Spaces:
Runtime error
Runtime error
File size: 5,217 Bytes
cbdf795 | 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 | #!/usr/bin/env python
import json
import logging
import os
import sys
import psycopg2
import s3fs
import torch
from langchain.embeddings.huggingface import HuggingFaceEmbeddings
from llama_index import (ServiceContext, SimpleDirectoryReader, StorageContext,
SummaryIndex, get_response_synthesizer,
set_global_service_context)
from llama_index.indices.document_summary import DocumentSummaryIndex
from llama_index.indices.vector_store import VectorStoreIndex
from llama_index.llms import OpenAI
from llama_index.schema import IndexNode
from llama_index.vector_stores import PGVectorStore
from sqlalchemy import make_url
# logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)
# logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))
def get_embed_model():
model_kwargs = {'device': 'cpu'}
if torch.cuda.is_available():
model_kwargs['device'] = 'cuda'
if torch.backends.mps.is_available():
model_kwargs['device'] = 'mps'
encode_kwargs = {'normalize_embeddings': True} # set True to compute cosine similarity
print("Loading model...")
try:
model_norm = HuggingFaceEmbeddings(
model_name="thenlper/gte-small",
model_kwargs=model_kwargs,
encode_kwargs=encode_kwargs,
)
except Exception as exception:
print(f"Model not found. Loading fake model...{exception}")
exit()
print("Model loaded.")
return model_norm
def create_table(db_name, connection_string):
conn = psycopg2.connect(connection_string)
conn.autocommit = True
with conn.cursor() as c:
c.execute(f"DROP DATABASE IF EXISTS {db_name}")
c.execute(f"CREATE DATABASE {db_name}")
return
def create_vector_store():
db_name = "helm"
connection_string = "postgresql://adrian@localhost:5432/postgres"
create_table(db_name, connection_string)
url = make_url(connection_string)
vector_store = PGVectorStore.from_params(
database=db_name,
host=url.host,
password=url.password,
port=url.port,
user=url.username,
table_name="f150_manual",
embed_dim=384,
hybrid_search=True,
text_search_config="english",
)
return vector_store
def get_remote_filesystem():
AWS_KEY = "AKIAWCUHDQXX3H7PPRXN"
AWS_SECRET = "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"
s3 = s3fs.S3FileSystem(
key=AWS_KEY,
secret=AWS_SECRET,
)
return s3
def create_vector_index():
docs = SimpleDirectoryReader(input_dir="docs/chapters").load_data()
vector_store = create_vector_store()
storage_context = StorageContext.from_defaults(vector_store=vector_store)
vector_index = VectorStoreIndex.from_documents(
docs,
storage_context=storage_context,
embedding_model=None,
show_progress=True,
chunk_size=1024,
chunk_overlap=20)
return vector_index
def create_recursive_index():
doc_dir = "./docs/chapters/"
doc_summaries = {}
titles = []
for filename in os.listdir(doc_dir):
print(filename)
title = filename.split(".")[0]
titles.append(title)
docs = SimpleDirectoryReader(input_files=[f"{doc_dir}{filename}"]).load_data()
docs[0].doc_id = title
doc_summaries[title] = docs
context_window = 4096
embed_model = get_embed_model()
chatgpt = OpenAI(temperature=0, model="gpt-3.5-turbo-16k")
service_context = ServiceContext.from_defaults(
llm=chatgpt,
embed_model=embed_model,
chunk_size=1024,
context_window=context_window)
s3 = get_remote_filesystem()
nodes = []
for title in titles:
print(title)
# build vector index
storage_context = StorageContext.from_defaults()
vector_index = VectorStoreIndex.from_documents(
doc_summaries[title],
service_context=service_context,
verbose=True,
storage_context=storage_context,
show_progress=True,
)
vector_index.storage_context.persist(f"f150-user-manual/recursive-agent/{title}/vector_index", fs=s3)
# build summary index
response_synthesizer = get_response_synthesizer(
response_mode="compact_accumulate", use_async=False
)
storage_context = StorageContext.from_defaults()
summary_index = DocumentSummaryIndex.from_documents(
doc_summaries[title],
service_context=service_context,
response_synthesizer=response_synthesizer,
verbose=True,
storage_context=storage_context,
show_progress=True,
)
print(summary_index.get_document_summary(title))
node = IndexNode(text=summary_index.get_document_summary(title), index_id=title)
nodes.append(node)
storage_context = StorageContext.from_defaults()
vector_index = VectorStoreIndex(
nodes,
service_context=service_context,
verbose=True,
storage_context=storage_context,
show_progress=True,)
vector_index.storage_context.persist("f150-user-manual/recursive-agent/vector_index", fs=s3)
def main():
embed_model = get_embed_model()
service_context = ServiceContext.from_defaults(embed_model=embed_model)
set_global_service_context(service_context)
create_vector_index();
create_recursive_index();
if __name__ == "__main__":
main() |