使用llama-index from_vector_store遇Pydantic验证错误求助

报错信息 #

pydantic_core._pydantic_core.ValidationError: NodeWithScore存在1个验证错误
node
输入应为有效字典或BaseNode实例 [type=model_type, input_value=TextNode(id='dcb36e46-7a...metadata_seperator='\n'), input_type=TextNode]

报错代码 #

import textwrap

from llama_index.core import VectorStoreIndex, StorageContext
from llama_index.legacy.vector_stores import PGVectorStore
from llama_index.llms.openai import OpenAI
from sqlalchemy import make_url
import os
import openai

# Get openAI api key by reading local .env file
openai.api_key = ("my-api-key")
os.environ["OPENAI_API_KEY"] = openai.api_key

print("Connecting to new vector db...")
connection_string = "postgresql://user:user@localhost:5432"

db_name = "vector_db"

url = make_url(connection_string)
vector_store = PGVectorStore.from_params(
    database=db_name,
    host=url.host,
    password=url.password,
    port=url.port,
    user=url.username,
    table_name="table_name",
    embed_dim=1536,  # openai embedding dimension
)

index = VectorStoreIndex.from_vector_store(
    vector_store=vector_store)

llm = OpenAI(model="gpt-3.5-turbo", temperature=0.1)

query_engine = index.as_query_engine(llm=llm)

response = query_engine.query("Give me a summary of the data with code AA10B")

print(textwrap.fill(str(response), 100))

问题描述 #

使用VectorStoreIndex.from_vector_store()方法连接PGVectorStore时触发上述验证错误，但改用VectorStoreIndex.from_documents()直接传入文档列表时代码可正常运行，不清楚问题原因。

可正常运行的代码 #

import textwrap

from llama_index.core import StorageContext
from llama_index.llms.openai import OpenAI
from llama_index.readers.database import DatabaseReader
from llama_index.core import VectorStoreIndex
from llama_index.vector_stores.postgres import PGVectorStore
import openai
import os
from sqlalchemy import make_url, create_engine, text
import psycopg2

# Get openAI api key by reading local .env file
openai.api_key = ("my-api-key")
os.environ["OPENAI_API_KEY"] = openai.api_key

engine = create_engine("postgresql+psycopg2://user:user@localhost/db")

reader = DatabaseReader(
    engine=engine
)

query = """query"""

documents = reader.load_data(query=query)

# Recreate database if exists
conn = psycopg2.connect("postgres://user:user@localhost:5432/db")
conn.autocommit = True
cur = conn.cursor()
cur.execute("DROP DATABASE IF EXISTS vector_db;")
cur.execute("CREATE DATABASE vector_db;")
conn.close()

conn = psycopg2.connect("postgres://user:user@localhost:5432/vector_db")
conn.autocommit = True
cur = conn.cursor()
cur.execute("CREATE EXTENSION vector;")
conn.close()

connection_string = "postgresql://user:user@localhost:5432"

db_name = "vector_db"

url = make_url(connection_string)
vector_store = PGVectorStore.from_params(
    database=db_name,
    host=url.host,
    password=url.password,
    port=url.port,
    user=url.username,
    table_name="table",
    embed_dim=1536,  # openai embedding dimension
    hnsw_kwargs={
        "hnsw_m": 16,
        "hnsw_ef_construction": 64,
        "hnsw_ef_search": 40,
        "hnsw_dist_method": "vector_cosine_ops",
    },
)

storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context, show_progress=True
)

llm = OpenAI(model="gpt-3.5-turbo", temperature=0.1)

query_engine = index.as_query_engine(llm=llm)

response = query_engine.query("Give me a summary of the data with code AA10B")

print(textwrap.fill(str(response), 100))

问题根源与解决方案 #

核心原因 #

报错本质是新旧版本PGVectorStore模块不兼容：

• 报错代码导入的是llama_index.legacy.vector_stores.PGVectorStore（旧版遗留模块）
• 正常代码导入的是llama_index.vector_stores.postgres.PGVectorStore（新版核心模块）

LlamaIndex迭代中重构了PGVectorStore的实现，旧版模块与新版VectorStoreIndex.from_vector_store()方法存在类型解析冲突，导致查询时无法正确识别存储的TextNode，触发Pydantic验证错误。

解决步骤 #

1. 替换导入路径
   将报错代码中的旧版导入：

   from llama_index.legacy.vector_stores import PGVectorStore
   

   替换为新版导入：

   from llama_index.vector_stores.postgres import PGVectorStore
   

2. 确保数据库结构兼容
   若之前用旧版PGVectorStore创建过数据表，需参考正常代码中的逻辑重建数据库（删除旧库→新建库→创建vector扩展），因为新旧版本的表结构存在差异。

3. 补充必要参数
   新版PGVectorStore建议添加hnsw_kwargs索引参数，提升查询性能和兼容性（参考正常代码中的配置）。

修正后的完整代码 #

import textwrap

from llama_index.core import VectorStoreIndex, StorageContext
from llama_index.vector_stores.postgres import PGVectorStore
from llama_index.llms.openai import OpenAI
from sqlalchemy import make_url
import os
import openai

# 设置OpenAI API密钥
openai.api_key = "my-api-key"
os.environ["OPENAI_API_KEY"] = openai.api_key

print("Connecting to new vector db...")
connection_string = "postgresql://user:user@localhost:5432"

db_name = "vector_db"

url = make_url(connection_string)
vector_store = PGVectorStore.from_params(
    database=db_name,
    host=url.host,
    password=url.password,
    port=url.port,
    user=url.username,
    table_name="table_name",
    embed_dim=1536,  # OpenAI Embedding维度
    hnsw_kwargs={
        "hnsw_m": 16,
        "hnsw_ef_construction": 64,
        "hnsw_ef_search": 40,
        "hnsw_dist_method": "vector_cosine_ops",
    },
)

storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_vector_store(
    vector_store=vector_store, storage_context=storage_context
)

llm = OpenAI(model="gpt-3.5-turbo", temperature=0.1)

query_engine = index.as_query_engine(llm=llm)

response = query_engine.query("Give me a summary of the data with code AA10B")

print(textwrap.fill(str(response), 100))

内容的提问来源于stack exchange，提问作者asemprini87