File size: 6,910 Bytes
9c1e8a7
 
5a84661
 
 
 
 
 
 
 
9c1e8a7
 
5a84661
9c1e8a7
 
 
 
 
 
5a84661
 
129499e
5a84661
 
 
 
 
 
 
 
 
9c1e8a7
5a84661
 
 
9c1e8a7
5a84661
 
 
 
 
 
 
 
9c1e8a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5a84661
 
 
 
 
 
9c1e8a7
 
 
 
 
5a84661
 
 
9c1e8a7
5a84661
 
 
 
 
8798577
9c1e8a7
5a84661
 
 
 
 
9c1e8a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5a84661
 
 
9c1e8a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
680fe32
5a84661
 
 
 
 
 
139a897
 
 
5a84661
680fe32
1d11211
9c1e8a7
 
da1bd08
5a84661
 
 
139a897
 
 
 
680fe32
1d11211
9c1e8a7
 
da1bd08
5a84661
 
129499e
 
 
 
 
5a84661
 
9c1e8a7
 
 
 
 
 
 
129499e
5a84661
 
 
 
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
203
204
205
206
207
208
209
210
211
212
213
214
215
import asyncio
import json
import logging
import os
import pickle

import chromadb
import logfire
from custom_retriever import CustomRetriever
from dotenv import load_dotenv
from llama_index.core import Document, SimpleKeywordTableIndex, VectorStoreIndex
from llama_index.core.ingestion import IngestionPipeline
from llama_index.core.node_parser import SentenceSplitter
from llama_index.core.retrievers import (
    KeywordTableSimpleRetriever,
    VectorIndexRetriever,
)
from llama_index.core.schema import NodeWithScore, QueryBundle
from llama_index.embeddings.cohere import CohereEmbedding
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.vector_stores.chroma import ChromaVectorStore
from utils import init_mongo_db

load_dotenv()

logger = logging.getLogger(__name__)
logging.basicConfig(level=logging.INFO)
logging.getLogger("httpx").setLevel(logging.WARNING)
logfire.configure()


if not os.path.exists("data/chroma-db-all_sources"):
    # Download the vector database from the Hugging Face Hub if it doesn't exist locally
    # https://huggingface.co/datasets/towardsai-buster/ai-tutor-vector-db/tree/main
    logfire.warn(
        f"Vector database does not exist at 'data/chroma-db-all_sources', downloading from Hugging Face Hub"
    )
    from huggingface_hub import snapshot_download

    snapshot_download(
        repo_id="towardsai-buster/ai-tutor-vector-db",
        local_dir="data",
        repo_type="dataset",
    )
    logfire.info(f"Downloaded vector database to 'data/chroma-db-all_sources'")


def create_docs(input_file: str) -> list[Document]:
    with open(input_file, "r") as f:
        documents = []
        for line in f:
            data = json.loads(line)
            documents.append(
                Document(
                    doc_id=data["doc_id"],
                    text=data["content"],
                    metadata={  # type: ignore
                        "url": data["url"],
                        "title": data["name"],
                        "tokens": data["tokens"],
                        "retrieve_doc": data["retrieve_doc"],
                        "source": data["source"],
                    },
                    excluded_llm_metadata_keys=[
                        "title",
                        "tokens",
                        "retrieve_doc",
                        "source",
                    ],
                    excluded_embed_metadata_keys=[
                        "url",
                        "tokens",
                        "retrieve_doc",
                        "source",
                    ],
                )
            )
    return documents


def setup_database(db_collection, dict_file_name):
    db = chromadb.PersistentClient(path=f"data/{db_collection}")
    chroma_collection = db.get_or_create_collection(db_collection)
    vector_store = ChromaVectorStore(chroma_collection=chroma_collection)
    embed_model = CohereEmbedding(
        api_key=os.environ["COHERE_API_KEY"],
        model_name="embed-english-v3.0",
        input_type="search_query",
    )

    index = VectorStoreIndex.from_vector_store(
        vector_store=vector_store,
        transformations=[SentenceSplitter(chunk_size=800, chunk_overlap=0)],
        show_progress=True,
        use_async=True,
    )
    vector_retriever = VectorIndexRetriever(
        index=index,
        similarity_top_k=15,
        embed_model=embed_model,
        use_async=True,
    )
    with open(f"data/{db_collection}/{dict_file_name}", "rb") as f:
        document_dict = pickle.load(f)

    with open("data/keyword_retriever_sync.pkl", "rb") as f:
        keyword_retriever: KeywordTableSimpleRetriever = pickle.load(f)

    # # Creating the keyword index and retriever
    # logfire.info("Creating nodes from documents")
    # documents = create_docs("data/all_sources_data.jsonl")
    # pipeline = IngestionPipeline(
    #     transformations=[SentenceSplitter(chunk_size=800, chunk_overlap=0)]
    # )
    # all_nodes = pipeline.run(documents=documents, show_progress=True)
    # # with open("data/all_nodes.pkl", "wb") as f:
    # #     pickle.dump(all_nodes, f)

    # # all_nodes = pickle.load(open("data/all_nodes.pkl", "rb"))
    # logfire.info(f"Number of nodes: {len(all_nodes)}")

    # keyword_index = SimpleKeywordTableIndex(
    #     nodes=all_nodes, max_keywords_per_chunk=10, show_progress=True, use_async=False
    # )
    # # with open("data/keyword_index.pkl", "wb") as f:
    # # pickle.dump(keyword_index, f)

    # # keyword_index = pickle.load(open("data/keyword_index.pkl", "rb"))

    # logfire.info("Creating keyword retriever")
    # keyword_retriever = KeywordTableSimpleRetriever(index=keyword_index)

    # with open("data/keyword_retriever_sync.pkl", "wb") as f:
    #     pickle.dump(keyword_retriever, f)

    return CustomRetriever(vector_retriever, document_dict, keyword_retriever, "OR")


# Setup retrievers
# custom_retriever_transformers: CustomRetriever = setup_database(
#     "chroma-db-transformers",
#     "document_dict_transformers.pkl",
# )
# custom_retriever_peft: CustomRetriever = setup_database(
#     "chroma-db-peft", "document_dict_peft.pkl"
# )
# custom_retriever_trl: CustomRetriever = setup_database(
#     "chroma-db-trl", "document_dict_trl.pkl"
# )
# custom_retriever_llama_index: CustomRetriever = setup_database(
#     "chroma-db-llama_index",
#     "document_dict_llama_index.pkl",
# )
# custom_retriever_openai_cookbooks: CustomRetriever = setup_database(
#     "chroma-db-openai_cookbooks",
#     "document_dict_openai_cookbooks.pkl",
# )
# custom_retriever_langchain: CustomRetriever = setup_database(
#     "chroma-db-langchain",
#     "document_dict_langchain.pkl",
# )

custom_retriever_all_sources: CustomRetriever = setup_database(
    "chroma-db-all_sources",
    "document_dict_all_sources.pkl",
)

# Constants
CONCURRENCY_COUNT = int(os.getenv("CONCURRENCY_COUNT", 64))
MONGODB_URI = os.getenv("MONGODB_URI")

AVAILABLE_SOURCES_UI = [
    "Transformers Docs",
    "PEFT Docs",
    "TRL Docs",
    "LlamaIndex Docs",
    "LangChain Docs",
    "OpenAI Cookbooks",
    "Towards AI Blog",
    # "All Sources",
    # "RAG Course",
]

AVAILABLE_SOURCES = [
    "transformers",
    "peft",
    "trl",
    "llama_index",
    "langchain",
    "openai_cookbooks",
    "tai_blog",
    # "all_sources",
    # "rag_course",
]

mongo_db = (
    init_mongo_db(uri=MONGODB_URI, db_name="towardsai-buster")
    if MONGODB_URI
    else logfire.warn("No mongodb uri found, you will not be able to save data.")
)

__all__ = [
    # "custom_retriever_transformers",
    # "custom_retriever_peft",
    # "custom_retriever_trl",
    # "custom_retriever_llama_index",
    # "custom_retriever_openai_cookbooks",
    # "custom_retriever_langchain",
    "custom_retriever_all_sources",
    "mongo_db",
    "CONCURRENCY_COUNT",
    "AVAILABLE_SOURCES_UI",
    "AVAILABLE_SOURCES",
]