向Qdrant添加向量时触发422验证错误，求问题排查方案

UnexpectedResponse   Traceback (most recent call last)<ipython-input-36-42a89db32382> in  
----> 3 add_vectors(embeddings, payload)

6 frames
/usr/local/lib/python3.10/dist-packages/qdrant_client/http/api_client.py in send(self, request, type_)
     95             except ValidationError as e:
     96                 raise ResponseHandlingException(e)
---&gt; 97         raise UnexpectedResponse.for_response(response)
     98 
     99     def send_inner(self, request: Request) -&gt; Response:

UnexpectedResponse: Unexpected Response: 422 (Unprocessable Entity)
Raw response content:
b'{"status":{"error":"Validation error in path parameters: [name: value \\"status=&lt;CollectionStatus.GREEN: \\'green\\'&gt; optimizer_status=&lt;OptimizersStatusOneOf.OK: \\'ok\\'&gt; vectors_count=0 indexed_vectors_co ..."}}'

以下是我的代码：

raw_text = "some long text...."

def get_chunks(raw_text):

    text_splitter = CharacterTextSplitter(       
      separator="\n",       
      chunk_size=100,        
      chunk_overlap=50,        
      length_function=len)
      chunks = text_splitter.split_text(raw_text)    
  return chunks
============
def get_embeddings(chunks, embedding_model_name="text-embedding-ada-002"):

    points = []    
    client = OpenAI(api_key=os.environ['OPENAI_API_KEY'])   
    embeddings = []    
    for chunk in chunks:        
      embeddings.append(client.embeddings.create(
      input=chunk, model=embedding_model_name).data[0].embedding)
          
    return embeddings
 
===========================================================
def add_vectors(vectors, payload):

    client = qdrant_client.QdrantClient(

        os.getenv("QDRANT_HOST"),

        api_key=os.getenv("QDRANT_API_KEY")

    )

    collection = client.get_collection(os.getenv("QDRANT_COLLECTION"))

    # Create a list of PointStruct objects

    points = [

        models.PointStruct(

            id=str(i),  # Assign unique IDs to points

            payload=payload,

            vector=vector

        )

        for i, vector in enumerate(vectors)

    ]


# Insert the points into the vector store

client.upsert(

    collection_name=collection,  # Replace with your collection name

    points=points

)

我的调用逻辑如下：

chunks = get_chunks(raw_text)
embeddings = get_embeddings(chunks)
payload = {"user": "gxxxx"}
add_vectors(embeddings, payload)

执行上述调用时触发了上述错误，我已尝试网络上多种解决方案但仍未解决，请问问题出在哪里？

错误原因分析 #

1. collection_name参数类型错误
   client.get_collection()返回的是CollectionInfo对象，而client.upsert()的collection_name参数需要字符串类型的集合名称。直接传入对象会导致Qdrant无法解析路径参数，触发422验证错误。
2. 代码缩进错误
   client.upsert()代码块缩进错误，位于add_vectors函数外部，导致函数执行时不会触发插入操作，且变量作用域存在问题。
3. payload复用问题（非直接错误，但需优化）
   所有向量点共用同一个payload，未关联对应的文本chunk，后续检索无法获取原始文本内容。

修正后的代码 #

修正add_vectors函数 #

def add_vectors(vectors, chunks, base_payload):
    client = qdrant_client.QdrantClient(
        os.getenv("QDRANT_HOST"),
        api_key=os.getenv("QDRANT_API_KEY")
    )
    collection_name = os.getenv("QDRANT_COLLECTION")

    # 为每个向量点生成包含chunk内容的payload
    points = [
        models.PointStruct(
            id=str(i),
            payload={**base_payload, "text": chunk},  # 合并基础payload与当前chunk文本
            vector=vector
        )
        for i, (vector, chunk) in enumerate(zip(vectors, chunks))
    ]

    # 执行插入操作，使用字符串类型的集合名称
    client.upsert(
        collection_name=collection_name,
        points=points
    )

修正调用逻辑 #

chunks = get_chunks(raw_text)
embeddings = get_embeddings(chunks)
base_payload = {"user": "gxxxx"}
add_vectors(embeddings, chunks, base_payload)

修正get_chunks函数的缩进错误 #

def get_chunks(raw_text):
    text_splitter = CharacterTextSplitter(       
        separator="\n",       
        chunk_size=100,        
        chunk_overlap=50,        
        length_function=len
    )
    chunks = text_splitter.split_text(raw_text)    
    return chunks

关键修正点说明 #

• 直接使用环境变量中的集合名称字符串，替代get_collection()返回的对象。
• 调整client.upsert()的缩进，确保它在add_vectors函数内部执行。
• 将文本chunk传入函数，为每个向量点添加对应的文本内容到payload，提升检索实用性。
• 修复get_chunks函数的缩进问题，避免语法错误。

内容的提问来源于stack exchange，提问作者Gautam Mukherjee