langchain/langchain/chains/graph_qa/cypher.py
Tomaz Bratanic fd866d1801
Update Cypher QA prompt (#5173)
# Improve Cypher QA prompt

The current QA prompt is optimized for networkX answer generation, which
returns all the possible triples.
However, Cypher search is a bit more focused and doesn't necessary
return all the context information.
Due to that reason, the model sometimes refuses to generate an answer
even though the information is provided:

![Screenshot from 2023-05-24
08-36-23](https://github.com/hwchase17/langchain/assets/19948365/351cf9c1-2567-447c-91fd-284ae3fa1ccf)


To fix this issue, I have updated the prompt. Interestingly, I tried
many variations with less instructions and they didn't work properly.
However, the current fix works nicely.
![Screenshot from 2023-05-24
08-37-25](https://github.com/hwchase17/langchain/assets/19948365/fc830603-e6ec-4a23-8a86-eaf572996014)
2023-05-24 08:31:30 -07:00

91 lines
2.9 KiB
Python

"""Question answering over a graph."""
from __future__ import annotations
from typing import Any, Dict, List, Optional
from pydantic import Field
from langchain.base_language import BaseLanguageModel
from langchain.callbacks.manager import CallbackManagerForChainRun
from langchain.chains.base import Chain
from langchain.chains.graph_qa.prompts import CYPHER_GENERATION_PROMPT, CYPHER_QA_PROMPT
from langchain.chains.llm import LLMChain
from langchain.graphs.neo4j_graph import Neo4jGraph
from langchain.prompts.base import BasePromptTemplate
class GraphCypherQAChain(Chain):
"""Chain for question-answering against a graph by generating Cypher statements."""
graph: Neo4jGraph = Field(exclude=True)
cypher_generation_chain: LLMChain
qa_chain: LLMChain
input_key: str = "query" #: :meta private:
output_key: str = "result" #: :meta private:
@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
@classmethod
def from_llm(
cls,
llm: BaseLanguageModel,
*,
qa_prompt: BasePromptTemplate = CYPHER_QA_PROMPT,
cypher_prompt: BasePromptTemplate = CYPHER_GENERATION_PROMPT,
**kwargs: Any,
) -> GraphCypherQAChain:
"""Initialize from LLM."""
qa_chain = LLMChain(llm=llm, prompt=qa_prompt)
cypher_generation_chain = LLMChain(llm=llm, prompt=cypher_prompt)
return cls(
qa_chain=qa_chain,
cypher_generation_chain=cypher_generation_chain,
**kwargs,
)
def _call(
self,
inputs: Dict[str, Any],
run_manager: Optional[CallbackManagerForChainRun] = None,
) -> Dict[str, str]:
"""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]
generated_cypher = self.cypher_generation_chain.run(
{"question": question, "schema": self.graph.get_schema}, callbacks=callbacks
)
_run_manager.on_text("Generated Cypher:", end="\n", verbose=self.verbose)
_run_manager.on_text(
generated_cypher, color="green", end="\n", verbose=self.verbose
)
context = self.graph.query(generated_cypher)
_run_manager.on_text("Full Context:", end="\n", verbose=self.verbose)
_run_manager.on_text(
str(context), color="green", end="\n", verbose=self.verbose
)
result = self.qa_chain(
{"question": question, "context": context},
callbacks=callbacks,
)
return {self.output_key: result[self.qa_chain.output_key]}