Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -13,12 +13,19 @@ from langchain_community.vectorstores import Qdrant
|
|
13 |
from langchain_core.runnables import RunnablePassthrough, RunnableParallel
|
14 |
import chainlit as cl
|
15 |
from pathlib import Path
|
16 |
-
from sentence_transformers import SentenceTransformer
|
17 |
|
18 |
load_dotenv()
|
19 |
|
20 |
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
|
21 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
@cl.on_chat_start
|
23 |
async def on_chat_start():
|
24 |
model = ChatOpenAI(streaming=True)
|
@@ -53,10 +60,10 @@ async def on_chat_start():
|
|
53 |
|
54 |
sentence_combined_documents = sentence_framework + sentence_blueprint
|
55 |
|
56 |
-
# Initialize the
|
57 |
-
embedding_model =
|
58 |
|
59 |
-
# Create the Qdrant vector store using the
|
60 |
sentence_vectorstore = Qdrant.from_documents(
|
61 |
documents=sentence_combined_documents,
|
62 |
embedding=embedding_model, # Ensure this is an instance
|
@@ -65,7 +72,11 @@ async def on_chat_start():
|
|
65 |
)
|
66 |
|
67 |
sentence_retriever = sentence_vectorstore.as_retriever()
|
68 |
-
|
|
|
|
|
|
|
|
|
69 |
# Set the retriever and prompt into session for reuse
|
70 |
cl.user_session.set("runnable", model)
|
71 |
cl.user_session.set("retriever", sentence_retriever)
|
@@ -83,6 +94,11 @@ async def on_message(message: cl.Message):
|
|
83 |
print(f"Received message: {message.content}")
|
84 |
|
85 |
# Retrieve relevant context from documents based on the user's message
|
|
|
|
|
|
|
|
|
|
|
86 |
relevant_docs = retriever.get_relevant_documents(message.content)
|
87 |
print(f"Retrieved {len(relevant_docs)} documents.")
|
88 |
|
|
|
13 |
from langchain_core.runnables import RunnablePassthrough, RunnableParallel
|
14 |
import chainlit as cl
|
15 |
from pathlib import Path
|
16 |
+
from sentence_transformers import SentenceTransformer
|
17 |
|
18 |
load_dotenv()
|
19 |
|
20 |
os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
|
21 |
|
22 |
+
class SentenceTransformerEmbedding:
|
23 |
+
def __init__(self, model_name):
|
24 |
+
self.model = SentenceTransformer(model_name)
|
25 |
+
|
26 |
+
def embed_documents(self, texts):
|
27 |
+
return self.model.encode(texts, convert_to_tensor=True).tolist() # Convert to list for compatibility
|
28 |
+
|
29 |
@cl.on_chat_start
|
30 |
async def on_chat_start():
|
31 |
model = ChatOpenAI(streaming=True)
|
|
|
60 |
|
61 |
sentence_combined_documents = sentence_framework + sentence_blueprint
|
62 |
|
63 |
+
# Initialize the custom embedding class
|
64 |
+
embedding_model = SentenceTransformerEmbedding('Cheselle/finetuned-arctic-sentence')
|
65 |
|
66 |
+
# Create the Qdrant vector store using the custom embedding model
|
67 |
sentence_vectorstore = Qdrant.from_documents(
|
68 |
documents=sentence_combined_documents,
|
69 |
embedding=embedding_model, # Ensure this is an instance
|
|
|
72 |
)
|
73 |
|
74 |
sentence_retriever = sentence_vectorstore.as_retriever()
|
75 |
+
|
76 |
+
# Check if retriever is initialized correctly
|
77 |
+
if sentence_retriever is None:
|
78 |
+
raise ValueError("Retriever is not initialized correctly.")
|
79 |
+
|
80 |
# Set the retriever and prompt into session for reuse
|
81 |
cl.user_session.set("runnable", model)
|
82 |
cl.user_session.set("retriever", sentence_retriever)
|
|
|
94 |
print(f"Received message: {message.content}")
|
95 |
|
96 |
# Retrieve relevant context from documents based on the user's message
|
97 |
+
if retriever is None:
|
98 |
+
print("Retriever is not available.")
|
99 |
+
await cl.Message(content="Sorry, the retriever is not initialized.").send()
|
100 |
+
return
|
101 |
+
|
102 |
relevant_docs = retriever.get_relevant_documents(message.content)
|
103 |
print(f"Retrieved {len(relevant_docs)} documents.")
|
104 |
|