AttributeError on calling LLMGraphTransformer.convert_to_graph_documents

[mintomohan]

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

from langchain_experimental.graph_transformers.llm import LLMGraphTransformer
from langchain.llms.bedrock import Bedrock
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.retrievers import WikipediaRetriever
from langchain_community.graphs.neo4j_graph import Neo4jGraph

bedrock=boto3.client(service_name='bedrock-runtime')

def prepare_graph(wiki_keyword):
    wiki_retriever = WikipediaRetriever(doc_content_chars_max=2000, top_k_results=1)
    docs = wiki_retriever.invoke(wiki_keyword)
    
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)
    doc_chunks = text_splitter.split_documents(docs)
    
    llm=Bedrock(model_id='mistral.mistral-7b-instruct-v0:2', client=bedrock)
    llm_transformer = LLMGraphTransformer(llm=llm)    
    
    graph_documents = llm_transformer.convert_to_graph_documents(doc_chunks)
    graph = Neo4jGraph()
    graph.add_graph_documents(graph_documents)

Error Message and Stack Trace (if applicable)

Traceback (most recent call last):
  File "...\...\load_data_graph.py", line 46, in <module>
    prepare_graph('Paris')
  File "...\...\load_data_graph.py", line 38, in prepare_graph
    graph_documents = llm_transformer.convert_to_graph_documents(doc_chunks)
  File "...\...\venv\lib\site-packages\langchain_experimental\graph_transformers\llm.py", line 646, in convert_to_graph_documents   
    return [self.process_response(document) for document in documents]
  File "...\...\venv\lib\site-packages\langchain_experimental\graph_transformers\llm.py", line 646, in <listcomp>
    return [self.process_response(document) for document in documents]
  File "...\...\venv\lib\site-packages\langchain_experimental\graph_transformers\llm.py", line 595, in process_response
    parsed_json = self.json_repair.loads(raw_schema.content)
AttributeError: 'str' object has no attribute 'content'

Description

I am trying to load a page from Wikipedia, split it and load to Neo4j using langchain

Wikipedia --> WikipediaRetriever --> RecursiveCharacterTextSplitter --> LLMGraphTransformer --> Neo4jGraph

LLM used is Mistral 7B (using AWS Bedrock)

System Info

langchain==0.1.17
langchain-community==0.0.36
langchain-core==0.1.52
langchain-experimental==0.0.58
langchain-text-splitters==0.0.1
langsmith==0.1.52

platform : windows

Python 3.10.8

[liugddx]

Let me see.

[UmutAlihan]

I am having the same issue here :'/

[somashekarbrdtscblr]

its able to run llama2 server and model but having this same bug/error

CODE:
%pip install --upgrade --quiet langchain langchain-community langchain-openai langchain-experimental neo4j json_repair

import os
import json
import types
from langchain_community.graphs import Neo4jGraph
from langchain_experimental.graph_transformers import LLMGraphTransformer
from langchain_openai import ChatOpenAI
from langchain_core.documents import Document
from langchain_community.llms import Ollama

llm = Ollama(temperature=0, model="llama2")
llm_transformer = LLMGraphTransformer(llm=llm)

os.environ["NEO4J_URI"] = "neo4j://localhost:7687"
os.environ["NEO4J_USERNAME"] = "neo4j"
os.environ["NEO4J_PASSWORD"] = "neo4jadmin"

graph = Neo4jGraph()

text = """Marie Curie, born in 1867, was a Polish and naturalised-French physicist and chemist who conducted pioneering research on radioactivity.She was the first woman to win a Nobel Prize, the first person to win a Nobel Prize twice, and the only person to win a Nobel Prize in two scientific fields.Her husband, Pierre Curie, was a co-winner of her first Nobel Prize, making them the first-ever married couple to win the Nobel Prize and launching the Curie family legacy of five Nobel Prizes.She was, in 1906, the first woman to become a professor at the University of Paris."""

documents = [Document(page_content=text)]
graph_documents = llm_transformer.convert_to_graph_documents(documents)

ERROR

AttributeError Traceback (most recent call last)
Cell In[27], line 2
1 documents = [Document(page_content=text)]
----> 2 graph_documents = llm_transformer.convert_to_graph_documents(documents)

File C:\tools\Anaconda3\lib\site-packages\langchain_experimental\graph_transformers\llm.py:646, in LLMGraphTransformer.convert_to_graph_documents(self, documents)
634 def convert_to_graph_documents(
635 self, documents: Sequence[Document]
636 ) -> List[GraphDocument]:
637 """Convert a sequence of documents into graph documents.
638
639 Args:
(...)
644 Sequence[GraphDocument]: The transformed documents as graphs.
645 """
--> 646 return [self.process_response(document) for document in documents]

File C:\tools\Anaconda3\lib\site-packages\langchain_experimental\graph_transformers\llm.py:646, in (.0)
634 def convert_to_graph_documents(
635 self, documents: Sequence[Document]
636 ) -> List[GraphDocument]:
637 """Convert a sequence of documents into graph documents.
638
639 Args:
(...)
644 Sequence[GraphDocument]: The transformed documents as graphs.
645 """
--> 646 return [self.process_response(document) for document in documents]

File C:\tools\Anaconda3\lib\site-packages\langchain_experimental\graph_transformers\llm.py:595, in LLMGraphTransformer.process_response(self, document)
593 nodes_set = set()
594 relationships = []
--> 595 parsed_json = self.json_repair.loads(raw_schema.content)
596 for rel in parsed_json:
597 # Nodes need to be deduplicated using a set
598 nodes_set.add((rel["head"], rel["head_type"]))

AttributeError: 'str' object has no attribute 'content'

[Yweik]

---

AttributeError Traceback (most recent call last)
in <cell line: 5>()
3 text_splitter = TokenTextSplitter(chunk_size=512, chunk_overlap=24)
4 documents = text_splitter.split_documents(docs)
----> 5 graph_documents = llm_transformer.convert_to_graph_documents(documents)

2 frames
/usr/local/lib/python3.10/dist-packages/langchain_experimental/graph_transformers/llm.py in process_response(self, document)
593 nodes_set = set()
594 relationships = []
--> 595 parsed_json = self.json_repair.loads(raw_schema.content)
596 for rel in parsed_json:
597 # Nodes need to be deduplicated using a set

AttributeError: 'str' object has no attribute 'content'
I encountered the same problem

[love11mishra]

I'm also facing the same error.

[francruzdiconium]

Running the same "Marie Curie" use case (with AzureOpeAi) i always get empty nodes. On Langsmith I can that the model is able to extract the nodes, but i see on the PydanticToolsParser step:

[big json here] are not valid JSON. Received JSONDecodeError Expecting value: line 1 column 1 (char 0)

[mintomohan]

Updates on the original issue:

(1) I switched from Mistral to OpenAI (GPT 3.5 Turbo) and encountered a different error message.

ValueError: Could not use APOC procedures.

(2) I referred to the following issue and adjusted the settings in Neo4j as suggested: #12901 (comment)

Currently, I am facing this error:

    graph_documents = llm_transformer.convert_to_graph_documents(doc_chunks)
  File "...\...\lib\site-packages\langchain_experimental\graph_transformers\llm.py", line 646, in convert_to_graph_documents
    return [self.process_response(document) for document in documents]
  File "...\...\lib\site-packages\langchain_experimental\graph_transformers\llm.py", line 646, in <listcomp>      
    return [self.process_response(document) for document in documents]
  File "...\...\lib\site-packages\langchain_experimental\graph_transformers\llm.py", line 598, in process_response    nodes_set.add((rel["head"], rel["head_type"]))
TypeError: string indices must be integers

Version Info:

langchain                                0.1.20
langchain-aws                            0.1.6
langchain-cohere                         0.1.5
langchain-community                      0.0.38
langchain-core                           0.1.52
langchain-experimental                   0.0.58
langchain-openai                         0.0.5
langchain-text-splitters                 0.0.1
neo4j                                    5.20.0