Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -17,23 +17,17 @@ from langchain_community.embeddings import HuggingFaceInferenceAPIEmbeddings
|
|
| 17 |
from langchain_community.vectorstores import FAISS
|
| 18 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
| 19 |
|
| 20 |
-
# ======================
|
| 21 |
-
# SECRETS CONFIGURATION
|
| 22 |
-
# ======================
|
| 23 |
-
# Get API keys from Hugging Face Secrets
|
| 24 |
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
|
| 25 |
HUGGINGFACE_ACCESS_TOKEN = os.environ.get("HUGGINGFACE_ACCESS_TOKEN")
|
| 26 |
|
| 27 |
-
# Validate required secrets
|
| 28 |
if not GOOGLE_API_KEY:
|
| 29 |
-
st.error("β GOOGLE_API_KEY not found.
|
| 30 |
st.stop()
|
| 31 |
|
| 32 |
if not HUGGINGFACE_ACCESS_TOKEN:
|
| 33 |
-
st.error("β HUGGINGFACE_ACCESS_TOKEN not found.
|
| 34 |
st.stop()
|
| 35 |
|
| 36 |
-
# Initialize LLM and embeddings with secrets
|
| 37 |
llm = ChatGoogleGenerativeAI(
|
| 38 |
model="gemini-1.5-pro",
|
| 39 |
google_api_key=GOOGLE_API_KEY
|
|
@@ -44,9 +38,6 @@ embeddings = HuggingFaceInferenceAPIEmbeddings(
|
|
| 44 |
model_name="BAAI/bge-small-en-v1.5"
|
| 45 |
)
|
| 46 |
|
| 47 |
-
# ======================
|
| 48 |
-
# DOCUMENT ANALYSIS CLASSES
|
| 49 |
-
# ======================
|
| 50 |
class KeyPoint(BaseModel):
|
| 51 |
point: str = Field(description="A key point extracted from the document.")
|
| 52 |
|
|
@@ -57,9 +48,6 @@ class DocumentAnalysis(BaseModel):
|
|
| 57 |
key_points: List[KeyPoint] = Field(description="List of key points from the document.")
|
| 58 |
summary: Summary = Field(description="Summary of the document.")
|
| 59 |
|
| 60 |
-
# ======================
|
| 61 |
-
# CHAIN SETUP
|
| 62 |
-
# ======================
|
| 63 |
parser = PydanticOutputParser(pydantic_object=DocumentAnalysis)
|
| 64 |
|
| 65 |
prompt_template = """
|
|
@@ -75,9 +63,6 @@ prompt = PromptTemplate(
|
|
| 75 |
|
| 76 |
chain = LLMChain(llm=llm, prompt=prompt, output_parser=parser)
|
| 77 |
|
| 78 |
-
# ======================
|
| 79 |
-
# UTILITY FUNCTIONS
|
| 80 |
-
# ======================
|
| 81 |
def analyze_text_structured(text):
|
| 82 |
return chain.run(text=text)
|
| 83 |
|
|
@@ -99,7 +84,7 @@ def create_pdf_report(analysis):
|
|
| 99 |
pdf.cell(200, 10, txt="PDF Analysis Report", ln=True, align='C')
|
| 100 |
pdf.cell(200, 10, txt=f"Generated on: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}", ln=True, align='C')
|
| 101 |
pdf.multi_cell(0, 10, txt=json_to_text(analysis))
|
| 102 |
-
return pdf.output(dest='S')
|
| 103 |
|
| 104 |
def create_word_report(analysis):
|
| 105 |
doc = Document()
|
|
@@ -112,9 +97,6 @@ def create_word_report(analysis):
|
|
| 112 |
docx_bytes.seek(0)
|
| 113 |
return docx_bytes.getvalue()
|
| 114 |
|
| 115 |
-
# ======================
|
| 116 |
-
# STREAMLIT UI
|
| 117 |
-
# ======================
|
| 118 |
st.set_page_config(page_title="Chat With PDF", page_icon="π")
|
| 119 |
|
| 120 |
def local_css():
|
|
@@ -201,7 +183,6 @@ def local_css():
|
|
| 201 |
|
| 202 |
local_css()
|
| 203 |
|
| 204 |
-
# Initialize session state
|
| 205 |
if "current_file" not in st.session_state:
|
| 206 |
st.session_state.current_file = None
|
| 207 |
if "pdf_summary" not in st.session_state:
|
|
@@ -217,14 +198,12 @@ if "vectorstore" not in st.session_state:
|
|
| 217 |
if "messages" not in st.session_state:
|
| 218 |
st.session_state.messages = []
|
| 219 |
|
| 220 |
-
# UI Components
|
| 221 |
st.markdown('<div class="main-header">', unsafe_allow_html=True)
|
| 222 |
st.markdown('<div class="flag-stripe"></div>', unsafe_allow_html=True)
|
| 223 |
st.title("π Chat With PDF")
|
| 224 |
st.caption("Your AI-powered Document Analyzer")
|
| 225 |
st.markdown('</div>', unsafe_allow_html=True)
|
| 226 |
|
| 227 |
-
# File Uploader
|
| 228 |
with st.container():
|
| 229 |
st.markdown('<div class="card animate-fadeIn">', unsafe_allow_html=True)
|
| 230 |
uploaded_file = st.file_uploader("Upload a PDF file", type="pdf")
|
|
@@ -274,8 +253,7 @@ with st.container():
|
|
| 274 |
)
|
| 275 |
st.markdown('</div>', unsafe_allow_html=True)
|
| 276 |
|
| 277 |
-
|
| 278 |
-
if "vectorstore" in st.session_state:
|
| 279 |
st.subheader("Chat with the Document")
|
| 280 |
|
| 281 |
for message in st.session_state.messages:
|
|
@@ -303,7 +281,6 @@ if "vectorstore" in st.session_state:
|
|
| 303 |
|
| 304 |
st.session_state.messages.append({"role": "assistant", "content": response.content})
|
| 305 |
|
| 306 |
-
# Footer
|
| 307 |
st.markdown(
|
| 308 |
f'<div class="footer">Analysis Time: {st.session_state.analysis_time:.1f}s | Powered by Google Generative AI</div>',
|
| 309 |
unsafe_allow_html=True
|
|
|
|
| 17 |
from langchain_community.vectorstores import FAISS
|
| 18 |
from langchain_text_splitters import RecursiveCharacterTextSplitter
|
| 19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
|
| 21 |
HUGGINGFACE_ACCESS_TOKEN = os.environ.get("HUGGINGFACE_ACCESS_TOKEN")
|
| 22 |
|
|
|
|
| 23 |
if not GOOGLE_API_KEY:
|
| 24 |
+
st.error("β GOOGLE_API_KEY not found.")
|
| 25 |
st.stop()
|
| 26 |
|
| 27 |
if not HUGGINGFACE_ACCESS_TOKEN:
|
| 28 |
+
st.error("β HUGGINGFACE_ACCESS_TOKEN not found.")
|
| 29 |
st.stop()
|
| 30 |
|
|
|
|
| 31 |
llm = ChatGoogleGenerativeAI(
|
| 32 |
model="gemini-1.5-pro",
|
| 33 |
google_api_key=GOOGLE_API_KEY
|
|
|
|
| 38 |
model_name="BAAI/bge-small-en-v1.5"
|
| 39 |
)
|
| 40 |
|
|
|
|
|
|
|
|
|
|
| 41 |
class KeyPoint(BaseModel):
|
| 42 |
point: str = Field(description="A key point extracted from the document.")
|
| 43 |
|
|
|
|
| 48 |
key_points: List[KeyPoint] = Field(description="List of key points from the document.")
|
| 49 |
summary: Summary = Field(description="Summary of the document.")
|
| 50 |
|
|
|
|
|
|
|
|
|
|
| 51 |
parser = PydanticOutputParser(pydantic_object=DocumentAnalysis)
|
| 52 |
|
| 53 |
prompt_template = """
|
|
|
|
| 63 |
|
| 64 |
chain = LLMChain(llm=llm, prompt=prompt, output_parser=parser)
|
| 65 |
|
|
|
|
|
|
|
|
|
|
| 66 |
def analyze_text_structured(text):
|
| 67 |
return chain.run(text=text)
|
| 68 |
|
|
|
|
| 84 |
pdf.cell(200, 10, txt="PDF Analysis Report", ln=True, align='C')
|
| 85 |
pdf.cell(200, 10, txt=f"Generated on: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}", ln=True, align='C')
|
| 86 |
pdf.multi_cell(0, 10, txt=json_to_text(analysis))
|
| 87 |
+
return bytes(pdf.output(dest='S').encode('latin-1'))
|
| 88 |
|
| 89 |
def create_word_report(analysis):
|
| 90 |
doc = Document()
|
|
|
|
| 97 |
docx_bytes.seek(0)
|
| 98 |
return docx_bytes.getvalue()
|
| 99 |
|
|
|
|
|
|
|
|
|
|
| 100 |
st.set_page_config(page_title="Chat With PDF", page_icon="π")
|
| 101 |
|
| 102 |
def local_css():
|
|
|
|
| 183 |
|
| 184 |
local_css()
|
| 185 |
|
|
|
|
| 186 |
if "current_file" not in st.session_state:
|
| 187 |
st.session_state.current_file = None
|
| 188 |
if "pdf_summary" not in st.session_state:
|
|
|
|
| 198 |
if "messages" not in st.session_state:
|
| 199 |
st.session_state.messages = []
|
| 200 |
|
|
|
|
| 201 |
st.markdown('<div class="main-header">', unsafe_allow_html=True)
|
| 202 |
st.markdown('<div class="flag-stripe"></div>', unsafe_allow_html=True)
|
| 203 |
st.title("π Chat With PDF")
|
| 204 |
st.caption("Your AI-powered Document Analyzer")
|
| 205 |
st.markdown('</div>', unsafe_allow_html=True)
|
| 206 |
|
|
|
|
| 207 |
with st.container():
|
| 208 |
st.markdown('<div class="card animate-fadeIn">', unsafe_allow_html=True)
|
| 209 |
uploaded_file = st.file_uploader("Upload a PDF file", type="pdf")
|
|
|
|
| 253 |
)
|
| 254 |
st.markdown('</div>', unsafe_allow_html=True)
|
| 255 |
|
| 256 |
+
if "vectorstore" in st.session_state and st.session_state.vectorstore is not None:
|
|
|
|
| 257 |
st.subheader("Chat with the Document")
|
| 258 |
|
| 259 |
for message in st.session_state.messages:
|
|
|
|
| 281 |
|
| 282 |
st.session_state.messages.append({"role": "assistant", "content": response.content})
|
| 283 |
|
|
|
|
| 284 |
st.markdown(
|
| 285 |
f'<div class="footer">Analysis Time: {st.session_state.analysis_time:.1f}s | Powered by Google Generative AI</div>',
|
| 286 |
unsafe_allow_html=True
|