连接Pinecone与OpenAI时出现MaxRetryError问题求助

我开发了一款简易应用，支持上传PDF文件、分割为文本块、生成Embedding并上传至Pinecone向量数据库。执行代码docsearch = Pinecone.from_texts([t.page_content for t in texts], embeddings, index_name=index_name)时触发如下MaxRetryError错误：

SSLEOFError                               Traceback (most recent call last)
File /usr/lib/python3/dist-packages/urllib3/connectionpool.py:699, in HTTPConnectionPool.urlopen(self, method, url, body, headers, retries, redirect, assert_same_host, timeout, pool_timeout, release_conn, chunked, body_pos, **response_kw)
    698 # Make the request on the httplib connection object.
--> 699 httplib_response = self._make_request(
    700     conn,
    701     method,
    702     url,
    703     timeout=timeout_obj,
    704     body=body,
    705     headers=headers,
    706     chunked=chunked,
    707 )
    709 # If we're going to release the connection in ``finally:``, then
    710 # the response doesn't need to know about the connection. Otherwise
    711 # it will also try to release it and we'll have a double-release
    712 # mess.

File /usr/lib/python3/dist-packages/urllib3/connectionpool.py:394, in HTTPConnectionPool._make_request(self, conn, method, url, timeout, chunked, **httplib_request_kw)
    393     else:
--> 394         conn.request(method, url, **httplib_request_kw)
    396 # We are swallowing BrokenPipeError (errno.EPIPE) since the server is
    397 # legitimately able to close the connection after sending a valid response.
    398 # With this behaviour, the received response is still readable.
...
--> 574     raise MaxRetryError(_pool, url, error or ResponseError(cause))
    576 log.debug("Incremented Retry for (url='%s'): %r", url, new_retry)
    578 return new_retry

MaxRetryError: HTTPSConnectionPool(host='langchain2-e630e5d.svc.asia-northeast1-gcp.pinecone.io', port=443): Max retries exceeded with url: /vectors/upsert (Caused by SSLError(SSLEOFError(8, 'EOF occurred in violation of protocol (_ssl.c:2396)')))

完整代码如下：

from langchain.text_splitter import RecursiveCharacterTextSplitter

# 加载数据
loader = UnstructuredPDFLoader("../data/field-guide-to-data-science.pdf")
# loader = OnlinePDFLoader("https://wolfpaulus.com/wp-content/uploads/2017/05/field-guide-to-data-science.pdf")

data = loader.load()

print (f'You have {len(data)} document(s) in your data')
print (f'There are {len(data[0].page_content)} characters in your document')

# 输出：
# You have 1 document(s) in your data
# There are 176584 characters in your document

# 将数据分割为更小的文本块
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
texts = text_splitter.split_documents(data)

print (f'Now you have {len(texts)} documents')

# 输出：
# Now you have 228 documents

# 生成文档Embedding以准备语义搜索
from langchain.vectorstores import Chroma, Pinecone
from langchain.embeddings.openai import OpenAIEmbeddings
import pinecone

OPENAI_API_KEY = '...'
PINECONE_API_KEY = '...'
PINECONE_API_ENV = 'us-east1-gcp'

embeddings = OpenAIEmbeddings(openai_api_key=OPENAI_API_KEY)

# 初始化Pinecone
pinecone.init(
    api_key=PINECONE_API_KEY,  # 在app.pinecone.io获取
    environment=PINECONE_API_ENV  # 控制台中API密钥旁查看
)
index_name = "langchain2"

docsearch = Pinecone.from_texts([t.page_content for t in texts], embeddings, index_name=index_name)

问题排查与解决方案 #

• 核心问题：环境配置不匹配
  错误信息中的Pinecone服务地址属于asia-northeast1-gcp环境，但代码里指定的PINECONE_API_ENV是us-east1-gcp，两者不匹配导致连接失败。登录Pinecone控制台，确认目标索引所在的环境，修改代码中的PINECONE_API_ENV为对应值（比如asia-northeast1-gcp）。

• 网络连接验证
  检查本地网络是否能访问Pinecone的目标服务地址，可执行如下命令测试：

  curl https://langchain2-e630e5d.svc.asia-northeast1-gcp.pinecone.io/vectors/upsert
  

  若无法连接，需排查防火墙规则、代理设置，确保网络能正常通达Pinecone的服务端口。

• 更新依赖版本
  版本不兼容可能引发SSL连接问题，更新相关依赖到稳定版本：

  pip install --upgrade pinecone-client langchain openai
  

• 分批次上传数据
  一次性上传228个文本块可能导致请求过大，尝试分批次上传：

  batch_size = 50
  for i in range(0, len(texts), batch_size):
      batch_texts = texts[i:i+batch_size]
      Pinecone.from_texts(
          [t.page_content for t in batch_texts], 
          embeddings, 
          index_name=index_name
      )
  

• 临时绕过SSL验证（仅测试用）
  若本地SSL证书存在问题，可临时关闭SSL验证（生产环境不建议使用）：

  import ssl
  ssl._create_default_https_context = ssl._create_unverified_context
  

内容的提问来源于stack exchange，提问作者Nordic Guy