Spaces:
Runtime error
Runtime error
File size: 5,797 Bytes
ed4d993 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 |
from pathlib import Path
from typing import List
from langchain_community.document_loaders import TextLoader
from langchain_community.graphs import Neo4jGraph
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.pydantic_v1 import BaseModel, Field
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain_text_splitters import TokenTextSplitter
from neo4j.exceptions import ClientError
txt_path = Path(__file__).parent / "dune.txt"
graph = Neo4jGraph()
# Embeddings & LLM models
embeddings = OpenAIEmbeddings()
embedding_dimension = 1536
llm = ChatOpenAI(temperature=0)
# Load the text file
loader = TextLoader(str(txt_path))
documents = loader.load()
# Ingest Parent-Child node pairs
parent_splitter = TokenTextSplitter(chunk_size=512, chunk_overlap=24)
child_splitter = TokenTextSplitter(chunk_size=100, chunk_overlap=24)
parent_documents = parent_splitter.split_documents(documents)
for i, parent in enumerate(parent_documents):
child_documents = child_splitter.split_documents([parent])
params = {
"parent_text": parent.page_content,
"parent_id": i,
"parent_embedding": embeddings.embed_query(parent.page_content),
"children": [
{
"text": c.page_content,
"id": f"{i}-{ic}",
"embedding": embeddings.embed_query(c.page_content),
}
for ic, c in enumerate(child_documents)
],
}
# Ingest data
graph.query(
"""
MERGE (p:Parent {id: $parent_id})
SET p.text = $parent_text
WITH p
CALL db.create.setVectorProperty(p, 'embedding', $parent_embedding)
YIELD node
WITH p
UNWIND $children AS child
MERGE (c:Child {id: child.id})
SET c.text = child.text
MERGE (c)<-[:HAS_CHILD]-(p)
WITH c, child
CALL db.create.setVectorProperty(c, 'embedding', child.embedding)
YIELD node
RETURN count(*)
""",
params,
)
# Create vector index for child
try:
graph.query(
"CALL db.index.vector.createNodeIndex('parent_document', "
"'Child', 'embedding', $dimension, 'cosine')",
{"dimension": embedding_dimension},
)
except ClientError: # already exists
pass
# Create vector index for parents
try:
graph.query(
"CALL db.index.vector.createNodeIndex('typical_rag', "
"'Parent', 'embedding', $dimension, 'cosine')",
{"dimension": embedding_dimension},
)
except ClientError: # already exists
pass
# Ingest hypothethical questions
class Questions(BaseModel):
"""Generating hypothetical questions about text."""
questions: List[str] = Field(
...,
description=(
"Generated hypothetical questions based on " "the information from the text"
),
)
questions_prompt = ChatPromptTemplate.from_messages(
[
(
"system",
(
"You are generating hypothetical questions based on the information "
"found in the text. Make sure to provide full context in the generated "
"questions."
),
),
(
"human",
(
"Use the given format to generate hypothetical questions from the "
"following input: {input}"
),
),
]
)
question_chain = questions_prompt | llm.with_structured_output(Questions)
for i, parent in enumerate(parent_documents):
questions = question_chain.invoke(parent.page_content).questions
params = {
"parent_id": i,
"questions": [
{"text": q, "id": f"{i}-{iq}", "embedding": embeddings.embed_query(q)}
for iq, q in enumerate(questions)
if q
],
}
graph.query(
"""
MERGE (p:Parent {id: $parent_id})
WITH p
UNWIND $questions AS question
CREATE (q:Question {id: question.id})
SET q.text = question.text
MERGE (q)<-[:HAS_QUESTION]-(p)
WITH q, question
CALL db.create.setVectorProperty(q, 'embedding', question.embedding)
YIELD node
RETURN count(*)
""",
params,
)
# Create vector index
try:
graph.query(
"CALL db.index.vector.createNodeIndex('hypothetical_questions', "
"'Question', 'embedding', $dimension, 'cosine')",
{"dimension": embedding_dimension},
)
except ClientError: # already exists
pass
# Ingest summaries
summary_prompt = ChatPromptTemplate.from_messages(
[
(
"system",
(
"You are generating concise and accurate summaries based on the "
"information found in the text."
),
),
(
"human",
("Generate a summary of the following input: {question}\n" "Summary:"),
),
]
)
summary_chain = summary_prompt | llm
for i, parent in enumerate(parent_documents):
summary = summary_chain.invoke({"question": parent.page_content}).content
params = {
"parent_id": i,
"summary": summary,
"embedding": embeddings.embed_query(summary),
}
graph.query(
"""
MERGE (p:Parent {id: $parent_id})
MERGE (p)-[:HAS_SUMMARY]->(s:Summary)
SET s.text = $summary
WITH s
CALL db.create.setVectorProperty(s, 'embedding', $embedding)
YIELD node
RETURN count(*)
""",
params,
)
# Create vector index
try:
graph.query(
"CALL db.index.vector.createNodeIndex('summary', "
"'Summary', 'embedding', $dimension, 'cosine')",
{"dimension": embedding_dimension},
)
except ClientError: # already exists
pass
|