"""Question answering over a graph."""
from __future__ import annotations
import re
from typing import Any, Dict, List, Optional, Union
from langchain.chains.base import Chain
from langchain.chains.llm import LLMChain
from langchain_core.callbacks import CallbackManagerForChainRun
from langchain_core.language_models import BaseLanguageModel
from langchain_core.messages import (
AIMessage,
BaseMessage,
SystemMessage,
ToolMessage,
)
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import (
BasePromptTemplate,
ChatPromptTemplate,
HumanMessagePromptTemplate,
MessagesPlaceholder,
)
from langchain_core.runnables import Runnable
from pydantic import Field
from langchain_community.chains.graph_qa.cypher_utils import (
CypherQueryCorrector,
Schema,
)
from langchain_community.chains.graph_qa.prompts import (
CYPHER_GENERATION_PROMPT,
CYPHER_QA_PROMPT,
)
from langchain_community.graphs.graph_store import GraphStore
INTERMEDIATE_STEPS_KEY = "intermediate_steps"
FUNCTION_RESPONSE_SYSTEM = """You are an assistant that helps to form nice and human
understandable answers based on the provided information from tools.
Do not add any other information that wasn't present in the tools, and use
very concise style in interpreting results!
"""
[docs]
def construct_schema(
structured_schema: Dict[str, Any],
include_types: List[str],
exclude_types: List[str],
) -> str:
"""Filter the schema based on included or excluded types"""
def filter_func(x: str) -> bool:
return x in include_types if include_types else x not in exclude_types
filtered_schema: Dict[str, Any] = {
"node_props": {
k: v
for k, v in structured_schema.get("node_props", {}).items()
if filter_func(k)
},
"rel_props": {
k: v
for k, v in structured_schema.get("rel_props", {}).items()
if filter_func(k)
},
"relationships": [
r
for r in structured_schema.get("relationships", [])
if all(filter_func(r[t]) for t in ["start", "end", "type"])
],
}
# Format node properties
formatted_node_props = []
for label, properties in filtered_schema["node_props"].items():
props_str = ", ".join(
[f"{prop['property']}: {prop['type']}" for prop in properties]
)
formatted_node_props.append(f"{label} {{{props_str}}}")
# Format relationship properties
formatted_rel_props = []
for rel_type, properties in filtered_schema["rel_props"].items():
props_str = ", ".join(
[f"{prop['property']}: {prop['type']}" for prop in properties]
)
formatted_rel_props.append(f"{rel_type} {{{props_str}}}")
# Format relationships
formatted_rels = [
f"(:{el['start']})-[:{el['type']}]->(:{el['end']})"
for el in filtered_schema["relationships"]
]
return "\n".join(
[
"Node properties are the following:",
",".join(formatted_node_props),
"Relationship properties are the following:",
",".join(formatted_rel_props),
"The relationships are the following:",
",".join(formatted_rels),
]
)
[docs]
def get_function_response(
question: str, context: List[Dict[str, Any]]
) -> List[BaseMessage]:
TOOL_ID = "call_H7fABDuzEau48T10Qn0Lsh0D"
messages = [
AIMessage(
content="",
additional_kwargs={
"tool_calls": [
{
"id": TOOL_ID,
"function": {
"arguments": '{"question":"' + question + '"}',
"name": "GetInformation",
},
"type": "function",
}
]
},
),
ToolMessage(content=str(context), tool_call_id=TOOL_ID),
]
return messages
[docs]
class GraphCypherQAChain(Chain):
"""Chain for question-answering against a graph by generating Cypher statements.
*Security note*: Make sure that the database connection uses credentials
that are narrowly-scoped to only include necessary permissions.
Failure to do so may result in data corruption or loss, since the calling
code may attempt commands that would result in deletion, mutation
of data if appropriately prompted or reading sensitive data if such
data is present in the database.
The best way to guard against such negative outcomes is to (as appropriate)
limit the permissions granted to the credentials used with this tool.
See https://python.langchain.com/docs/security for more information.
"""
graph: GraphStore = Field(exclude=True)
cypher_generation_chain: LLMChain
qa_chain: Union[LLMChain, Runnable]
graph_schema: str
input_key: str = "query" #: :meta private:
output_key: str = "result" #: :meta private:
top_k: int = 10
"""Number of results to return from the query"""
return_intermediate_steps: bool = False
"""Whether or not to return the intermediate steps along with the final answer."""
return_direct: bool = False
"""Whether or not to return the result of querying the graph directly."""
cypher_query_corrector: Optional[CypherQueryCorrector] = None
"""Optional cypher validation tool"""
use_function_response: bool = False
"""Whether to wrap the database context as tool/function response"""
allow_dangerous_requests: bool = False
"""Forced user opt-in to acknowledge that the chain can make dangerous requests.
*Security note*: Make sure that the database connection uses credentials
that are narrowly-scoped to only include necessary permissions.
Failure to do so may result in data corruption or loss, since the calling
code may attempt commands that would result in deletion, mutation
of data if appropriately prompted or reading sensitive data if such
data is present in the database.
The best way to guard against such negative outcomes is to (as appropriate)
limit the permissions granted to the credentials used with this tool.
See https://python.langchain.com/docs/security for more information.
"""
def __init__(self, **kwargs: Any) -> None:
"""Initialize the chain."""
super().__init__(**kwargs)
if self.allow_dangerous_requests is not True:
raise ValueError(
"In order to use this chain, you must acknowledge that it can make "
"dangerous requests by setting `allow_dangerous_requests` to `True`."
"You must narrowly scope the permissions of the database connection "
"to only include necessary permissions. Failure to do so may result "
"in data corruption or loss or reading sensitive data if such data is "
"present in the database."
"Only use this chain if you understand the risks and have taken the "
"necessary precautions. "
"See https://python.langchain.com/docs/security for more information."
)
@property
def input_keys(self) -> List[str]:
"""Return the input keys.
:meta private:
"""
return [self.input_key]
@property
def output_keys(self) -> List[str]:
"""Return the output keys.
:meta private:
"""
_output_keys = [self.output_key]
return _output_keys
@property
def _chain_type(self) -> str:
return "graph_cypher_chain"
[docs]
@classmethod
def from_llm(
cls,
llm: Optional[BaseLanguageModel] = None,
*,
qa_prompt: Optional[BasePromptTemplate] = None,
cypher_prompt: Optional[BasePromptTemplate] = None,
cypher_llm: Optional[BaseLanguageModel] = None,
qa_llm: Optional[Union[BaseLanguageModel, Any]] = None,
exclude_types: List[str] = [],
include_types: List[str] = [],
validate_cypher: bool = False,
qa_llm_kwargs: Optional[Dict[str, Any]] = None,
cypher_llm_kwargs: Optional[Dict[str, Any]] = None,
use_function_response: bool = False,
function_response_system: str = FUNCTION_RESPONSE_SYSTEM,
**kwargs: Any,
) -> GraphCypherQAChain:
"""Initialize from LLM."""
if not cypher_llm and not llm:
raise ValueError("Either `llm` or `cypher_llm` parameters must be provided")
if not qa_llm and not llm:
raise ValueError("Either `llm` or `qa_llm` parameters must be provided")
if cypher_llm and qa_llm and llm:
raise ValueError(
"You can specify up to two of 'cypher_llm', 'qa_llm'"
", and 'llm', but not all three simultaneously."
)
if cypher_prompt and cypher_llm_kwargs:
raise ValueError(
"Specifying cypher_prompt and cypher_llm_kwargs together is"
" not allowed. Please pass prompt via cypher_llm_kwargs."
)
if qa_prompt and qa_llm_kwargs:
raise ValueError(
"Specifying qa_prompt and qa_llm_kwargs together is"
" not allowed. Please pass prompt via qa_llm_kwargs."
)
use_qa_llm_kwargs = qa_llm_kwargs if qa_llm_kwargs is not None else {}
use_cypher_llm_kwargs = (
cypher_llm_kwargs if cypher_llm_kwargs is not None else {}
)
if "prompt" not in use_qa_llm_kwargs:
use_qa_llm_kwargs["prompt"] = (
qa_prompt if qa_prompt is not None else CYPHER_QA_PROMPT
)
if "prompt" not in use_cypher_llm_kwargs:
use_cypher_llm_kwargs["prompt"] = (
cypher_prompt if cypher_prompt is not None else CYPHER_GENERATION_PROMPT
)
qa_llm = qa_llm or llm
if use_function_response:
try:
qa_llm.bind_tools({}) # type: ignore[union-attr]
response_prompt = ChatPromptTemplate.from_messages(
[
SystemMessage(content=function_response_system),
HumanMessagePromptTemplate.from_template("{question}"),
MessagesPlaceholder(variable_name="function_response"),
]
)
qa_chain = response_prompt | qa_llm | StrOutputParser() # type: ignore
except (NotImplementedError, AttributeError):
raise ValueError("Provided LLM does not support native tools/functions")
else:
qa_chain = LLMChain(llm=qa_llm, **use_qa_llm_kwargs) # type: ignore[arg-type]
cypher_generation_chain = LLMChain(
llm=cypher_llm or llm, # type: ignore[arg-type]
**use_cypher_llm_kwargs, # type: ignore[arg-type]
)
if exclude_types and include_types:
raise ValueError(
"Either `exclude_types` or `include_types` "
"can be provided, but not both"
)
graph_schema = construct_schema(
kwargs["graph"].get_structured_schema, include_types, exclude_types
)
cypher_query_corrector = None
if validate_cypher:
corrector_schema = [
Schema(el["start"], el["type"], el["end"])
for el in kwargs["graph"].structured_schema.get("relationships")
]
cypher_query_corrector = CypherQueryCorrector(corrector_schema)
return cls(
graph_schema=graph_schema,
qa_chain=qa_chain,
cypher_generation_chain=cypher_generation_chain,
cypher_query_corrector=cypher_query_corrector,
use_function_response=use_function_response,
**kwargs,
)
def _call(
self,
inputs: Dict[str, Any],
run_manager: Optional[CallbackManagerForChainRun] = None,
) -> Dict[str, Any]:
"""Generate Cypher statement, use it to look up in db and answer question."""
_run_manager = run_manager or CallbackManagerForChainRun.get_noop_manager()
callbacks = _run_manager.get_child()
question = inputs[self.input_key]
args = {
"question": question,
"schema": self.graph_schema,
}
args.update(inputs)
intermediate_steps: List = []
generated_cypher = self.cypher_generation_chain.run(args, callbacks=callbacks)
# Extract Cypher code if it is wrapped in backticks
generated_cypher = extract_cypher(generated_cypher)
# Correct Cypher query if enabled
if self.cypher_query_corrector:
generated_cypher = self.cypher_query_corrector(generated_cypher)
_run_manager.on_text("Generated Cypher:", end="\n", verbose=self.verbose)
_run_manager.on_text(
generated_cypher, color="green", end="\n", verbose=self.verbose
)
intermediate_steps.append({"query": generated_cypher})
# Retrieve and limit the number of results
# Generated Cypher be null if query corrector identifies invalid schema
if generated_cypher:
context = self.graph.query(generated_cypher)[: self.top_k]
else:
context = []
if self.return_direct:
final_result = context
else:
_run_manager.on_text("Full Context:", end="\n", verbose=self.verbose)
_run_manager.on_text(
str(context), color="green", end="\n", verbose=self.verbose
)
intermediate_steps.append({"context": context})
if self.use_function_response:
function_response = get_function_response(question, context)
final_result = self.qa_chain.invoke( # type: ignore
{"question": question, "function_response": function_response},
)
else:
result = self.qa_chain.invoke( # type: ignore
{"question": question, "context": context},
callbacks=callbacks,
)
final_result = result[self.qa_chain.output_key] # type: ignore
chain_result: Dict[str, Any] = {self.output_key: final_result}
if self.return_intermediate_steps:
chain_result[INTERMEDIATE_STEPS_KEY] = intermediate_steps
return chain_result
