Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,116 +1,3 @@
|
|
1 |
-
import os
|
2 |
-
import time
|
3 |
import gradio as gr
|
4 |
-
from langchain_community.vectorstores import FAISS
|
5 |
-
from langchain_community.embeddings import HuggingFaceEmbeddings
|
6 |
-
from langchain.prompts import PromptTemplate
|
7 |
-
from langchain.memory import ConversationBufferWindowMemory
|
8 |
-
from langchain.chains import ConversationalRetrievalChain
|
9 |
-
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
10 |
-
import torch
|
11 |
|
12 |
-
|
13 |
-
def load_embeddings():
|
14 |
-
return HuggingFaceEmbeddings(model_name="law-ai/InLegalBERT")
|
15 |
-
|
16 |
-
def load_faiss_db():
|
17 |
-
embeddings = load_embeddings()
|
18 |
-
return FAISS.load_local("ipc_embed_db", embeddings, allow_dangerous_deserialization=True)
|
19 |
-
|
20 |
-
# Load embeddings and FAISS database
|
21 |
-
embeddings = load_embeddings()
|
22 |
-
db = load_faiss_db()
|
23 |
-
db_retriever = db.as_retriever(search_type="similarity", search_kwargs={"k": 3})
|
24 |
-
|
25 |
-
# Define prompt template
|
26 |
-
prompt_template = """
|
27 |
-
<s>[INST]
|
28 |
-
As a legal chatbot specializing in the Indian Penal Code, you are tasked with providing highly accurate and contextually appropriate responses. Ensure your answers meet these criteria:
|
29 |
-
- Respond in a bullet-point format to clearly delineate distinct aspects of the legal query.
|
30 |
-
- Each point should accurately reflect the breadth of the legal provision in question, avoiding over-specificity unless directly relevant to the user's query.
|
31 |
-
- Clarify the general applicability of the legal rules or sections mentioned, highlighting any common misconceptions or frequently misunderstood aspects.
|
32 |
-
- Limit responses to essential information that directly addresses the user's question, providing concise yet comprehensive explanations.
|
33 |
-
- Avoid assuming specific contexts or details not provided in the query, focusing on delivering universally applicable legal interpretations unless otherwise specified.
|
34 |
-
- Conclude with a brief summary that captures the essence of the legal discussion and corrects any common misinterpretations related to the topic.
|
35 |
-
|
36 |
-
CONTEXT: {context}
|
37 |
-
CHAT HISTORY: {chat_history}
|
38 |
-
QUESTION: {question}
|
39 |
-
ANSWER:
|
40 |
-
- [Detail the first key aspect of the law, ensuring it reflects general application]
|
41 |
-
- [Provide a concise explanation of how the law is typically interpreted or applied]
|
42 |
-
- [Correct a common misconception or clarify a frequently misunderstood aspect]
|
43 |
-
- [Detail any exceptions to the general rule, if applicable]
|
44 |
-
- [Include any additional relevant information that directly relates to the user's query]
|
45 |
-
</s>[INST]
|
46 |
-
"""
|
47 |
-
|
48 |
-
prompt = PromptTemplate(template=prompt_template, input_variables=['context', 'question', 'chat_history'])
|
49 |
-
|
50 |
-
# Load the InLegalBERT model and tokenizer
|
51 |
-
tokenizer = AutoTokenizer.from_pretrained("law-ai/InLegalBERT")
|
52 |
-
model = AutoModelForSequenceClassification.from_pretrained("law-ai/InLegalBERT")
|
53 |
-
|
54 |
-
# Function to get the model's response
|
55 |
-
def get_inlegalbert_response(question):
|
56 |
-
inputs = tokenizer(question, return_tensors="pt")
|
57 |
-
with torch.no_grad():
|
58 |
-
outputs = model(**inputs)
|
59 |
-
logits = outputs.logits
|
60 |
-
response = tokenizer.decode(torch.argmax(logits, dim=-1))
|
61 |
-
return response
|
62 |
-
|
63 |
-
# Define a wrapper for the model
|
64 |
-
class InLegalBERTWrapper:
|
65 |
-
def __init__(self, model, tokenizer):
|
66 |
-
self.model = model
|
67 |
-
self.tokenizer = tokenizer
|
68 |
-
|
69 |
-
def __call__(self, prompt, **kwargs):
|
70 |
-
return {"text": get_inlegalbert_response(prompt)}
|
71 |
-
|
72 |
-
llm = InLegalBERTWrapper(model, tokenizer)
|
73 |
-
|
74 |
-
qa = ConversationalRetrievalChain.from_llm(
|
75 |
-
llm=llm,
|
76 |
-
memory=ConversationBufferWindowMemory(k=2, memory_key="chat_history", return_messages=True),
|
77 |
-
retriever=db_retriever,
|
78 |
-
combine_docs_chain_kwargs={'prompt': prompt}
|
79 |
-
)
|
80 |
-
|
81 |
-
def extract_answer(full_response):
|
82 |
-
answer_start = full_response.find("Response:")
|
83 |
-
if answer_start != -1:
|
84 |
-
answer_start += len("Response:")
|
85 |
-
return full_response[answer_start:].strip()
|
86 |
-
return full_response
|
87 |
-
|
88 |
-
def chat(input_prompt, messages):
|
89 |
-
if "messages" not in messages:
|
90 |
-
messages["messages"] = []
|
91 |
-
|
92 |
-
messages["messages"].append({"role": "user", "content": input_prompt})
|
93 |
-
|
94 |
-
result = qa.invoke(input=input_prompt)
|
95 |
-
answer = extract_answer(result["answer"])
|
96 |
-
|
97 |
-
messages["messages"].append({"role": "assistant", "content": answer})
|
98 |
-
|
99 |
-
return [(message["role"], message["content"]) for message in messages["messages"]], messages
|
100 |
-
|
101 |
-
with gr.Blocks() as demo:
|
102 |
-
gr.Markdown("## Stat.ai Legal Assistant")
|
103 |
-
chatbot = gr.Chatbot()
|
104 |
-
state = gr.State({"messages": []})
|
105 |
-
msg = gr.Textbox(placeholder="Ask Stat.ai")
|
106 |
-
|
107 |
-
def user_input(message, history):
|
108 |
-
history["messages"].append({"role": "user", "content": message})
|
109 |
-
return "", history
|
110 |
-
|
111 |
-
msg.submit(user_input, [msg, state], [msg, state], queue=False).then(
|
112 |
-
chat, [msg, state], [chatbot, state]
|
113 |
-
)
|
114 |
-
|
115 |
-
if __name__ == "__main__":
|
116 |
-
demo.launch()
|
|
|
|
|
|
|
1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
+
gr.load("models/law-ai/InLegalBERT").launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|