LangChain RdfLib-sparql error

[apollosbangalu]

Please I appreciate any good help.

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

import os
from dotenv import load_dotenv
from dotenv import load_dotenv
from langchain.chains import GraphSparqlQAChain
from langchain.graphs import RdfGraph
from rdflib import Graph
from langchain_openai import ChatOpenAI

# Load environment variables
load_dotenv()

# Initialize the ChatOpenAI model
llm = ChatOpenAI(model_name="gpt-4-0613", temperature=0)

def clean_sparql_query(query):
    """
    Remove backticks from the generated SPARQL query.
    """
    return query.replace('`', '')

def get_user_input():
    """
    Get the source URI from command line input.
    """
    return input("Enter the source URI for the RDF graph: ")

def create_qa_chain(source_uri):
    """
    Create and return the GraphSparqlQAChain.
    """
    graph = RdfGraph(source_file=source_uri)
    print("\nFirst 3 triples in the graph:")
    for i, triple in enumerate(graph.graph):
        if i < 3:
            print(triple)
        else:
            break
    
    chain = GraphSparqlQAChain.from_llm(
        llm, 
        graph=graph, 
        verbose=True,
        return_intermediate_steps=True,
        return_direct=True
    )
    
    # Create a wrapper function to clean the SPARQL query
    def cleaned_chain(query):
        result = chain(query)
        if isinstance(result, dict) and 'intermediate_steps' in result:
            result['intermediate_steps'][0]['query'] = clean_sparql_query(result['intermediate_steps'][0]['query'])
        return result
    
    return cleaned_chain, graph

def refresh_graph(graph, source_uri):
    """
    Refresh the RDF graph with the latest data from the source URI.
    """
    new_graph = RdfGraph(source_file=source_uri)
    graph.graph = new_graph.graph
    print("\nGraph refreshed. First 3 triples in the updated graph:")
    for i, triple in enumerate(graph.graph):
        if i < 3:
            print(triple)
        else:
            break

def main():
    """
    Main function to run the RDFLib SPARQL app.
    """
    source_uri = get_user_input()
    qa_chain, graph = create_qa_chain(source_uri)
    
    if not graph or not qa_chain:
        print("Error: Failed to initialize the graph or QA chain. Cannot proceed with questions.")
        return

    print("\nGraph and QA chain initialized successfully. You can now ask questions about the RDF graph.")
    
    while True:
        refresh = input("\nDo you want to refresh the graph? (yes/no): ").lower()
        if refresh == 'yes':
            refresh_graph(graph, source_uri)
            qa_chain, _ = create_qa_chain(source_uri)
        
        user_input = input("\nEnter your question or 'quit' to exit: ")
        if user_input.lower() == 'quit':
            break
        else:
            result = qa_chain(user_input)
            print("\nAnswer:", result['result'] if isinstance(result, dict) else result)

if __name__ == "__main__":
    main()

Description

I am trying to implement the exaple at https://python.langchain.com/v0.2/docs/integrations/graphs/rdflib_sparql/ but I always get the error
Enter your question or 'quit' to exit: who is Bernes-Lee

Entering new GraphSparqlQAChain chain...
Identified intent:
SELECT
Generated SPARQL:

PREFIX foaf: <http://xmlns.com/foaf/0.1/>
SELECT ?person
WHERE {
    ?person foaf:name "Bernes-Lee" .
}

Traceback (most recent call last):
File "E:\My Python Works\WebRdfQaApp\main.py", line 96, in
main()
File "E:\My Python Works\WebRdfQaApp\main.py", line 92, in main
result = qa_chain(user_input)
^^^^^^^^^^^^^^^^^^^^
File "E:\My Python Works\WebRdfQaApp\main.py", line 49, in cleaned_chain
result = chain(query)
^^^^^^^^^^^^
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\langchain\chains\base.py", line 312, in call
raise e
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\langchain\chains\base.py", line 306, in call
self._call(inputs, run_manager=run_manager)
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\langchain\chains\graph_qa\sparql.py", line 122, in _call
context = self.graph.query(generated_sparql)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\langchain_community\graphs\rdf_graph.py", line 208, in query
res = self.graph.query(query)
^^^^^^^^^^^^^^^^^^^^^^^
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\rdflib\graph.py", line 1567, in query
return result(processor.query(query_object, initBindings, initNs, **kwargs)) # type: ignore[arg-type]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\rdflib\plugins\sparql\processor.py", line 136, in query
parsetree = parseQuery(strOrQuery)
^^^^^^^^^^^^^^^^^^^^^^
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\rdflib\plugins\sparql\parser.py", line 1542, in parseQuery
return Query.parseString(q, parseAll=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\pyparsing\util.py", line 256, in _inner
return fn(self, *args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\My Python Works\WebRdfQaApp\install4aider-env\Lib\site-packages\pyparsing\core.py", line 1200, in parse_string
raise exc.with_traceback(None)
pyparsing.exceptions.ParseException: Expected {SelectQuery | ConstructQuery | DescribeQuery | AskQuery}, found '`' (at char 0), (line:1, col:1)

System Info

aiohttp==3.9.5
aiosignal==1.3.1
annotated-types==0.7.0
anyio==4.4.0
attrs==23.2.0
certifi==2024.7.4
charset-normalizer==3.3.2
colorama==0.4.6
dataclasses-json==0.6.7
distro==1.9.0
frozenlist==1.4.1
greenlet==3.0.3
h11==0.14.0
httpcore==1.0.5
httpx==0.27.0
idna==3.7
install==1.3.5
isodate==0.6.1
jsonpatch==1.33
jsonpointer==3.0.0
langchain==0.0.352
langchain-community==0.0.10
langchain-core==0.1.23
langchain-openai==0.0.2
langchain-text-splitters==0.2.2
langsmith==0.0.87
marshmallow==3.21.3
multidict==6.0.5
mypy-extensions==1.0.0
numpy==1.26.4
openai==1.35.13
orjson==3.10.6
packaging==23.2
pydantic==2.8.2
pydantic_core==2.20.1
pyparsing==3.1.2
python-dotenv==1.0.0
PyYAML==6.0.1
rdflib==6.3.2
regex==2024.5.15
requests==2.32.3
six==1.16.0
sniffio==1.3.1
SQLAlchemy==2.0.31
tenacity==8.5.0
tiktoken==0.5.2
tqdm==4.66.4
typing-inspect==0.9.0
typing_extensions==4.12.2
urllib3==2.2.2
yarl==1.9.4

[dosubot[bot]]

Hey @apollosbangalu! I'm here to help you with any coding issues or questions you may have. Let's get that pesky bug sorted out together!

To resolve the parsing error in your SPARQL query, you can use the _get_prepared_sparql_query method, which attempts to prepare the SPARQL query and, if it fails, uses the sparql_fix_chain to correct the query. Here is the relevant code snippet:

def _get_prepared_sparql_query(
    self,
    _run_manager: CallbackManagerForChainRun,
    callbacks: CallbackManager,
    generated_sparql: str,
    ontology_schema: str,
) -> str:
    try:
        return self._prepare_sparql_query(_run_manager, generated_sparql)
    except Exception as e:
        retries = 0
        error_message = str(e)
        self._log_invalid_sparql_query(
            _run_manager, generated_sparql, error_message
        )

        while retries < self.max_fix_retries:
            try:
                sparql_fix_chain_result = self.sparql_fix_chain.invoke(
                    {
                        "error_message": error_message,
                        "generated_sparql": generated_sparql,
                        "schema": ontology_schema,
                    },
                    callbacks=callbacks,
                )
                generated_sparql = sparql_fix_chain_result[
                    self.sparql_fix_chain.output_key
                ]
                return self._prepare_sparql_query(_run_manager, generated_sparql)
            except Exception as e:
                retries += 1
                parse_exception = str(e)
                self._log_invalid_sparql_query(
                    _run_manager, generated_sparql, parse_exception
                )

    raise ValueError("The generated SPARQL query is invalid.")

This method will attempt to prepare the SPARQL query using prepareQuery. If it encounters an error, it will log the invalid query and use the sparql_fix_chain to attempt to correct the query up to a maximum number of retries (max_fix_retries). If it still fails after the retries, it raises a ValueError.

Additionally, ensure that your SPARQL query is a valid CONSTRUCT query, as the OntotextGraphDBGraph class only supports CONSTRUCT queries for ontology schema operations. You can validate your query using the _validate_user_query method:

@staticmethod
def _validate_user_query(query_ontology: str) -> None:
    """
    Validate the query is a valid SPARQL CONSTRUCT query
    """
    from pyparsing import ParseException
    from rdflib.plugins.sparql import prepareQuery

    if not isinstance(query_ontology, str):
        raise TypeError("Ontology query must be provided as string.")
    try:
        parsed_query = prepareQuery(query_ontology)
    except ParseException as e:
        raise ValueError("Ontology query is not a valid SPARQL query.", e)

    if parsed_query.algebra.name != "ConstructQuery":
        raise ValueError(
            "Invalid query type. Only CONSTRUCT queries are supported."
        )

Make sure your query conforms to these requirements to avoid parsing errors.

Additionally, you can use the GRAPHDB_SPARQL_FIX_TEMPLATE from LangChain to help correct the SPARQL query. Here is the template:

GRAPHDB_SPARQL_FIX_TEMPLATE = """
This following SPARQL query delimited by triple backticks

{generated_sparql}

is not valid.
The error delimited by triple backticks is

{error_message}

Give me a correct version of the SPARQL query.
Do not change the logic of the query.
Do not include any explanations or apologies in your responses.
Do not wrap the query in backticks.
Do not include any text except the SPARQL query generated.
The ontology schema delimited by triple backticks in Turtle format is:

{schema}

"""

This template can be used to generate a corrected version of the SPARQL query by providing the invalid query, the error message, and the ontology schema.

For practical examples and tests, you can refer to the integration tests provided in the LangChain repository, which demonstrate how to handle various scenarios, including valid and invalid SPARQL queries, and how the sparql_fix_chain is utilized to correct them ^[1][2][3][4].

To continue talking to Dosu, mention @dosu.

[anubhutisingh-19]

@apollosbangalu Hey, I am facing the same issue and was wondering if you were able to solve this or find any workaround. Any help would be appreciated.

[nadgeri14]

I think the problem is the LLM generating this, so I devised this workaround.

instruction = "ADDITIONAL INSTRUCTION: When you make the final query, remove these ``` quotes and only have the query \n"
query = "QUERY: What is Tim Berners-Lee's work homepage?"
prompt = instruction + query
print(prompt)
chain.run(prompt)

[anubhutisingh-19]

Thank you so much! This isn't giving me any errors, but giving out blank result now. Did you face the same issue?

[nadgeri14]

No, I get the desired result. The result is highly dependent on the underlying LLM and its complexity. You could try multiple times by changing the query prompt.

[julfr]

@nadgeri14 thank you so much, i faced the same issue (i use just "gpt-3.5-turbo" which is by default) and your workaround helpd