search

microsoft / graphrag

A modular graph-based Retrieval-Augmented Generation (RAG) system
https://microsoft.github.io/graphrag/
MIT License
18.82k stars 1.84k forks source link

[Bug]: TypeError: Query column vector must be a vector. Got list<item: double>. #1335

Open zel2023 opened 1 week ago

zel2023 commented 1 week ago

Do you need to file an issue?

Describe the bug

I attempted to refer to https://github.com/microsoft/graphrag/blob/main/docs/examples_notebooks/local_search.ipynb to write a Python file for running local search but failed. However, using https://github.com/microsoft/graphrag/blob/main/docs/examples_notebooks/global_search.ipynb as a reference, I successfully wrote a Python file to run global search. Additionally, I successfully ran the local search by referring to https://github.com/microsoft/graphrag/blob/94f1e62e5c06795fc8c361dba6580bb76d6e77ce/docs/get_started.md. Below is the error message:

Entity count: 3 Relationship count: 2 Report records: 1 Text unit records: 1 Traceback (most recent call last): File "/data/zelongzheng/graphrag-main/local_search.py", line 172, in asyncio.run(main()) File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/asyncio/runners.py", line 190, in run return runner.run(main) ^^^^^^^^^^^^^^^^ File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/asyncio/runners.py", line 118, in run return self._loop.run_until_complete(task) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/asyncio/base_events.py", line 654, in run_until_complete return future.result() ^^^^^^^^^^^^^^^ File "/data/zelongzheng/graphrag-main/local_search.py", line 167, in main result = await search_engine.asearch("what is the relationship between xiaozhang and xiaoming?") ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/data/zelongzheng/graphrag-main/graphrag/query/structured_search/local_search/search.py", line 67, in asearch context_text, context_records = self.context_builder.build_context( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/data/zelongzheng/graphrag-main/graphrag/query/structured_search/local_search/mixed_context.py", line 140, in build_context selected_entities = map_query_to_entities( ^^^^^^^^^^^^^^^^^^^^^^ File "/data/zelongzheng/graphrag-main/graphrag/query/context_builder/entity_extraction.py", line 57, in map_query_to_entities search_results = text_embedding_vectorstore.similarity_search_by_text( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/data/zelongzheng/graphrag-main/graphrag/vector_stores/lancedb.py", line 136, in similarity_search_by_text return self.similarity_search_by_vector(query_embedding, k) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/data/zelongzheng/graphrag-main/graphrag/vector_stores/lancedb.py", line 115, in similarity_search_by_vector .to_list() ^^^^^^^^^ File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/site-packages/lancedb/query.py", line 320, in to_list return self.to_arrow().to_pylist() ^^^^^^^^^^^^^^^ File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/site-packages/lancedb/query.py", line 647, in to_arrow return self.to_batches().read_all() ^^^^^^^^^^^^^^^^^ File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/site-packages/lancedb/query.py", line 678, in to_batches result_set = self._table._execute_query(query, batch_size) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/site-packages/lancedb/table.py", line 1742, in _execute_query return ds.scanner( ^^^^^^^^^^^ File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/site-packages/lance/dataset.py", line 369, in scanner builder = builder.nearest(**nearest) ^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/zelongzheng/anaconda3/envs/graphrag/lib/python3.11/site-packages/lance/dataset.py", line 2449, in nearest raise TypeError( TypeError: Query column vector must be a vector. Got list.

Steps to reproduce

1.pip install graphrag==0.3.6

2.Build the graph and use the command python -m graphrag.query --root ./ragtest --method local "what is the relationship between xiaozhang and xiaoming?" to confirm successful execution.

3.Write a python file using https://github.com/microsoft/graphrag/blob/main/docs/examples_notebooks/local_search.ipynb as a reference. Modify certain parts of this file: change INPUT_DIR, comment out all variables related to covariates since no related files were generated when building the graph, set API_KEY, llm_model (deepseek-chat), embedding_model (text-embedding-3-small), and api_base="https://api.agicto.cn/v1".

4.Run the file using Python.

Expected Behavior

I expect it will response for what I ask.

GraphRAG Config Used

graphrag=0.3.6
llm = "deepseek-chat"
embedding_model = "text-embedding-3-small"
api_base = "https://api.agicto.cn/v1"

Logs and screenshots

Image

Additional Information

EniacTNB commented 3 days ago

Did you solve that?

zel2023 commented 3 days ago

I did not solve that.

yuanzhoulvpi2017 commented 2 days ago

I got same issue ,this is a tiny code, This bug is really weird.

import numpy as np
import json
import pyarrow as pa
import lancedb

class TextEmbeder:
    def __init__(self) -> None:
        pass

    def encode(self, x):
        return np.abs(np.around(np.random.randn(3), 3)).tolist()

textembeder = TextEmbeder()
textembeder.encode("a")[:3]

def build_fake_data():
    res = []
    for index in range(1000):
        id_ = str(index)
        text = f"hello test {index}"
        vector = textembeder.encode(text)
        extra_data = json.dumps(
            {"attr1": index, "attr2": index * 2, "attr3": "B站"}, ensure_ascii=True
        )

        res.append(
            {"id": id_, "text": text, "extra_data": extra_data, "vector": vector}
        )

    return res

fake_data_list = build_fake_data()
fake_data_list[0].keys()

db_url = "data/database"
db_table_name = "smalltest"
db_connection = lancedb.connect(db_url)

db_connection.table_names()

schema = pa.schema(
    [
        pa.field("vector", pa.list_(pa.float64())),
        pa.field("id", pa.string()),
        pa.field("text", pa.string()),
        pa.field("extra_data", pa.string()),
    ]
)

db_connection.create_table(name=db_table_name, schema=schema, mode="overwrite")

table = db_connection.open_table(db_table_name)
table.add(fake_data_list)

query_vector = textembeder.encode("hh")
query_vector[:4]

docs = (
    table.search(query=query_vector, vector_column_name="vector", query_type="vector")
    .limit(4)
    .to_list()
)

print(docs)

env:

 pip show lancedb            
Name: lancedb
Version: 0.13.0
Summary: lancedb
Home-page: 
Author: 
Author-email: LanceDB Devs <dev@lancedb.com>
License: Apache-2.0
Location: /data2/miniconda3/envs/hz_grahrag/lib/python3.11/site-packages
Requires: attrs, cachetools, deprecation, overrides, packaging, pydantic, pylance, requests, retry, tqdm
Required-by: graphrag

Name: pyarrow
Version: 15.0.2
Summary: Python library for Apache Arrow
Home-page: https://arrow.apache.org/
Author: 
Author-email: 
License: Apache License, Version 2.0
Location: /data2/miniconda3/envs/hz_grahrag/lib/python3.11/site-packages
Requires: numpy
Required-by: datasets, datashaper, graphrag, pylance, streamlit
yuanzhoulvpi2017 commented 2 days ago

I had fix this issue,in schema,need set vectorsize, `pa.field("vector", pa.list(pa.float32(), list_size=DIM_VALUE)),`

schema = pa.schema(
    [
        pa.field("vector", pa.list_(pa.float32(), list_size=DIM_VALUE)),
        pa.field("id", pa.string()),
        pa.field("text", pa.string()),
        pa.field("extra_data", pa.string()),
    ]
)
zel2023 commented 1 day ago

Congratulations! However, there is no “schema” in my code：

import os

import pandas as pd
import tiktoken
import asyncio
from graphrag.query.context_builder.entity_extraction import EntityVectorStoreKey
from graphrag.query.indexer_adapters import (
    read_indexer_covariates,
    read_indexer_entities,
    read_indexer_relationships,
    read_indexer_reports,
    read_indexer_text_units,
)
from graphrag.query.input.loaders.dfs import (
    store_entity_semantic_embeddings,
)
from graphrag.query.llm.oai.chat_openai import ChatOpenAI
from graphrag.query.llm.oai.embedding import OpenAIEmbedding
from graphrag.query.llm.oai.typing import OpenaiApiType
from graphrag.query.question_gen.local_gen import LocalQuestionGen
from graphrag.query.structured_search.local_search.mixed_context import (
    LocalSearchMixedContext,
)
from graphrag.query.structured_search.local_search.search import LocalSearch
from graphrag.vector_stores.lancedb import LanceDBVectorStore

INPUT_DIR = "xx"
LANCEDB_URI = f"{INPUT_DIR}/lancedb"

COMMUNITY_REPORT_TABLE = "create_final_community_reports"
ENTITY_TABLE = "create_final_nodes"
ENTITY_EMBEDDING_TABLE = "create_final_entities"
RELATIONSHIP_TABLE = "create_final_relationships"
#COVARIATE_TABLE = "create_final_covariates"
TEXT_UNIT_TABLE = "create_final_text_units"
COMMUNITY_LEVEL = 2

# read nodes table to get community and degree data
entity_df = pd.read_parquet(f"{INPUT_DIR}/{ENTITY_TABLE}.parquet")
entity_embedding_df = pd.read_parquet(f"{INPUT_DIR}/{ENTITY_EMBEDDING_TABLE}.parquet")

entities = read_indexer_entities(entity_df, entity_embedding_df, COMMUNITY_LEVEL)

# load description embeddings to an in-memory lancedb vectorstore
# to connect to a remote db, specify url and port values.
description_embedding_store = LanceDBVectorStore(
    collection_name="entity_description_embeddings",
)
description_embedding_store.connect(db_uri=LANCEDB_URI)
entity_description_embeddings = store_entity_semantic_embeddings(
    entities=entities, vectorstore=description_embedding_store
)

print(f"Entity count: {len(entity_df)}")
entity_df.head()

relationship_df = pd.read_parquet(f"{INPUT_DIR}/{RELATIONSHIP_TABLE}.parquet")
relationships = read_indexer_relationships(relationship_df)

print(f"Relationship count: {len(relationship_df)}")
relationship_df.head()

# NOTE: covariates are turned off by default, because they generally need prompt tuning to be valuable
# Please see the GRAPHRAG_CLAIM_* settings
#covariate_df = pd.read_parquet(f"{INPUT_DIR}/{COVARIATE_TABLE}.parquet")

#claims = read_indexer_covariates(covariate_df)

#print(f"Claim records: {len(claims)}")
#covariates = {"claims": claims}

report_df = pd.read_parquet(f"{INPUT_DIR}/{COMMUNITY_REPORT_TABLE}.parquet")
reports = read_indexer_reports(report_df, entity_df, COMMUNITY_LEVEL)

print(f"Report records: {len(report_df)}")
report_df.head()

text_unit_df = pd.read_parquet(f"{INPUT_DIR}/{TEXT_UNIT_TABLE}.parquet")
text_units = read_indexer_text_units(text_unit_df)

print(f"Text unit records: {len(text_unit_df)}")
text_unit_df.head()

api_key = "xx"
llm_model = "deepseek-chat"
embedding_model = "text-embedding-3-small"

#embedding_model = os.environ["text-embedding-3-small"]

llm = ChatOpenAI(
    api_key=api_key,
    model=llm_model,
    api_type=OpenaiApiType.OpenAI,  # OpenaiApiType.OpenAI or OpenaiApiType.AzureOpenAI
    max_retries=20,
    api_base="https://api.agicto.cn/v1"
)

token_encoder = tiktoken.get_encoding("cl100k_base")

text_embedder = OpenAIEmbedding(
    api_key=api_key,
    api_base="https://api.agicto.cn/v1",
    api_type=OpenaiApiType.OpenAI,
    model=embedding_model,
    deployment_name=embedding_model,
    max_retries=20,
)

context_builder = LocalSearchMixedContext(
    community_reports=reports,
    text_units=text_units,
    entities=entities,
    relationships=relationships,
    # if you did not run covariates during indexing, set this to None
    covariates=None,
    entity_text_embeddings=description_embedding_store,
    embedding_vectorstore_key=EntityVectorStoreKey.TITLE,  # if the vectorstore uses entity title as ids, set this to EntityVectorStoreKey.TITLE
    text_embedder=text_embedder,
    token_encoder=token_encoder,
)

local_context_params = {
    "text_unit_prop": 0.5,
    "community_prop": 0.1,
    "conversation_history_max_turns": 5,
    "conversation_history_user_turns_only": True,
    "top_k_mapped_entities": 10,
    "top_k_relationships": 10,
    "include_entity_rank": True,
    "include_relationship_weight": True,
    "include_community_rank": False,
    "return_candidate_context": False,
    "embedding_vectorstore_key": EntityVectorStoreKey.TITLE,  # set this to EntityVectorStoreKey.TITLE if the vectorstore uses entity title as ids
    "max_tokens": 12_000,  # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 5000)
}

llm_params = {
    "max_tokens": 2_000,  # change this based on the token limit you have on your model (if you are using a model with 8k limit, a good setting could be 1000=1500)
    "temperature": 0.0,
}

search_engine = LocalSearch(
    llm=llm,
    context_builder=context_builder,
    token_encoder=token_encoder,
    llm_params=llm_params,
    context_builder_params=local_context_params,
    response_type="multiple paragraphs",  # free form text describing the response type and format, can be anything, e.g. prioritized list, single paragraph, multiple paragraphs, multiple-page report
)

async def main():
    result = await search_engine.asearch("what is the relationship between xiaozhang and xiaoming?")
    print(result.response)

if __name__ == "__main__":
    asyncio.run(main())
EachSheep commented 1 day ago

I meet this problem too, do anyone have any suggestions?

Taylor180520 commented 12 hours ago

i had the same issue, and my problem is solved by adjusting the command. Previously, I use "graphrag.index --root ./ragtest" for indexing, and "graphrag.query --root ./ragtest --method local "explain the relationship between Jay and May." for querying.

Then i use "python -m graphrag.index --root ./ragtest" and "python -m graphrag.query --root ./ragtest --method local "explain the relationship between Jay and May.", it solves my problem.

To sum up, in my particular case, i didn't include "pyhton -m" in my execution command, and it turns out problematic.