Spaces:
Sleeping
Sleeping
Update create_and_upload_index.py
Browse files- create_and_upload_index.py +118 -118
create_and_upload_index.py
CHANGED
|
@@ -1,119 +1,119 @@
|
|
| 1 |
-
|
| 2 |
-
import os
|
| 3 |
-
import json
|
| 4 |
-
from tqdm import tqdm
|
| 5 |
-
from azure.core.credentials import AzureKeyCredential
|
| 6 |
-
from azure.search.documents.indexes import SearchIndexClient
|
| 7 |
-
from azure.search.documents.indexes.models import (
|
| 8 |
-
SearchIndex,
|
| 9 |
-
SearchField,
|
| 10 |
-
SearchFieldDataType,
|
| 11 |
-
SimpleField,
|
| 12 |
-
SearchableField,
|
| 13 |
-
VectorSearch,
|
| 14 |
-
HnswAlgorithmConfiguration,
|
| 15 |
-
VectorSearchProfile
|
| 16 |
-
)
|
| 17 |
-
from azure.search.documents import SearchClient
|
| 18 |
-
from openai import AzureOpenAI
|
| 19 |
-
from dotenv import load_dotenv
|
| 20 |
-
|
| 21 |
-
load_dotenv()
|
| 22 |
-
|
| 23 |
-
# ---------------- CONFIG ----------------
|
| 24 |
-
AZURE_SEARCH_ENDPOINT = os.getenv("AZURE_SEARCH_ENDPOINT") # e.g., https://your-search.search.windows.net
|
| 25 |
-
AZURE_SEARCH_KEY = os.getenv("AZURE_SEARCH_KEY")
|
| 26 |
-
AZURE_SEARCH_INDEX_NAME = "it-ticket-solutions-index"
|
| 27 |
-
|
| 28 |
-
HF_TOKEN = os.getenv("HF_TOKEN")
|
| 29 |
-
API_BASE_URL = os.getenv("API_BASE_URL")
|
| 30 |
-
MODEL_NAME = os.getenv("MODEL_NAME") # e.g.,
|
| 31 |
-
AZURE_OPENAI_API_VERSION = "2024-02-15-preview"
|
| 32 |
-
|
| 33 |
-
DATA_FILE = "data/knowledge_base.json"
|
| 34 |
-
VECTOR_DIMENSIONS = 1536 # Based on OpenAI embeddings
|
| 35 |
-
# ----------------------------------------
|
| 36 |
-
|
| 37 |
-
# ---------- Clients ----------
|
| 38 |
-
credential = AzureKeyCredential(AZURE_SEARCH_KEY)
|
| 39 |
-
index_client = SearchIndexClient(endpoint=AZURE_SEARCH_ENDPOINT, credential=credential)
|
| 40 |
-
search_client = SearchClient(endpoint=AZURE_SEARCH_ENDPOINT, index_name=AZURE_SEARCH_INDEX_NAME, credential=credential)
|
| 41 |
-
openai_client = AzureOpenAI(api_key=HF_TOKEN, api_version=AZURE_OPENAI_API_VERSION, azure_endpoint=API_BASE_URL)
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
# ---------- Index Creation ----------
|
| 45 |
-
def create_index():
|
| 46 |
-
print("📦 Creating index...")
|
| 47 |
-
|
| 48 |
-
fields = [
|
| 49 |
-
SimpleField(name="id", type=SearchFieldDataType.String, key=True, filterable=True, sortable=True, facetable=True),
|
| 50 |
-
SearchableField(name="category", type=SearchFieldDataType.String, filterable=True, sortable=True, facetable=True),
|
| 51 |
-
SearchableField(name="problem", type=SearchFieldDataType.String, filterable=True, sortable=True, facetable=True),
|
| 52 |
-
SearchableField(name="solution", type=SearchFieldDataType.String, filterable=True, sortable=True, facetable=True),
|
| 53 |
-
SearchField(
|
| 54 |
-
name="embedding",
|
| 55 |
-
type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
|
| 56 |
-
searchable=True,
|
| 57 |
-
vector_search_dimensions=VECTOR_DIMENSIONS,
|
| 58 |
-
vector_search_profile_name="default"
|
| 59 |
-
)
|
| 60 |
-
]
|
| 61 |
-
|
| 62 |
-
vector_search = VectorSearch(
|
| 63 |
-
algorithms=[HnswAlgorithmConfiguration(name="default")],
|
| 64 |
-
profiles=[VectorSearchProfile(name="default", algorithm_configuration_name="default")]
|
| 65 |
-
)
|
| 66 |
-
|
| 67 |
-
index = SearchIndex(name=AZURE_SEARCH_INDEX_NAME, fields=fields, vector_search=vector_search)
|
| 68 |
-
|
| 69 |
-
try:
|
| 70 |
-
# Try to get the index; if it exists, we skip creation
|
| 71 |
-
index_client.get_index(name=AZURE_SEARCH_INDEX_NAME)
|
| 72 |
-
print(f"ℹ️ Index '{AZURE_SEARCH_INDEX_NAME}' already exists. Skipping creation.")
|
| 73 |
-
except Exception:
|
| 74 |
-
# If not found, create the index
|
| 75 |
-
index_client.create_index(index)
|
| 76 |
-
print(f"✅ Index '{AZURE_SEARCH_INDEX_NAME}' created successfully.")
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
# ---------- Embedding ----------
|
| 80 |
-
def embed_text(text: str):
|
| 81 |
-
response = openai_client.embeddings.create(
|
| 82 |
-
input=[text],
|
| 83 |
-
model=MODEL_NAME
|
| 84 |
-
)
|
| 85 |
-
return response.data[0].embedding
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
# ---------- Upload ----------
|
| 89 |
-
def load_data():
|
| 90 |
-
with open(DATA_FILE, "r", encoding="utf-8") as f:
|
| 91 |
-
return json.load(f)
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
def upload_documents(docs):
|
| 95 |
-
print(f"\n📤 Uploading {len(docs)} documents...")
|
| 96 |
-
batch_size = 10
|
| 97 |
-
for i in tqdm(range(0, len(docs), batch_size)):
|
| 98 |
-
chunk = docs[i:i+batch_size]
|
| 99 |
-
search_client.upload_documents(documents=chunk)
|
| 100 |
-
print("✅ Upload completed.")
|
| 101 |
-
|
| 102 |
-
# ---------- Main ----------
|
| 103 |
-
def main():
|
| 104 |
-
create_index()
|
| 105 |
-
|
| 106 |
-
raw_docs = load_data()
|
| 107 |
-
enriched_docs = []
|
| 108 |
-
|
| 109 |
-
print("🔍 Generating embeddings...")
|
| 110 |
-
for doc in tqdm(raw_docs):
|
| 111 |
-
embedding = embed_text(doc["problem"])
|
| 112 |
-
doc["embedding"] = embedding
|
| 113 |
-
enriched_docs.append(doc)
|
| 114 |
-
|
| 115 |
-
upload_documents(enriched_docs)
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
if __name__ == "__main__":
|
| 119 |
main()
|
|
|
|
| 1 |
+
|
| 2 |
+
import os
|
| 3 |
+
import json
|
| 4 |
+
from tqdm import tqdm
|
| 5 |
+
from azure.core.credentials import AzureKeyCredential
|
| 6 |
+
from azure.search.documents.indexes import SearchIndexClient
|
| 7 |
+
from azure.search.documents.indexes.models import (
|
| 8 |
+
SearchIndex,
|
| 9 |
+
SearchField,
|
| 10 |
+
SearchFieldDataType,
|
| 11 |
+
SimpleField,
|
| 12 |
+
SearchableField,
|
| 13 |
+
VectorSearch,
|
| 14 |
+
HnswAlgorithmConfiguration,
|
| 15 |
+
VectorSearchProfile
|
| 16 |
+
)
|
| 17 |
+
from azure.search.documents import SearchClient
|
| 18 |
+
from openai import AzureOpenAI
|
| 19 |
+
from dotenv import load_dotenv
|
| 20 |
+
|
| 21 |
+
load_dotenv()
|
| 22 |
+
|
| 23 |
+
# ---------------- CONFIG ----------------
|
| 24 |
+
AZURE_SEARCH_ENDPOINT = os.getenv("AZURE_SEARCH_ENDPOINT") # e.g., https://your-search.search.windows.net
|
| 25 |
+
AZURE_SEARCH_KEY = os.getenv("AZURE_SEARCH_KEY")
|
| 26 |
+
AZURE_SEARCH_INDEX_NAME = "it-ticket-solutions-index"
|
| 27 |
+
|
| 28 |
+
HF_TOKEN = os.getenv("HF_TOKEN")
|
| 29 |
+
API_BASE_URL = os.getenv("API_BASE_URL")
|
| 30 |
+
MODEL_NAME = os.getenv("MODEL_NAME") # e.g.,
|
| 31 |
+
AZURE_OPENAI_API_VERSION = "2024-02-15-preview"
|
| 32 |
+
|
| 33 |
+
DATA_FILE = "data/knowledge_base.json"
|
| 34 |
+
VECTOR_DIMENSIONS = 1536 # Based on OpenAI embeddings
|
| 35 |
+
# ----------------------------------------
|
| 36 |
+
|
| 37 |
+
# ---------- Clients ----------
|
| 38 |
+
credential = AzureKeyCredential(AZURE_SEARCH_KEY)
|
| 39 |
+
index_client = SearchIndexClient(endpoint=AZURE_SEARCH_ENDPOINT, credential=credential)
|
| 40 |
+
search_client = SearchClient(endpoint=AZURE_SEARCH_ENDPOINT, index_name=AZURE_SEARCH_INDEX_NAME, credential=credential)
|
| 41 |
+
openai_client = AzureOpenAI(api_key=HF_TOKEN, api_version=AZURE_OPENAI_API_VERSION, azure_endpoint=API_BASE_URL)
|
| 42 |
+
|
| 43 |
+
|
| 44 |
+
# ---------- Index Creation ----------
|
| 45 |
+
def create_index():
|
| 46 |
+
print("📦 Creating index...")
|
| 47 |
+
|
| 48 |
+
fields = [
|
| 49 |
+
SimpleField(name="id", type=SearchFieldDataType.String, key=True, filterable=True, sortable=True, facetable=True),
|
| 50 |
+
SearchableField(name="category", type=SearchFieldDataType.String, filterable=True, sortable=True, facetable=True),
|
| 51 |
+
SearchableField(name="problem", type=SearchFieldDataType.String, filterable=True, sortable=True, facetable=True),
|
| 52 |
+
SearchableField(name="solution", type=SearchFieldDataType.String, filterable=True, sortable=True, facetable=True),
|
| 53 |
+
SearchField(
|
| 54 |
+
name="embedding",
|
| 55 |
+
type=SearchFieldDataType.Collection(SearchFieldDataType.Single),
|
| 56 |
+
searchable=True,
|
| 57 |
+
vector_search_dimensions=VECTOR_DIMENSIONS,
|
| 58 |
+
vector_search_profile_name="default"
|
| 59 |
+
)
|
| 60 |
+
]
|
| 61 |
+
|
| 62 |
+
vector_search = VectorSearch(
|
| 63 |
+
algorithms=[HnswAlgorithmConfiguration(name="default")],
|
| 64 |
+
profiles=[VectorSearchProfile(name="default", algorithm_configuration_name="default")]
|
| 65 |
+
)
|
| 66 |
+
|
| 67 |
+
index = SearchIndex(name=AZURE_SEARCH_INDEX_NAME, fields=fields, vector_search=vector_search)
|
| 68 |
+
|
| 69 |
+
try:
|
| 70 |
+
# Try to get the index; if it exists, we skip creation
|
| 71 |
+
index_client.get_index(name=AZURE_SEARCH_INDEX_NAME)
|
| 72 |
+
print(f"ℹ️ Index '{AZURE_SEARCH_INDEX_NAME}' already exists. Skipping creation.")
|
| 73 |
+
except Exception:
|
| 74 |
+
# If not found, create the index
|
| 75 |
+
index_client.create_index(index)
|
| 76 |
+
print(f"✅ Index '{AZURE_SEARCH_INDEX_NAME}' created successfully.")
|
| 77 |
+
|
| 78 |
+
|
| 79 |
+
# ---------- Embedding ----------
|
| 80 |
+
def embed_text(text: str):
|
| 81 |
+
response = openai_client.embeddings.create(
|
| 82 |
+
input=[text],
|
| 83 |
+
model=MODEL_NAME
|
| 84 |
+
)
|
| 85 |
+
return response.data[0].embedding
|
| 86 |
+
|
| 87 |
+
|
| 88 |
+
# ---------- Upload ----------
|
| 89 |
+
def load_data():
|
| 90 |
+
with open(DATA_FILE, "r", encoding="utf-8") as f:
|
| 91 |
+
return json.load(f)
|
| 92 |
+
|
| 93 |
+
|
| 94 |
+
def upload_documents(docs):
|
| 95 |
+
print(f"\n📤 Uploading {len(docs)} documents...")
|
| 96 |
+
batch_size = 10
|
| 97 |
+
for i in tqdm(range(0, len(docs), batch_size)):
|
| 98 |
+
chunk = docs[i:i+batch_size]
|
| 99 |
+
search_client.upload_documents(documents=chunk)
|
| 100 |
+
print("✅ Upload completed.")
|
| 101 |
+
|
| 102 |
+
# ---------- Main ----------
|
| 103 |
+
def main():
|
| 104 |
+
create_index()
|
| 105 |
+
|
| 106 |
+
raw_docs = load_data()
|
| 107 |
+
enriched_docs = []
|
| 108 |
+
|
| 109 |
+
print("🔍 Generating embeddings...")
|
| 110 |
+
for doc in tqdm(raw_docs):
|
| 111 |
+
embedding = embed_text(doc["problem"])
|
| 112 |
+
doc["embedding"] = embedding
|
| 113 |
+
enriched_docs.append(doc)
|
| 114 |
+
|
| 115 |
+
upload_documents(enriched_docs)
|
| 116 |
+
|
| 117 |
+
|
| 118 |
+
if __name__ == "__main__":
|
| 119 |
main()
|