Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,11 +2,10 @@ import streamlit as st
|
|
| 2 |
import tempfile
|
| 3 |
import logging
|
| 4 |
from typing import List
|
| 5 |
-
from
|
| 6 |
-
#from langchain_community.embeddings import HuggingFaceEmbeddings
|
| 7 |
from langchain_huggingface import HuggingFaceEmbeddings
|
| 8 |
-
from
|
| 9 |
-
from
|
| 10 |
from langchain.chains.summarize import load_summarize_chain
|
| 11 |
from langchain.schema import Document
|
| 12 |
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
|
@@ -35,7 +34,7 @@ def load_embeddings():
|
|
| 35 |
def load_llm(model_name):
|
| 36 |
"""Load and cache the language model."""
|
| 37 |
try:
|
| 38 |
-
pipe = pipeline("
|
| 39 |
return HuggingFacePipeline(pipeline=pipe)
|
| 40 |
except Exception as e:
|
| 41 |
logger.error(f"Failed to load LLM: {e}")
|
|
@@ -48,7 +47,7 @@ def process_pdf(file) -> List[Document]:
|
|
| 48 |
with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as temp_file:
|
| 49 |
temp_file.write(file.getvalue())
|
| 50 |
temp_file_path = temp_file.name
|
| 51 |
-
|
| 52 |
loader = PyPDFLoader(file_path=temp_file_path)
|
| 53 |
pages = loader.load()
|
| 54 |
text_splitter = RecursiveCharacterTextSplitter(chunk_size=4000, chunk_overlap=200)
|
|
@@ -73,7 +72,6 @@ def summarize_report(documents: List[Document], llm) -> str:
|
|
| 73 |
try:
|
| 74 |
prompt_template = """
|
| 75 |
<s>[INST] You are an advanced AI assistant with expertise in summarizing technical documents. Your goal is to create a clear, concise, and well-organized summary using Markdown formatting. Focus on extracting and presenting the essential points of the document effectively.
|
| 76 |
-
|
| 77 |
*Instructions:*
|
| 78 |
- Analyze the provided context and input carefully.
|
| 79 |
- Identify and highlight the key points, main arguments, and important details.
|
|
@@ -82,30 +80,24 @@ def summarize_report(documents: List[Document], llm) -> str:
|
|
| 82 |
- Use **text** for important terms or concepts.
|
| 83 |
- Provide a brief introduction, followed by the main points, and a concluding summary if applicable.
|
| 84 |
- Ensure the summary is easy to read and understand, avoiding unnecessary jargon.
|
| 85 |
-
|
| 86 |
*Example Summary Format:*
|
| 87 |
-
|
| 88 |
# Overview
|
| 89 |
*Document Title:* Technical Analysis Report
|
| 90 |
-
|
| 91 |
*Summary:*
|
| 92 |
The report provides an in-depth analysis of the recent technical advancements in AI. It covers key areas such as ...
|
| 93 |
-
|
| 94 |
# Key Findings
|
| 95 |
- *Finding 1:* Description of finding 1.
|
| 96 |
- *Finding 2:* Description of finding 2.
|
| 97 |
-
|
| 98 |
# Conclusion
|
| 99 |
The analysis highlights the significant advancements and future directions for AI technology.
|
| 100 |
-
|
| 101 |
*Your Response:* [/INST]</s> {input}
|
| 102 |
Context: {context}
|
| 103 |
"""
|
| 104 |
|
| 105 |
prompt = PromptTemplate.from_template(prompt_template)
|
| 106 |
chain = load_summarize_chain(llm, chain_type="stuff", prompt=prompt)
|
| 107 |
-
summary = chain.
|
| 108 |
-
return summary
|
| 109 |
|
| 110 |
except Exception as e:
|
| 111 |
logger.error(f"Error summarizing report: {e}")
|
|
@@ -114,7 +106,7 @@ def summarize_report(documents: List[Document], llm) -> str:
|
|
| 114 |
|
| 115 |
def main():
|
| 116 |
st.title("Report Summarizer")
|
| 117 |
-
|
| 118 |
model_option = st.sidebar.text_input("Enter model name", value=DEFAULT_MODEL)
|
| 119 |
|
| 120 |
uploaded_file = st.sidebar.file_uploader("Upload your Report", type="pdf")
|
|
|
|
| 2 |
import tempfile
|
| 3 |
import logging
|
| 4 |
from typing import List
|
| 5 |
+
from langchain.document_loaders import PyPDFLoader
|
|
|
|
| 6 |
from langchain_huggingface import HuggingFaceEmbeddings
|
| 7 |
+
from langchain.vectorstores import FAISS
|
| 8 |
+
from langchain.llms import HuggingFacePipeline
|
| 9 |
from langchain.chains.summarize import load_summarize_chain
|
| 10 |
from langchain.schema import Document
|
| 11 |
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
|
|
|
| 34 |
def load_llm(model_name):
|
| 35 |
"""Load and cache the language model."""
|
| 36 |
try:
|
| 37 |
+
pipe = pipeline("text-generation", model=model_name, max_length=1024)
|
| 38 |
return HuggingFacePipeline(pipeline=pipe)
|
| 39 |
except Exception as e:
|
| 40 |
logger.error(f"Failed to load LLM: {e}")
|
|
|
|
| 47 |
with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as temp_file:
|
| 48 |
temp_file.write(file.getvalue())
|
| 49 |
temp_file_path = temp_file.name
|
| 50 |
+
|
| 51 |
loader = PyPDFLoader(file_path=temp_file_path)
|
| 52 |
pages = loader.load()
|
| 53 |
text_splitter = RecursiveCharacterTextSplitter(chunk_size=4000, chunk_overlap=200)
|
|
|
|
| 72 |
try:
|
| 73 |
prompt_template = """
|
| 74 |
<s>[INST] You are an advanced AI assistant with expertise in summarizing technical documents. Your goal is to create a clear, concise, and well-organized summary using Markdown formatting. Focus on extracting and presenting the essential points of the document effectively.
|
|
|
|
| 75 |
*Instructions:*
|
| 76 |
- Analyze the provided context and input carefully.
|
| 77 |
- Identify and highlight the key points, main arguments, and important details.
|
|
|
|
| 80 |
- Use **text** for important terms or concepts.
|
| 81 |
- Provide a brief introduction, followed by the main points, and a concluding summary if applicable.
|
| 82 |
- Ensure the summary is easy to read and understand, avoiding unnecessary jargon.
|
|
|
|
| 83 |
*Example Summary Format:*
|
|
|
|
| 84 |
# Overview
|
| 85 |
*Document Title:* Technical Analysis Report
|
|
|
|
| 86 |
*Summary:*
|
| 87 |
The report provides an in-depth analysis of the recent technical advancements in AI. It covers key areas such as ...
|
|
|
|
| 88 |
# Key Findings
|
| 89 |
- *Finding 1:* Description of finding 1.
|
| 90 |
- *Finding 2:* Description of finding 2.
|
|
|
|
| 91 |
# Conclusion
|
| 92 |
The analysis highlights the significant advancements and future directions for AI technology.
|
|
|
|
| 93 |
*Your Response:* [/INST]</s> {input}
|
| 94 |
Context: {context}
|
| 95 |
"""
|
| 96 |
|
| 97 |
prompt = PromptTemplate.from_template(prompt_template)
|
| 98 |
chain = load_summarize_chain(llm, chain_type="stuff", prompt=prompt)
|
| 99 |
+
summary = chain.run(documents)
|
| 100 |
+
return summary
|
| 101 |
|
| 102 |
except Exception as e:
|
| 103 |
logger.error(f"Error summarizing report: {e}")
|
|
|
|
| 106 |
|
| 107 |
def main():
|
| 108 |
st.title("Report Summarizer")
|
| 109 |
+
|
| 110 |
model_option = st.sidebar.text_input("Enter model name", value=DEFAULT_MODEL)
|
| 111 |
|
| 112 |
uploaded_file = st.sidebar.file_uploader("Upload your Report", type="pdf")
|