Spaces:
Sleeping
Sleeping
danicafisher
commited on
Commit
•
d643080
1
Parent(s):
d78b065
Reverts
Browse files- app.py +51 -65
- requirements.txt +3 -2
app.py
CHANGED
@@ -5,20 +5,20 @@ from aimakerspace.openai_utils.prompts import (
|
|
5 |
UserRolePrompt,
|
6 |
SystemRolePrompt,
|
7 |
)
|
8 |
-
|
9 |
from aimakerspace.vectordatabase import VectorDatabase
|
10 |
-
|
11 |
import chainlit as cl
|
12 |
# import asyncio
|
13 |
-
from operator import itemgetter
|
14 |
import nest_asyncio
|
15 |
nest_asyncio.apply()
|
16 |
from langchain_community.document_loaders import PyMuPDFLoader
|
17 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
18 |
-
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
|
19 |
-
from langchain_community.vectorstores import Qdrant
|
20 |
-
from langchain.prompts import ChatPromptTemplate
|
21 |
-
from langchain_core.runnables import RunnablePassthrough
|
22 |
|
23 |
|
24 |
filepath_NIST = "data/NIST.AI.600-1.pdf"
|
@@ -36,15 +36,15 @@ text_splitter = RecursiveCharacterTextSplitter(
|
|
36 |
|
37 |
rag_documents = text_splitter.split_documents(documents)
|
38 |
|
39 |
-
embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
|
40 |
|
41 |
-
vectorstore = Qdrant.from_documents(
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
)
|
47 |
-
retriever = qdrant_vectorstore.as_retriever()
|
48 |
|
49 |
RAG_PROMPT = """\
|
50 |
Given a provided context and question, you must answer the question based only on context.
|
@@ -55,46 +55,46 @@ Context: {context}
|
|
55 |
Question: {question}
|
56 |
"""
|
57 |
|
58 |
-
prompt = ChatPromptTemplate.from_template(RAG_PROMPT)
|
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 |
# ------------------------------------------------------------
|
@@ -112,36 +112,22 @@ async def start_chat():
|
|
112 |
# }
|
113 |
|
114 |
# # Create a dict vector store
|
115 |
-
|
116 |
-
|
117 |
# vector_db = await vector_db.abuild_from_list(split_documents_NIST)
|
118 |
# vector_db = await vector_db.abuild_from_list(split_documents_Blueprint)
|
119 |
|
120 |
# # chat_openai = ChatOpenAI()
|
121 |
-
|
122 |
|
123 |
|
124 |
# # Create a chain
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
# )
|
129 |
-
primary_llm = ChatOpenAI(model_name="gpt-4o-mini", temperature=0)
|
130 |
-
|
131 |
-
rag_chain = (
|
132 |
-
# INVOKE CHAIN WITH: {"question" : "<<SOME USER QUESTION>>"}
|
133 |
-
# "question" : populated by getting the value of the "question" key
|
134 |
-
# "context" : populated by getting the value of the "question" key and chaining it into the base_retriever
|
135 |
-
{"context": itemgetter("question") | retriever, "question": itemgetter("question")}
|
136 |
-
# "context" : is assigned to a RunnablePassthrough object (will not be called or considered in the next step)
|
137 |
-
# by getting the value of the "context" key from the previous step
|
138 |
-
| RunnablePassthrough.assign(context=itemgetter("context"))
|
139 |
-
# "response" : the "context" and "question" values are used to format our prompt object and then piped
|
140 |
-
# into the LLM and stored in a key called "response"
|
141 |
-
# "context" : populated by getting the value of the "context" key from the previous step
|
142 |
-
| {"response": prompt | primary_llm, "context": itemgetter("context")}
|
143 |
)
|
144 |
|
|
|
145 |
# cl.user_session.set("settings", settings)
|
146 |
cl.user_session.set("chain", rag_chain)
|
147 |
|
|
|
5 |
UserRolePrompt,
|
6 |
SystemRolePrompt,
|
7 |
)
|
8 |
+
from aimakerspace.openai_utils.embedding import EmbeddingModel
|
9 |
from aimakerspace.vectordatabase import VectorDatabase
|
10 |
+
from aimakerspace.openai_utils.chatmodel import ChatOpenAI
|
11 |
import chainlit as cl
|
12 |
# import asyncio
|
13 |
+
# from operator import itemgetter
|
14 |
import nest_asyncio
|
15 |
nest_asyncio.apply()
|
16 |
from langchain_community.document_loaders import PyMuPDFLoader
|
17 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
18 |
+
# from langchain_openai import ChatOpenAI, OpenAIEmbeddings
|
19 |
+
# from langchain_community.vectorstores import Qdrant
|
20 |
+
# from langchain.prompts import ChatPromptTemplate
|
21 |
+
# from langchain_core.runnables import RunnablePassthrough
|
22 |
|
23 |
|
24 |
filepath_NIST = "data/NIST.AI.600-1.pdf"
|
|
|
36 |
|
37 |
rag_documents = text_splitter.split_documents(documents)
|
38 |
|
39 |
+
# embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
|
40 |
|
41 |
+
# vectorstore = Qdrant.from_documents(
|
42 |
+
# documents=rag_documents,
|
43 |
+
# embedding=embeddings,
|
44 |
+
# location=":memory:",
|
45 |
+
# collection_name="Implications of AI"
|
46 |
+
# )
|
47 |
+
# retriever = qdrant_vectorstore.as_retriever()
|
48 |
|
49 |
RAG_PROMPT = """\
|
50 |
Given a provided context and question, you must answer the question based only on context.
|
|
|
55 |
Question: {question}
|
56 |
"""
|
57 |
|
58 |
+
# prompt = ChatPromptTemplate.from_template(RAG_PROMPT)
|
59 |
|
60 |
+
RAG_PROMPT_TEMPLATE = """ \
|
61 |
+
Use the provided context to answer the user's query.
|
62 |
+
You may not answer the user's query unless there is specific context in the following text.
|
63 |
+
If you do not know the answer, or cannot answer, please respond with "I don't know".
|
64 |
+
"""
|
65 |
|
66 |
+
rag_prompt = SystemRolePrompt(RAG_PROMPT_TEMPLATE)
|
67 |
|
68 |
+
USER_PROMPT_TEMPLATE = """ \
|
69 |
+
Context:
|
70 |
+
{context}
|
71 |
+
User Query:
|
72 |
+
{user_query}
|
73 |
+
"""
|
74 |
|
75 |
+
user_prompt = UserRolePrompt(USER_PROMPT_TEMPLATE)
|
76 |
|
77 |
+
class RetrievalAugmentedQAPipeline:
|
78 |
+
def __init__(self, llm: ChatOpenAI(), vector_db_retriever: VectorDatabase) -> None:
|
79 |
+
self.llm = llm
|
80 |
+
self.vector_db_retriever = vector_db_retriever
|
81 |
|
82 |
+
async def arun_pipeline(self, user_query: str):
|
83 |
+
context_list = self.vector_db_retriever.search_by_text(user_query, k=4)
|
84 |
|
85 |
+
context_prompt = ""
|
86 |
+
for context in context_list:
|
87 |
+
context_prompt += context[0] + "\n"
|
88 |
|
89 |
+
formatted_system_prompt = rag_prompt.create_message()
|
90 |
|
91 |
+
formatted_user_prompt = user_prompt.create_message(user_query=user_query, context=context_prompt)
|
92 |
|
93 |
+
async def generate_response():
|
94 |
+
async for chunk in self.llm.astream([formatted_system_prompt, formatted_user_prompt]):
|
95 |
+
yield chunk
|
96 |
|
97 |
+
return {"response": generate_response(), "context": context_list}
|
98 |
|
99 |
|
100 |
# ------------------------------------------------------------
|
|
|
112 |
# }
|
113 |
|
114 |
# # Create a dict vector store
|
115 |
+
vector_db = VectorDatabase()
|
116 |
+
vector_db = await vector_db.abuild_from_list(rag_documents)
|
117 |
# vector_db = await vector_db.abuild_from_list(split_documents_NIST)
|
118 |
# vector_db = await vector_db.abuild_from_list(split_documents_Blueprint)
|
119 |
|
120 |
# # chat_openai = ChatOpenAI()
|
121 |
+
llm = ChatOpenAI(model="gpt-4o-mini", tags=["base_llm"])
|
122 |
|
123 |
|
124 |
# # Create a chain
|
125 |
+
rag_chain = RetrievalAugmentedQAPipeline(
|
126 |
+
vector_db_retriever=vector_db,
|
127 |
+
llm=llm
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
128 |
)
|
129 |
|
130 |
+
|
131 |
# cl.user_session.set("settings", settings)
|
132 |
cl.user_session.set("chain", rag_chain)
|
133 |
|
requirements.txt
CHANGED
@@ -7,5 +7,6 @@ langchain
|
|
7 |
langchain-core
|
8 |
langchain-community
|
9 |
langchain-text-splitters
|
10 |
-
langchain-openai
|
11 |
-
qdrant-client
|
|
|
|
7 |
langchain-core
|
8 |
langchain-community
|
9 |
langchain-text-splitters
|
10 |
+
# langchain-openai
|
11 |
+
# qdrant-client
|
12 |
+
# langchain-qdrant
|