Spaces:
Sleeping
Sleeping
Upload app.py
Browse files
app.py
CHANGED
@@ -12,8 +12,8 @@ from tqdm import tqdm
|
|
12 |
from datasets import load_dataset
|
13 |
import pandas as pd
|
14 |
from sentence_transformers import SentenceTransformer
|
15 |
-
|
16 |
-
import chromadb.config
|
17 |
|
18 |
# --- Page Config (MUST BE FIRST Streamlit call) ---
|
19 |
st.set_page_config(layout="wide")
|
@@ -25,7 +25,7 @@ LOCAL_EMBEDDING_MODEL = 'BAAI/bge-m3' # Local model for QUERY embedding
|
|
25 |
HF_GENERATION_MODEL = "google/gemma-3-27b-it" # HF model for generation
|
26 |
HF_DATASET_ID = "Zwounds/Libguides_Embeddings" # Your HF Dataset ID
|
27 |
PARQUET_FILENAME = "libguides_embeddings.parquet" # Filename within the dataset
|
28 |
-
ADD_BATCH_SIZE = 500 # Batch size for adding to
|
29 |
TOP_K = 10
|
30 |
INITIAL_N_RESULTS = 50
|
31 |
MAX_NEW_TOKENS = 512
|
@@ -129,12 +129,18 @@ generation_client = initialize_hf_client()
|
|
129 |
embedding_model = load_local_embedding_model()
|
130 |
# ---
|
131 |
|
132 |
-
# --- Setup ChromaDB Collection (using Session State) ---
|
133 |
-
# This function now attempts to load or create the collection and stores it in session state
|
134 |
def setup_chroma_collection():
|
|
|
135 |
if 'chroma_collection' in st.session_state and st.session_state.chroma_collection is not None:
|
136 |
-
|
137 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
138 |
|
139 |
# Proceed with setup only if essential components are loaded
|
140 |
if not embedding_model or not generation_client:
|
@@ -147,17 +153,23 @@ def setup_chroma_collection():
|
|
147 |
st.error("Failed to load embedding data. Cannot initialize vector database.")
|
148 |
return None
|
149 |
|
|
|
|
|
|
|
|
|
|
|
|
|
150 |
try:
|
151 |
-
logging.info("Initializing
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
-
allow_reset=True # Optional: Allows resetting
|
157 |
-
)
|
158 |
)
|
|
|
|
|
159 |
|
160 |
-
# Check if collection exists and delete if it does
|
161 |
try:
|
162 |
existing_collections = [col.name for col in chroma_client.list_collections()]
|
163 |
if COLLECTION_NAME in existing_collections:
|
@@ -166,7 +178,6 @@ def setup_chroma_collection():
|
|
166 |
except Exception as delete_e:
|
167 |
logging.warning(f"Could not check/delete existing collection (might be okay): {delete_e}")
|
168 |
|
169 |
-
|
170 |
logging.info(f"Creating collection: {COLLECTION_NAME}")
|
171 |
collection_instance = chroma_client.create_collection(
|
172 |
name=COLLECTION_NAME,
|
@@ -234,7 +245,6 @@ def setup_chroma_collection():
|
|
234 |
return None
|
235 |
|
236 |
# --- Initialize collection ---
|
237 |
-
# Call the setup function which populates session state if needed
|
238 |
collection = setup_chroma_collection()
|
239 |
# ---
|
240 |
|
|
|
12 |
from datasets import load_dataset
|
13 |
import pandas as pd
|
14 |
from sentence_transformers import SentenceTransformer
|
15 |
+
import tempfile # Added for temporary directory
|
16 |
+
import chromadb.config # Added for Settings
|
17 |
|
18 |
# --- Page Config (MUST BE FIRST Streamlit call) ---
|
19 |
st.set_page_config(layout="wide")
|
|
|
25 |
HF_GENERATION_MODEL = "google/gemma-3-27b-it" # HF model for generation
|
26 |
HF_DATASET_ID = "Zwounds/Libguides_Embeddings" # Your HF Dataset ID
|
27 |
PARQUET_FILENAME = "libguides_embeddings.parquet" # Filename within the dataset
|
28 |
+
ADD_BATCH_SIZE = 500 # Batch size for adding to Chroma
|
29 |
TOP_K = 10
|
30 |
INITIAL_N_RESULTS = 50
|
31 |
MAX_NEW_TOKENS = 512
|
|
|
129 |
embedding_model = load_local_embedding_model()
|
130 |
# ---
|
131 |
|
132 |
+
# --- Setup ChromaDB Collection (using Session State and Temp Dir) ---
|
|
|
133 |
def setup_chroma_collection():
|
134 |
+
"""Loads data from HF, sets up ChromaDB in a temp dir, populates it, and returns the collection."""
|
135 |
if 'chroma_collection' in st.session_state and st.session_state.chroma_collection is not None:
|
136 |
+
# Basic check: see if collection is queryable
|
137 |
+
try:
|
138 |
+
st.session_state.chroma_collection.peek(1) # Try a lightweight operation
|
139 |
+
logging.info("Using existing Chroma collection from session state.")
|
140 |
+
return st.session_state.chroma_collection
|
141 |
+
except Exception as e:
|
142 |
+
logging.warning(f"Error accessing existing collection in session state ({e}), re-initializing.")
|
143 |
+
st.session_state.chroma_collection = None # Force re-init
|
144 |
|
145 |
# Proceed with setup only if essential components are loaded
|
146 |
if not embedding_model or not generation_client:
|
|
|
153 |
st.error("Failed to load embedding data. Cannot initialize vector database.")
|
154 |
return None
|
155 |
|
156 |
+
# Create a temporary directory for this session
|
157 |
+
# Note: This directory might be cleaned up automatically depending on the OS/environment
|
158 |
+
# In HF Spaces ephemeral storage, it will likely be wiped on restart anyway.
|
159 |
+
temp_dir = tempfile.mkdtemp()
|
160 |
+
logging.info(f"Created temporary directory for ChromaDB: {temp_dir}")
|
161 |
+
|
162 |
try:
|
163 |
+
logging.info("Initializing ChromaDB client with temporary storage...")
|
164 |
+
settings = chromadb.config.Settings(
|
165 |
+
persist_directory=temp_dir,
|
166 |
+
anonymized_telemetry=False,
|
167 |
+
is_persistent=True # Explicitly set for PersistentClient behavior in temp dir
|
|
|
|
|
168 |
)
|
169 |
+
# Use the standard Client, but point it to the temp directory
|
170 |
+
chroma_client = chromadb.Client(settings=settings)
|
171 |
|
172 |
+
# Check if collection exists and delete if it does
|
173 |
try:
|
174 |
existing_collections = [col.name for col in chroma_client.list_collections()]
|
175 |
if COLLECTION_NAME in existing_collections:
|
|
|
178 |
except Exception as delete_e:
|
179 |
logging.warning(f"Could not check/delete existing collection (might be okay): {delete_e}")
|
180 |
|
|
|
181 |
logging.info(f"Creating collection: {COLLECTION_NAME}")
|
182 |
collection_instance = chroma_client.create_collection(
|
183 |
name=COLLECTION_NAME,
|
|
|
245 |
return None
|
246 |
|
247 |
# --- Initialize collection ---
|
|
|
248 |
collection = setup_chroma_collection()
|
249 |
# ---
|
250 |
|