llm_transformer.convert_to_graph_documents TypeError: list indices must be integers or slices, not str about langchain

[VJJJJJJ1]

I run the following code and faced no error:

from langchain_huggingface import HuggingFacePipeline
from langchain_text_splitters.character import RecursiveCharacterTextSplitter
from langchain.document_loaders import TextLoader
from langchain_community.graphs import Neo4jGraph
from langchain_experimental.graph_transformers.llm import LLMGraphTransformer

llm = HuggingFacePipeline.from_model_id(model_id='baichuan-inc/Baichuan2-7B-Chat',task="text-generation")

loader = TextLoader('doc.txt')
documents = loader.load() # + docx_documents
text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=800, chunk_overlap=200)
texts = text_splitter.split_documents(documents)
graph = Neo4jGraph(url = 'bolt://localhost:7687',database='neo4j',username='neo4j',password='')
llm_transformer = LLMGraphTransformer(llm=llm)
graph_documents = llm_transformer.convert_to_graph_documents(texts)

The doc txt was as follow

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.

Can you double check your input file or possible share a sample because I feel like that might be causing a problem.

this is one part of my doc samples:

卷一　建　　置-->  第一章　沿　　革-->    第一节　隶　属-->      三、中华人民共和国成立后的隶属

三、中华人民共和国成立后的隶属
1949年9月9日，晋江县人民政府成立，归属福建省人民政府第五行政督察专员公署(治设泉州城区)。1950年4月，属泉州行政督察专员公署。1950年9月，属晋江区行政督察专员公署。1955年4月，属晋江专区专员公署。1967年6月，由晋江专区军事管制委员会管辖。1968年9月，属晋江专区革命委员会。1971年7月，属晋江地区革命委员会。1980年1月，属晋江地区行政公署。1986年1月至1988年12月，属泉州市。




卷一　建　　置-->  第一章　沿　　革-->    第二节　境域析变

第二节　境域析变
晋江县位于福建省东南沿海，晋江下游。东经118°24′～118°46′，北纬24°30′～24°54′。东濒**海峡，西接南安县，南与金门隔海相望，北邻鲤城区。南北长42公里，东西宽37公里。总面积809.24平方公里。建县时境域包括今惠安县、鲤城区。宋代兼辖澎湖岛。后几经析变形成今晋江县域。




卷一　建　　置-->  第一章　沿　　革-->    第二节　境域析变-->      一、开元建县

一、开元建县
据清道光《晋江县志》记载，晋江之名“以晋南渡时，衣冠避此者多沿江而居，故名”。西晋末年，北方士族为避兵燹，纷纷南迁，部分在今晋江两岸定居，劳动生息。晋江之名，即始于此。建县时遂以江名命县。唐初，晋江一带属南安县地。景云二年(711)，改武荣州为泉州(即今泉州)，属闽州都督府。州治无县，刺史冯仁知以此为由，呈请置县。于是在唐开元六年(718)，析南安县东南部设置新县，即为晋江县之始建。县治在今鲤城区内，州县同城而治。晋江县唐建县境域示意图晋江县当代境域析变示意图

and I run this code will meet the same error:

doc = Document(page_content="Elon Musk is suing OpenAI")
graph_documents = llm_transformer.convert_to_graph_documents([doc])

maybe the version is incorrect?

[VJJJJJJ1]

this code will meet the same error:
doc = Document(page_content="Elon Musk is suing OpenAI")
graph_documents = llm_transformer.convert_to_graph_documents([doc])

[keenborder786]

what is the llm you are using?

[VJJJJJJ1]

what is the llm you are using?

baichuan2-7b-chat. Thank u for your reply

[VJJJJJJ1]

what is the llm you are using?

my code about using llm as follows:

model_path='../../Models/Baichuan2-7B-Chat'
model = HuggingFacePipeline.from_model_id(model_id=model_path,
                                              task="text-generation",
                                              model_kwargs={
                                                  "torch_dtype": load_type,
                                                  "low_cpu_mem_usage": True,
                                                  "temperature": 0.2,
                                                  "max_length": 1000,
                                                  "device_map": "auto",
                                                  "repetition_penalty": 1.1,
                                                  "trust_remote_code": True,
                                                  "quantization_config": quantization_config, }
                                              )

[keenborder786]

okay I am checking it. Will get back to you.

[keenborder786]

I run the following code and faced no error:

from langchain_huggingface import HuggingFacePipeline
from langchain_text_splitters.character import RecursiveCharacterTextSplitter
from langchain.document_loaders import TextLoader
from langchain_community.graphs import Neo4jGraph
from langchain_experimental.graph_transformers.llm import LLMGraphTransformer

llm = HuggingFacePipeline.from_model_id(model_id='baichuan-inc/Baichuan2-7B-Chat',task="text-generation")

loader = TextLoader('doc.txt')
documents = loader.load() # + docx_documents
text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=800, chunk_overlap=200)
texts = text_splitter.split_documents(documents)
graph = Neo4jGraph(url = 'bolt://localhost:7687',database='neo4j',username='neo4j',password='')
llm_transformer = LLMGraphTransformer(llm=llm)
graph_documents = llm_transformer.convert_to_graph_documents(texts)

The doc txt was as follow

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.

Can you double check your input file or possible share a sample because I feel like that might be causing a problem.

[drahmad89]

i have the same issue

[li-hhhh]

我也遇到了相同的问题，请问最后是如何解决的呢？谢谢

[dumanting]

这是来自QQ邮箱的假期自动回复邮件。   您好，我最近正在休假中，无法亲自回复您的邮件。我将在假期结束后，尽快给您回复。

[SatSadhu]

Could someone solve the error?

[VJJJJJJ1]

I have not solved this error

[VJJJJJJ1]

I run the following code and faced no error:

from langchain_huggingface import HuggingFacePipeline
from langchain_text_splitters.character import RecursiveCharacterTextSplitter
from langchain.document_loaders import TextLoader
from langchain_community.graphs import Neo4jGraph
from langchain_experimental.graph_transformers.llm import LLMGraphTransformer

llm = HuggingFacePipeline.from_model_id(model_id='baichuan-inc/Baichuan2-7B-Chat',task="text-generation")

loader = TextLoader('doc.txt')
documents = loader.load() # + docx_documents
text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=800, chunk_overlap=200)
texts = text_splitter.split_documents(documents)
graph = Neo4jGraph(url = 'bolt://localhost:7687',database='neo4j',username='neo4j',password='')
llm_transformer = LLMGraphTransformer(llm=llm)
graph_documents = llm_transformer.convert_to_graph_documents(texts)

The doc txt was as follow

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.

Can you double check your input file or possible share a sample because I feel like that might be causing a problem.

this is one part of my doc samples:

卷一　建　　置-->  第一章　沿　　革-->    第一节　隶　属-->      三、中华人民共和国成立后的隶属

三、中华人民共和国成立后的隶属
1949年9月9日，晋江县人民政府成立，归属福建省人民政府第五行政督察专员公署(治设泉州城区)。1950年4月，属泉州行政督察专员公署。1950年9月，属晋江区行政督察专员公署。1955年4月，属晋江专区专员公署。1967年6月，由晋江专区军事管制委员会管辖。1968年9月，属晋江专区革命委员会。1971年7月，属晋江地区革命委员会。1980年1月，属晋江地区行政公署。1986年1月至1988年12月，属泉州市。




卷一　建　　置-->  第一章　沿　　革-->    第二节　境域析变

第二节　境域析变
晋江县位于福建省东南沿海，晋江下游。东经118°24′～118°46′，北纬24°30′～24°54′。东濒**海峡，西接南安县，南与金门隔海相望，北邻鲤城区。南北长42公里，东西宽37公里。总面积809.24平方公里。建县时境域包括今惠安县、鲤城区。宋代兼辖澎湖岛。后几经析变形成今晋江县域。




卷一　建　　置-->  第一章　沿　　革-->    第二节　境域析变-->      一、开元建县

一、开元建县
据清道光《晋江县志》记载，晋江之名“以晋南渡时，衣冠避此者多沿江而居，故名”。西晋末年，北方士族为避兵燹，纷纷南迁，部分在今晋江两岸定居，劳动生息。晋江之名，即始于此。建县时遂以江名命县。唐初，晋江一带属南安县地。景云二年(711)，改武荣州为泉州(即今泉州)，属闽州都督府。州治无县，刺史冯仁知以此为由，呈请置县。于是在唐开元六年(718)，析南安县东南部设置新县，即为晋江县之始建。县治在今鲤城区内，州县同城而治。晋江县唐建县境域示意图晋江县当代境域析变示意图

and I run this code will meet the same error:

doc = Document(page_content="Elon Musk is suing OpenAI")
graph_documents = llm_transformer.convert_to_graph_documents([doc])

maybe the version is incorrect?

@keenborder786 could my sample cause the problem? it confuses me a long time.

[VJJJJJJ1]

I run the following code and faced no error:

from langchain_huggingface import HuggingFacePipeline
from langchain_text_splitters.character import RecursiveCharacterTextSplitter
from langchain.document_loaders import TextLoader
from langchain_community.graphs import Neo4jGraph
from langchain_experimental.graph_transformers.llm import LLMGraphTransformer

llm = HuggingFacePipeline.from_model_id(model_id='baichuan-inc/Baichuan2-7B-Chat',task="text-generation")

loader = TextLoader('doc.txt')
documents = loader.load() # + docx_documents
text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=800, chunk_overlap=200)
texts = text_splitter.split_documents(documents)
graph = Neo4jGraph(url = 'bolt://localhost:7687',database='neo4j',username='neo4j',password='')
llm_transformer = LLMGraphTransformer(llm=llm)
graph_documents = llm_transformer.convert_to_graph_documents(texts)

The doc txt was as follow

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.

Can you double check your input file or possible share a sample because I feel like that might be causing a problem.

this is one part of my doc samples:

卷一　建　　置-->  第一章　沿　　革-->    第一节　隶　属-->      三、中华人民共和国成立后的隶属

三、中华人民共和国成立后的隶属
1949年9月9日，晋江县人民政府成立，归属福建省人民政府第五行政督察专员公署(治设泉州城区)。1950年4月，属泉州行政督察专员公署。1950年9月，属晋江区行政督察专员公署。1955年4月，属晋江专区专员公署。1967年6月，由晋江专区军事管制委员会管辖。1968年9月，属晋江专区革命委员会。1971年7月，属晋江地区革命委员会。1980年1月，属晋江地区行政公署。1986年1月至1988年12月，属泉州市。




卷一　建　　置-->  第一章　沿　　革-->    第二节　境域析变

第二节　境域析变
晋江县位于福建省东南沿海，晋江下游。东经118°24′～118°46′，北纬24°30′～24°54′。东濒**海峡，西接南安县，南与金门隔海相望，北邻鲤城区。南北长42公里，东西宽37公里。总面积809.24平方公里。建县时境域包括今惠安县、鲤城区。宋代兼辖澎湖岛。后几经析变形成今晋江县域。




卷一　建　　置-->  第一章　沿　　革-->    第二节　境域析变-->      一、开元建县

一、开元建县
据清道光《晋江县志》记载，晋江之名“以晋南渡时，衣冠避此者多沿江而居，故名”。西晋末年，北方士族为避兵燹，纷纷南迁，部分在今晋江两岸定居，劳动生息。晋江之名，即始于此。建县时遂以江名命县。唐初，晋江一带属南安县地。景云二年(711)，改武荣州为泉州(即今泉州)，属闽州都督府。州治无县，刺史冯仁知以此为由，呈请置县。于是在唐开元六年(718)，析南安县东南部设置新县，即为晋江县之始建。县治在今鲤城区内，州县同城而治。晋江县唐建县境域示意图晋江县当代境域析变示意图

and I run this code will meet the same error:

doc = Document(page_content="Elon Musk is suing OpenAI")
graph_documents = llm_transformer.convert_to_graph_documents([doc])

maybe the version is incorrect?

@keenborder786 could my sample cause the problem? it confuses me a long time.

[SatSadhu]

I was able to put together a solution to the problem...
What I did was edit the 'llm.py', in your case (@VJJJJJJ1) the path is -> /root/miniconda3/envs/rag/lib/python3.10/site-packages/langchain_experimental/graph_transformers/llm.py

What I did was go to line 714 and rename the function called 'process_response' as 'process_response_old' and then create another function called 'process_response' which is the following:

def process_response(self, document: Document) -> GraphDocument: 
   """
   Processes a single document, transforming it into a graph document using
   an LLM based on the model's schema and constraints.
   """
   text = document.page_content
   raw_schema = self.chain.invoke({"input": text})
   if self._function_call:
       raw_schema = cast(Dict[Any, Any], raw_schema)
       nodes, relationships = _convert_to_graph_document(raw_schema)
   else:
       nodes_set = set()
       relationships = []
       if not isinstance(raw_schema, str):
           raw_schema = raw_schema.content
       parsed_json = self.json_repair.loads(raw_schema)
       print(parsed_json)
       
       properties = parsed_json.get('properties', {})

       # Handle different types of values: strings and lists
       def get_first_element(value):
           return value[0] if isinstance(value, list) else value

       head = get_first_element(properties.get('head', 'Unknown'))
       head_type = get_first_element(properties.get('head_type', 'Unknown'))
       tail = get_first_element(properties.get('tail', 'Unknown'))
       tail_type = get_first_element(properties.get('tail_type', 'Unknown'))
       relation = get_first_element(properties.get('relation', 'Unknown'))

       # Nodes need to be deduplicated using a set
       nodes_set.add((head, head_type))
       nodes_set.add((tail, tail_type))

       source_node = Node(id=head, type=head_type)
       target_node = Node(id=tail, type=tail_type)
       relationships.append(
           Relationship(
               source=source_node, target=target_node, type=relation
           )
       )
       
       # Create nodes list
       nodes = [Node(id=el[0], type=el[1]) for el in list(nodes_set)]

   # Strict mode filtering
   if self.strict_mode and (self.allowed_nodes or self.allowed_relationships):
       if self.allowed_nodes:
           lower_allowed_nodes = [el.lower() for el in self.allowed_nodes]
           nodes = [
               node for node in nodes if node.type.lower() in lower_allowed_nodes
           ]
           relationships = [
               rel
               for rel in relationships
               if rel.source.type.lower() in lower_allowed_nodes
               and rel.target.type.lower() in lower_allowed_nodes
           ]
       if self.allowed_relationships:
           relationships = [
               rel
               for rel in relationships
               if rel.type.lower()
               in [el.lower() for el in self.allowed_relationships]
           ]

   return GraphDocument(nodes=nodes, relationships=relationships, source=document)

I hope it has been helpfull! It worked for me...

[ingmars1709]

I had the same error when using llama2:

nodes_set.add((rel["head"], rel["head_type"]))
TypeError: list indices must be integers or slices, not str

When I changed the model string variable to 'llama3' I got correct results using the old implementation of process_response using the following set-up:

loader = TextLoader('doc.txt')
documents = loader.load()

text_splitter = RecursiveCharacterTextSplitter(chunk_size=800, chunk_overlap=200)
texts = text_splitter.split_documents(documents)

llm=ChatOllama(model="llama3") # make sure you run llama3 model

llm_transformer = LLMGraphTransformer(llm=llm)
graph_documents = llm_transformer.convert_to_graph_documents(texts)

print(f"Nodes:{graph_documents[0].nodes}")
print(f"Relationships:{graph_documents[0].relationships}")

[Otnielush]

It happens when extracting information from LLM output with input prompt (for me it was by default).
can be fixed by:

pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=2000, return_full_text=True)
llm = HuggingFacePipeline(pipeline=pipe)