Spaces:
Sleeping
Sleeping
Commit
•
55216cd
1
Parent(s):
3a8c659
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,153 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import io
|
3 |
+
import re
|
4 |
+
import numpy as np
|
5 |
+
import pytesseract
|
6 |
+
from PIL import Image
|
7 |
+
from typing import List
|
8 |
+
from sentence_transformers import SentenceTransformer
|
9 |
+
from langchain_community.vectorstores import Chroma
|
10 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
11 |
+
from langchain_community.embeddings import SentenceTransformerEmbeddings
|
12 |
+
from groq import Groq
|
13 |
+
import gradio as gr
|
14 |
+
import requests
|
15 |
+
|
16 |
+
# Ensure the Tesseract OCR path is set correctly
|
17 |
+
pytesseract.pytesseract.tesseract_cmd = r'/usr/bin/tesseract'
|
18 |
+
|
19 |
+
GROQ_API_KEY = "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"
|
20 |
+
|
21 |
+
def preprocess_text(text):
|
22 |
+
try:
|
23 |
+
text = text.replace('\n', ' ').replace('\r', ' ')
|
24 |
+
text = re.sub(r'[^\x00-\x7F]+', ' ', text)
|
25 |
+
text = text.lower()
|
26 |
+
text = re.sub(r'[^\w\s]', '', text)
|
27 |
+
text = re.sub(r'\s+', ' ', text).strip()
|
28 |
+
return text
|
29 |
+
except Exception as e:
|
30 |
+
print("Failed to preprocess text:", e)
|
31 |
+
return ""
|
32 |
+
|
33 |
+
def fetch_text_file_from_huggingface_space():
|
34 |
+
url = "https://huggingface.co/spaces/Luciferalive/goosev9/blob/main/extracted_text.txt"
|
35 |
+
try:
|
36 |
+
response = requests.get(url)
|
37 |
+
response.raise_for_status()
|
38 |
+
text_content = response.text
|
39 |
+
print("Successfully downloaded the text file")
|
40 |
+
return text_content
|
41 |
+
except Exception as e:
|
42 |
+
print(f"Failed to download the text file: {e}")
|
43 |
+
return ""
|
44 |
+
|
45 |
+
def create_vector_store(text_content):
|
46 |
+
embeddings = SentenceTransformerEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
|
47 |
+
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
|
48 |
+
texts = text_splitter.split_text(text_content)
|
49 |
+
if not texts:
|
50 |
+
print("No text chunks created.")
|
51 |
+
return None
|
52 |
+
|
53 |
+
vector_store = Chroma.from_texts(texts, embeddings, collection_name="insurance_cosine")
|
54 |
+
print("Vector DB Successfully Created!")
|
55 |
+
return vector_store
|
56 |
+
|
57 |
+
def load_vector_store():
|
58 |
+
embeddings = SentenceTransformerEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2")
|
59 |
+
try:
|
60 |
+
db = Chroma(embedding_function=embeddings, collection_name="insurance_cosine")
|
61 |
+
print("Vector DB Successfully Loaded!")
|
62 |
+
return db
|
63 |
+
except Exception as e:
|
64 |
+
print("Failed to load Vector DB:", e)
|
65 |
+
return None
|
66 |
+
|
67 |
+
def answer_query(query):
|
68 |
+
try:
|
69 |
+
vector_store = load_vector_store()
|
70 |
+
if not vector_store:
|
71 |
+
return None
|
72 |
+
|
73 |
+
docs = vector_store.similarity_search(query)
|
74 |
+
print(f"\n\nDocuments retrieved: {len(docs)}")
|
75 |
+
|
76 |
+
if not docs:
|
77 |
+
print("No documents match the query.")
|
78 |
+
return None
|
79 |
+
|
80 |
+
docs_content = [doc.page_content for doc in docs]
|
81 |
+
all_docs_content = " ".join(docs_content)
|
82 |
+
|
83 |
+
client = Groq(api_key=GROQ_API_KEY)
|
84 |
+
template = """
|
85 |
+
### [INST] Instruction:
|
86 |
+
You are an AI assistant named Goose. Your purpose is to provide accurate, relevant, and helpful information to users in a friendly, warm, and supportive manner, similar to ChatGPT. When responding to queries, please keep the following guidelines in mind:
|
87 |
+
- When someone says hi, or small talk, only respond in a sentence.
|
88 |
+
- Retrieve relevant information from your knowledge base to formulate accurate and informative responses.
|
89 |
+
- Always maintain a positive, friendly, and encouraging tone in your interactions with users.
|
90 |
+
- Strictly write crisp and clear answers, don't write unnecessary stuff.
|
91 |
+
- Only answer the asked question, don't hallucinate or print any pre-information.
|
92 |
+
- After providing the answer, always ask for any other help needed in the next paragraph.
|
93 |
+
- Writing in bullet format is our top preference.
|
94 |
+
Remember, your goal is to be a reliable, friendly, and supportive AI assistant that provides accurate information while creating a positive user experience, just like ChatGPT. Adapt your communication style to best suit each user's needs and preferences.
|
95 |
+
### Docs: {docs}
|
96 |
+
### Question: {question}
|
97 |
+
"""
|
98 |
+
|
99 |
+
chat_completion = client.chat.completions.create(
|
100 |
+
messages=[
|
101 |
+
{
|
102 |
+
"role": "system",
|
103 |
+
"content": template.format(docs=all_docs_content, question=query)
|
104 |
+
},
|
105 |
+
{
|
106 |
+
"role": "user",
|
107 |
+
"content": query
|
108 |
+
}
|
109 |
+
],
|
110 |
+
model="llama3-8b-8192",
|
111 |
+
)
|
112 |
+
|
113 |
+
answer = chat_completion.choices[0].message.content.strip()
|
114 |
+
return answer
|
115 |
+
except Exception as e:
|
116 |
+
print("An error occurred while getting the answer: ", str(e))
|
117 |
+
return None
|
118 |
+
|
119 |
+
def process_query(query):
|
120 |
+
try:
|
121 |
+
response = answer_query(query)
|
122 |
+
if response:
|
123 |
+
return "Answer: " + response
|
124 |
+
else:
|
125 |
+
return "No answer found."
|
126 |
+
except Exception as e:
|
127 |
+
print("An error occurred while getting the answer: ", str(e))
|
128 |
+
return "An error occurred: " + str(e)
|
129 |
+
|
130 |
+
|
131 |
+
# Set up the Gradio interface
|
132 |
+
def launch_assistant():
|
133 |
+
text_content = fetch_text_file_from_huggingface_space()
|
134 |
+
if not text_content.strip():
|
135 |
+
print("No text content fetched.")
|
136 |
+
return
|
137 |
+
|
138 |
+
vector_store = create_vector_store(text_content)
|
139 |
+
if not vector_store:
|
140 |
+
print("Failed to create Vector DB.")
|
141 |
+
return
|
142 |
+
|
143 |
+
iface = gr.Interface(
|
144 |
+
fn=process_query,
|
145 |
+
inputs=gr.Textbox(lines=7, label="Enter your question"),
|
146 |
+
outputs="text",
|
147 |
+
title="Goose AI Assistant",
|
148 |
+
description="Ask a question and get an answer from the AI assistant."
|
149 |
+
)
|
150 |
+
|
151 |
+
iface.launch()
|
152 |
+
|
153 |
+
launch_assistant()
|