u786e\u5173\u952e\u5b57\u7684\u60c5\u51b5\u4e0b\uff0c\u4e5f\u80fd\u5c06\u56fe\u50cf\u8fde\u63a5\u5230\u63cf\u8ff0\u6027\u6587\u672c\u3002\u672c\u6559\u7a0b\u5141\u8bb8\u60a8\u63a2\u7d22\u5e76\u5c06\u8fd9\u4e9b\u6280\u672f\u5e94\u7528\u4e8e\u60a8\u7684\u9879\u76ee\u3002<\/p>\n<p>\u793a\u4f8b Notebook\uff1ahttps:\/\/github.com\/vectrix-ai\/vectrix-graphs\/blob\/main\/examples\/multi-model-embeddings.ipynb<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Voyage AI \u7684 Voyager 3 \u662f\u4e00\u79cd\u65b0\u7684\u6700\u5148\u8fdb\u7684\u6a21\u578b\uff0c\u5b83\u5141\u8bb8\u60a8\u5c06\u6587\u672c\u548c\u56fe\u50cf\u5d4c\u5165\u5230\u540c\u4e00\u7a7a\u95f4\u4e2d\u3002\u5728\u8fd9\u7bc7\u6587\u7ae0\u4e2d\uff0c\u6211\u5c06\u89e3\u91ca\u5982\u4f55\u4ece\u6742\u5fd7\u4e2d\u63d0\u53d6\u8fd9\u4e9b\u591a\u6a21\u6001\u5d4c\u5165\uff0c\u5c06\u5b83\u4eec\u5b58\u50a8\u5728\u5411\u91cf\u6570\u636e\u5e93\uff08Weaviate\uff09\u4e2d\uff0c\u5e76\u4f7f\u7528\u76f8\u540c\u7684\u5d4c\u5165\u5411\u91cf\u5bf9\u6587\u672c\u548c\u56fe\u50cf\u6267\u884c\u76f8\u4f3c\u6027\u641c\u7d22\u3002 &nbsp; \u5c06\u56fe\u50cf\u548c\u6587\u672c\u5d4c\u5165\u5230\u540c\u4e00\u7a7a\u95f4\u4e2d\uff0c\u5c06\u4f7f\u6211\u4eec\u80fd\u591f\u5bf9\u591a\u6a21\u6001\u5185\u5bb9\uff08\u5982\u7f51\u9875\u3001PDF \u6587\u4ef6\u3001\u6742\u5fd7\u3001\u4e66\u7c4d\u3001\u5ba3\u4f20\u518c\u548c\u5404\u79cd\u8bba\u6587\uff09\u6267\u884c\u9ad8\u5ea6\u7cbe\u786e\u7684\u641c\u7d22\u3002\u4e3a\u4ec0\u4e48\u8fd9\u79cd\u6280\u672f\u5982\u6b64\u6709\u8da3\uff1f\u5c06\u6587\u672c\u548c\u56fe\u50cf\u5d4c\u5165\u5230\u540c\u4e00\u7a7a\u95f4\u7684\u4e3b\u8981\u4ee4\u4eba\u5174\u594b\u4e4b\u5904\u5728\u4e8e\uff0c\u60a8\u53ef\u4ee5\u641c\u7d22\u548c\u68c0\u7d22\u4e0e\u7279\u5b9a\u56fe\u50cf\u76f8\u5173\u7684\u6587\u672c\uff0c\u53cd\u4e4b\u4ea6\u7136\u3002\u4f8b\u5982\uff0c\u5982\u679c\u60a8\u6b63\u5728\u641c\u7d22\u732b\uff0c\u60a8\u5c06\u627e\u5230\u663e\u793a\u732b\u7684\u56fe\u7247\uff0c\u4f46\u60a8\u4e5f\u4f1a\u5f97\u5230\u5f15\u7528\u8fd9\u4e9b\u56fe\u50cf\u7684\u6587\u672c\uff0c\u5373\u4f7f\u6587\u672c\u6ca1\u6709\u660e\u786e\u5730\u8bf4\u51fa\u201c\u732b\u201d\u8fd9\u4e2a\u8bcd\u3002 \u8ba9\u6211\u5c55\u793a\u4e00\u4e0b\u4f20\u7edf\u7684\u6587\u672c\u5d4c\u5165\u76f8\u4f3c\u6027\u641c\u7d22\u548c\u591a\u6a21\u6001\u5d4c\u5165\u7a7a\u95f4\u4e4b\u95f4\u7684\u533a\u522b\uff1a \u793a\u4f8b\u95ee\u9898\uff1a\u6742\u5fd7\u4e0a\u5173\u4e8e\u732b\u8bf4\u4e86\u4ec0\u4e48\uff1f \u4e00\u5f20\u6765\u81ea\u6444\u5f71\u6742\u5fd7\u7684\u622a\u56fe \u2014\u2014 OUTDOOR &nbsp; \u5e38\u89c4\u76f8\u4f3c\u6027\u641c\u7d22\u7b54\u6848 \u63d0\u4f9b\u7684\u641c\u7d22\u7ed3\u679c\u4e0d\u5305\u542b\u5173\u4e8e\u732b\u7684\u5177\u4f53\u4fe1\u606f\u3002\u5b83\u4eec\u63d0\u5230\u4e86\u52a8\u7269\u8096\u50cf\u548c\u6444\u5f71\u6280\u5de7\uff0c\u4f46\u6ca1\u6709\u660e\u786e\u63d0\u53ca\u732b\u6216\u4e0e\u5176\u76f8\u5173\u7684\u7ec6\u8282\u3002 \u5982\u4e0a\u56fe\u6240\u793a\uff0c\u201c\u732b\u201d\u8fd9\u4e2a\u8bcd\u6ca1\u6709\u88ab\u63d0\u53ca\uff1b\u53ea\u6709\u4e00\u5f20\u56fe\u7247\u548c\u5173\u4e8e\u5982\u4f55\u62cd\u6444\u52a8\u7269\u7167\u7247\u7684\u89e3\u91ca\u3002\u7531\u4e8e\u6ca1\u6709\u63d0\u5230\u201c\u732b\u201d\u8fd9\u4e2a\u8bcd\uff0c\u5e38\u89c4\u7684\u76f8\u4f3c\u6027\u641c\u7d22\u6ca1\u6709\u4ea7\u751f\u4efb\u4f55\u7ed3\u679c\u3002 \u591a\u6a21\u6001\u641c\u7d22\u7b54\u6848 \u8fd9\u672c\u6742\u5fd7\u520a\u767b\u4e86\u4e00\u5f20\u732b\u7684\u8096\u50cf\uff0c\u7a81\u51fa\u4e86\u5176\u9762\u90e8\u7279\u5f81\u548c\u6027\u683c\u7684\u7cbe\u7ec6\u6355\u6349\u3002\u6587\u5b57\u5f3a\u8c03\u4e86\u5236\u4f5c\u7cbe\u826f\u7684\u52a8\u7269\u8096\u50cf\u5982\u4f55\u6df1\u5165\u5230\u4e3b\u9898\u7684\u7075\u9b42\uff0c\u5e76\u901a\u8fc7\u5f15\u4eba\u6ce8\u76ee\u7684\u773c\u795e\u4ea4\u6d41\u4e0e\u89c2\u770b\u8005\u5efa\u7acb\u60c5\u611f\u8054\u7cfb\u3002 \u4f7f\u7528\u591a\u6a21\u6001\u641c\u7d22\uff0c\u6211\u4eec\u5c06\u627e\u5230\u4e00\u5f20\u732b\u7684\u56fe\u7247\uff0c\u7136\u540e\u5c06\u76f8\u5173\u7684\u6587\u5b57\u94fe\u63a5\u5230\u5b83\u3002\u5c06\u8fd9\u4e9b\u6570\u636e\u63d0\u4f9b\u7ed9\u6a21\u578b\u5c06\u4f7f\u5176\u80fd\u591f\u66f4\u597d\u5730\u56de\u7b54\u548c\u7406\u89e3\u4e0a\u4e0b\u6587\u3002 &nbsp; \u5982\u4f55\u6784\u5efa\u591a\u6a21\u6001\u5d4c\u5165\u548c\u68c0\u7d22\u7ba1\u9053 \u73b0\u5728\uff0c\u6211\u5c06\u5206\u51e0\u4e2a\u6b65\u9aa4\u63cf\u8ff0\u8fd9\u6837\u4e00\u4e2a\u7ba1\u9053\u7684\u5de5\u4f5c\u539f\u7406\uff1a \u6211\u4eec\u5c06\u4f7f\u7528&nbsp;\uff08\u4e00\u4e2a\u7528\u4e8e\u6570\u636e\u63d0\u53d6\u7684\u5f3a\u5927 Python \u5e93\uff09\u4ece PDF \u6587\u4ef6\u4e2d\u63d0\u53d6\u6587\u672c\u548c\u56fe\u50cf\u3002 \u6211\u4eec\u5c06\u4f7f\u7528&nbsp;&nbsp;\u6a21\u578b\u4e3a\u540c\u4e00\u5411\u91cf\u7a7a\u95f4\u5185\u7684\u6587\u672c\u548c\u56fe\u50cf\u521b\u5efa\u591a\u6a21\u6001\u5411\u91cf\u3002 \u6211\u4eec\u5c06\u628a\u5b83\u63d2\u5165\u5230\u5411\u91cf\u5b58\u50a8 () \u4e2d\u3002 \u6700\u540e\uff0c\u6211\u4eec\u5c06\u6267\u884c\u76f8\u4f3c\u6027\u641c\u7d22\u5e76\u6bd4\u8f83\u6587\u672c\u548c\u56fe\u50cf\u7684\u7ed3\u679c\u3002 \u7b2c 1 \u6b65\uff1a\u8bbe\u7f6e\u5411\u91cf\u5b58\u50a8\u5e76\u4ece\u6587\u4ef6 (PDF) \u4e2d\u63d0\u53d6\u56fe\u50cf\u548c\u6587\u672c \u5728\u8fd9\u91cc\uff0c\u6211\u4eec\u5fc5\u987b\u505a\u4e00\u4e9b\u624b\u52a8\u5de5\u4f5c\u3002\u901a\u5e38\uff0cWeaviate \u662f\u4e00\u4e2a\u975e\u5e38\u6613\u4e8e\u4f7f\u7528\u7684\u5411\u91cf\u5b58\u50a8\uff0c\u5b83\u4f1a\u5728\u63d2\u5165\u65f6\u81ea\u52a8\u8f6c\u6362\u6570\u636e\u5e76\u6dfb\u52a0\u5d4c\u5165\u3002\u4f46\u662f\uff0c\u6ca1\u6709\u7528\u4e8e Voyager Multimodal v3 \u7684\u63d2\u4ef6\uff0c\u56e0\u6b64\u6211\u4eec\u5fc5\u987b\u624b\u52a8\u8ba1\u7b97\u5d4c\u5165\u3002\u5728\u8fd9\u79cd\u60c5\u51b5\u4e0b\uff0c\u6211\u4eec\u5fc5\u987b\u521b\u5efa\u4e00\u4e2a\u96c6\u5408\u800c\u4e0d\u5b9a\u4e49\u5411\u91cf\u5316\u5668\u6a21\u5757\u3002 import weaviate from weaviate.classes.config import Configure client = weaviate.connect_to_local() collection_name [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[13],"tags":[],"class_list":["post-55859","post","type-post","status-publish","format-standard","hentry","category-ai"],"_links":{"self":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts\/55859","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/comments?post=55859"}],"version-history":[{"count":0,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts\/55859\/revisions"}],"wp:attachment":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/media?parent=55859"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/categories?post=55859"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/tags?post=55859"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}