Spaces:
Sleeping
Sleeping
Upload 2 files
Browse files- app.py +36 -64
- requirements.txt +0 -1
app.py
CHANGED
@@ -7,21 +7,18 @@ from aimakerspace.openai_utils.prompts import (
|
|
7 |
SystemRolePrompt,
|
8 |
AssistantRolePrompt,
|
9 |
)
|
10 |
-
|
11 |
-
|
12 |
-
from
|
13 |
-
# from aimakerspace.openai_utils.chatmodel import ChatOpenAI
|
14 |
import chainlit as cl
|
15 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
16 |
-
# from
|
17 |
-
# from langchain_openai.embeddings import OpenAIEmbeddings
|
18 |
-
from sentence_transformers import SentenceTransformer
|
19 |
from langchain_huggingface import HuggingFaceEmbeddings
|
20 |
-
from langchain_community.vectorstores import FAISS
|
21 |
-
from langchain_openai.embeddings import OpenAIEmbeddings
|
22 |
-
from langchain_core.documents import Document
|
23 |
from dotenv import load_dotenv
|
24 |
-
|
25 |
|
26 |
load_dotenv()
|
27 |
|
@@ -37,27 +34,27 @@ Question:
|
|
37 |
"""
|
38 |
user_role_prompt = UserRolePrompt(user_prompt_template)
|
39 |
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
|
45 |
-
|
46 |
-
|
47 |
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
|
52 |
-
|
53 |
|
54 |
-
|
55 |
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
|
60 |
-
|
61 |
|
62 |
text_splitter = RecursiveCharacterTextSplitter()
|
63 |
|
@@ -80,8 +77,8 @@ def process_text_file(file: AskFileResponse):
|
|
80 |
documents = pdf_loader.load()
|
81 |
else:
|
82 |
raise ValueError("Provide a .txt or .pdf file")
|
83 |
-
|
84 |
-
|
85 |
return text_splitter.split_documents(documents)
|
86 |
|
87 |
|
@@ -100,7 +97,8 @@ async def on_chat_start():
|
|
100 |
max_files=10
|
101 |
).send()
|
102 |
|
103 |
-
|
|
|
104 |
for file in files:
|
105 |
|
106 |
msg = cl.Message(
|
@@ -110,50 +108,26 @@ async def on_chat_start():
|
|
110 |
|
111 |
# load the file
|
112 |
texts = process_text_file(file)
|
113 |
-
|
114 |
print(f"Processing {len(texts)} text chunks")
|
115 |
|
116 |
# Create a dict vector store
|
117 |
-
# vector_db = VectorDatabase()
|
118 |
-
# vector_db = await vector_db.abuild_from_list(texts)
|
119 |
|
120 |
-
|
121 |
|
122 |
-
|
123 |
-
# retrieval_augmented_qa_pipeline = RetrievalAugmentedQAPipeline(
|
124 |
-
# vector_db_retriever=vector_db,
|
125 |
-
# llm=chat_openai
|
126 |
-
# )
|
127 |
-
|
128 |
-
# model = SentenceTransformer("acpotts/finetuned_arctic")
|
129 |
-
|
130 |
-
finetune_embeddings = HuggingFaceEmbeddings(model_name='acpotts/finetuned_arctic')
|
131 |
|
132 |
-
|
133 |
-
|
134 |
-
|
135 |
-
|
136 |
-
from langchain_core.output_parsers import StrOutputParser
|
137 |
-
from langchain_core.runnables import RunnablePassthrough, RunnableParallel
|
138 |
-
|
139 |
-
rag_llm = ChatOpenAI(
|
140 |
-
model="gpt-4o-mini",
|
141 |
-
temperature=0
|
142 |
-
)
|
143 |
-
|
144 |
-
|
145 |
-
finetune_rag_chain = (
|
146 |
-
{"context": itemgetter("question") | finetune_retriever, "question": itemgetter("question")}
|
147 |
-
| RunnablePassthrough.assign(context=itemgetter("context"))
|
148 |
-
| {"response": system_template | rag_llm | StrOutputParser(), "context": itemgetter("context")}
|
149 |
)
|
150 |
-
|
151 |
|
152 |
# Let the user know that the system is ready
|
153 |
msg.content = f"Processing `{file.name}` done. You can now ask questions!"
|
154 |
await msg.update()
|
155 |
|
156 |
-
cl.user_session.set("chain",
|
157 |
|
158 |
|
159 |
@cl.on_message
|
@@ -161,9 +135,7 @@ async def main(message):
|
|
161 |
chain = cl.user_session.get("chain")
|
162 |
|
163 |
msg = cl.Message(content="")
|
164 |
-
|
165 |
-
# result = await chain.arun_pipeline(message.content)
|
166 |
-
result = await chain.arun_pipeline({'question': message.content})
|
167 |
|
168 |
async for stream_resp in result["response"]:
|
169 |
await msg.stream_token(stream_resp)
|
|
|
7 |
SystemRolePrompt,
|
8 |
AssistantRolePrompt,
|
9 |
)
|
10 |
+
from aimakerspace.openai_utils.embedding import EmbeddingModel
|
11 |
+
from aimakerspace.vectordatabase import VectorDatabase
|
12 |
+
from aimakerspace.openai_utils.chatmodel import ChatOpenAI
|
|
|
13 |
import chainlit as cl
|
14 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
15 |
+
# from sentence_transformers import SentenceTransformer
|
|
|
|
|
16 |
from langchain_huggingface import HuggingFaceEmbeddings
|
17 |
+
# from langchain_community.vectorstores import FAISS
|
18 |
+
# from langchain_openai.embeddings import OpenAIEmbeddings
|
19 |
+
# from langchain_core.documents import Document
|
20 |
from dotenv import load_dotenv
|
21 |
+
|
22 |
|
23 |
load_dotenv()
|
24 |
|
|
|
34 |
"""
|
35 |
user_role_prompt = UserRolePrompt(user_prompt_template)
|
36 |
|
37 |
+
class RetrievalAugmentedQAPipeline:
|
38 |
+
def __init__(self, llm: ChatOpenAI(), vector_db_retriever: VectorDatabase) -> None:
|
39 |
+
self.llm = llm
|
40 |
+
self.vector_db_retriever = vector_db_retriever
|
41 |
|
42 |
+
async def arun_pipeline(self, user_query: str):
|
43 |
+
context_list = self.vector_db_retriever.search_by_text(user_query, k=4)
|
44 |
|
45 |
+
context_prompt = ""
|
46 |
+
for context in context_list:
|
47 |
+
context_prompt += context[0] + "\n"
|
48 |
|
49 |
+
formatted_system_prompt = system_role_prompt.create_message()
|
50 |
|
51 |
+
formatted_user_prompt = user_role_prompt.create_message(question=user_query, context=context_prompt)
|
52 |
|
53 |
+
async def generate_response():
|
54 |
+
async for chunk in self.llm.astream([formatted_system_prompt, formatted_user_prompt]):
|
55 |
+
yield chunk
|
56 |
|
57 |
+
return {"response": generate_response(), "context": context_list}
|
58 |
|
59 |
text_splitter = RecursiveCharacterTextSplitter()
|
60 |
|
|
|
77 |
documents = pdf_loader.load()
|
78 |
else:
|
79 |
raise ValueError("Provide a .txt or .pdf file")
|
80 |
+
texts = [x.page_content for x in text_splitter.transform_documents(documents)]
|
81 |
+
|
82 |
return text_splitter.split_documents(documents)
|
83 |
|
84 |
|
|
|
97 |
max_files=10
|
98 |
).send()
|
99 |
|
100 |
+
embedding_model = HuggingFaceEmbeddings(model_name='acpotts/finetuned_arctic')
|
101 |
+
vector_db = VectorDatabase(embedding_model=embedding_model)
|
102 |
for file in files:
|
103 |
|
104 |
msg = cl.Message(
|
|
|
108 |
|
109 |
# load the file
|
110 |
texts = process_text_file(file)
|
111 |
+
|
112 |
print(f"Processing {len(texts)} text chunks")
|
113 |
|
114 |
# Create a dict vector store
|
|
|
|
|
115 |
|
116 |
+
vector_db = await vector_db.abuild_from_list(texts)
|
117 |
|
118 |
+
chat_openai = ChatOpenAI()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
119 |
|
120 |
+
#Create a chain
|
121 |
+
retrieval_augmented_qa_pipeline = RetrievalAugmentedQAPipeline(
|
122 |
+
vector_db_retriever=vector_db,
|
123 |
+
llm=chat_openai
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
124 |
)
|
|
|
125 |
|
126 |
# Let the user know that the system is ready
|
127 |
msg.content = f"Processing `{file.name}` done. You can now ask questions!"
|
128 |
await msg.update()
|
129 |
|
130 |
+
cl.user_session.set("chain", retrieval_augmented_qa_pipeline)
|
131 |
|
132 |
|
133 |
@cl.on_message
|
|
|
135 |
chain = cl.user_session.get("chain")
|
136 |
|
137 |
msg = cl.Message(content="")
|
138 |
+
result = await chain.arun_pipeline(message.content)
|
|
|
|
|
139 |
|
140 |
async for stream_resp in result["response"]:
|
141 |
await msg.stream_token(stream_resp)
|
requirements.txt
CHANGED
@@ -7,4 +7,3 @@ pypdf
|
|
7 |
sentence_transformers
|
8 |
langchain_text_splitters
|
9 |
langchain-community
|
10 |
-
faiss-cpu
|
|
|
7 |
sentence_transformers
|
8 |
langchain_text_splitters
|
9 |
langchain-community
|
|