Spaces:
Running
Running
Jatin Mehra
commited on
Commit
·
a193f24
1
Parent(s):
f50be30
Refactor app.py to implement FastAPI for PDF processing, session management, and chat functionality
Browse files
app.py
CHANGED
@@ -1,196 +1,266 @@
|
|
1 |
import os
|
2 |
-
import
|
3 |
-
import json
|
4 |
-
import streamlit as st
|
5 |
-
from streamlit_chat import message
|
6 |
-
from preprocessing import Model
|
7 |
-
from io import BytesIO
|
8 |
import pickle
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
-
#
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
15 |
)
|
16 |
|
17 |
-
#
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
}
|
35 |
-
|
36 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
}
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
agent_response = st.session_state["assistant"].get_response(
|
63 |
-
user_input,
|
64 |
-
st.session_state["temperature"],
|
65 |
-
st.session_state["max_tokens"],
|
66 |
-
st.session_state["model"]
|
67 |
-
)
|
68 |
-
|
69 |
-
st.session_state["messages"].append((user_input, True))
|
70 |
-
st.session_state["messages"].append((agent_response, False))
|
71 |
-
st.session_state["user_input"] = ""
|
72 |
-
|
73 |
-
# Save chat history temporarily on local storage
|
74 |
-
with open("chat_history.pkl", "wb") as f:
|
75 |
-
pickle.dump(st.session_state["messages"], f)
|
76 |
-
|
77 |
-
def process_file():
|
78 |
-
"""
|
79 |
-
Processes the uploaded PDF file and appends its content to the context.
|
80 |
-
"""
|
81 |
-
for file in st.session_state["file_uploader"]:
|
82 |
-
with tempfile.NamedTemporaryFile(delete=False) as tf:
|
83 |
-
tf.write(file.getbuffer())
|
84 |
-
file_path = tf.name
|
85 |
-
|
86 |
-
with st.session_state["process_file_spinner"], st.spinner(f"Processing {file.name}..."):
|
87 |
-
try:
|
88 |
-
st.session_state["assistant"].add_to_context(file_path)
|
89 |
-
except Exception as e:
|
90 |
-
st.error(f"Failed to process file {file.name}: {str(e)}")
|
91 |
-
os.remove(file_path)
|
92 |
-
|
93 |
-
def download_chat_history():
|
94 |
-
"""
|
95 |
-
Allows users to download chat history in HTML or JSON format.
|
96 |
-
"""
|
97 |
-
# Convert messages to JSON format
|
98 |
-
chat_data = [{"role": "user" if is_user else "assistant", "content": msg} for msg, is_user in st.session_state["messages"]]
|
99 |
-
|
100 |
-
# Download as JSON
|
101 |
-
json_data = json.dumps(chat_data, indent=4)
|
102 |
-
st.download_button(
|
103 |
-
label="💾 Download Chat History as JSON",
|
104 |
-
data=json_data,
|
105 |
-
file_name="chat_history.json",
|
106 |
-
mime="application/json"
|
107 |
-
)
|
108 |
-
|
109 |
-
# Download as HTML
|
110 |
-
html_data = "<html><body><h1>Chat History</h1><ul>"
|
111 |
-
for entry in chat_data:
|
112 |
-
role = "User" if entry["role"] == "user" else "Assistant"
|
113 |
-
html_data += f"<li><strong>{role}:</strong> {entry['content']}</li>"
|
114 |
-
html_data += "</ul></body></html>"
|
115 |
-
st.download_button(
|
116 |
-
label="💾 Download Chat History as HTML",
|
117 |
-
data=html_data,
|
118 |
-
file_name="chat_history.html",
|
119 |
-
mime="text/html"
|
120 |
-
)
|
121 |
-
|
122 |
-
def main_page():
|
123 |
-
"""
|
124 |
-
Main function to set up the Streamlit UI and handle user interactions.
|
125 |
-
"""
|
126 |
-
# Initialize session state variables
|
127 |
-
if "messages" not in st.session_state:
|
128 |
-
st.session_state["messages"] = []
|
129 |
-
|
130 |
-
if "assistant" not in st.session_state:
|
131 |
-
st.session_state["assistant"] = Model()
|
132 |
-
|
133 |
-
if "user_input" not in st.session_state:
|
134 |
-
st.session_state["user_input"] = ""
|
135 |
-
|
136 |
-
if "temperature" not in st.session_state:
|
137 |
-
st.session_state["temperature"] = 0.5
|
138 |
-
|
139 |
-
if "max_tokens" not in st.session_state:
|
140 |
-
st.session_state["max_tokens"] = 550
|
141 |
-
|
142 |
-
if "model" not in st.session_state:
|
143 |
-
st.session_state["model"] = "llama-3.1-8b-instant"
|
144 |
-
|
145 |
-
# File uploader
|
146 |
-
st.subheader("📤 Upload Your PDF Documents")
|
147 |
-
st.file_uploader(
|
148 |
-
"Choose PDF files to analyze",
|
149 |
-
type=["pdf"],
|
150 |
-
key="file_uploader",
|
151 |
-
on_change=process_file,
|
152 |
-
accept_multiple_files=True,
|
153 |
-
)
|
154 |
-
|
155 |
-
st.session_state["process_file_spinner"] = st.empty()
|
156 |
-
|
157 |
-
# Document management section
|
158 |
-
if st.session_state["assistant"].contexts:
|
159 |
-
st.subheader("🗂️ Manage Uploaded Documents")
|
160 |
-
for i, context in enumerate(st.session_state["assistant"].contexts):
|
161 |
-
st.text_area(f"Document {i+1} Context", context[:500] + "..." if len(context) > 500 else context, height=100)
|
162 |
-
if st.button(f"Remove Document {i+1}"):
|
163 |
-
st.session_state["assistant"].remove_from_context(i)
|
164 |
-
|
165 |
-
# Model settings
|
166 |
-
with st.expander("⚙️ Customize AI Settings", expanded=True):
|
167 |
-
st.slider("Sampling Temperature", min_value=0.0, max_value=1.0, step=0.1, key="temperature", help="Higher values make output more random.")
|
168 |
-
st.slider("Max Tokens", min_value=750, max_value=5000, step=50, key="max_tokens", help="Limits the length of the response.")
|
169 |
-
st.selectbox("Choose AI Model", ["llama-3.1-8b-instant", "llama3-70b-8192", "gemma-7b-it"], key="model")
|
170 |
-
|
171 |
-
# Display messages and input box
|
172 |
-
display_messages()
|
173 |
-
st.text_input("Type your query and hit Enter", key="user_input", on_change=process_user_input, placeholder="Ask something about your documents...")
|
174 |
|
175 |
-
|
176 |
-
|
177 |
-
download_chat_history()
|
178 |
-
|
179 |
-
# Developer info and bug report
|
180 |
-
st.subheader("🐞 Bug Report")
|
181 |
-
st.markdown("""
|
182 |
-
If you encounter any bugs or issues while using the app, please send a bug report to the developer. You can include a screenshot (optional) to help identify the problem.\n
|
183 |
-
""")
|
184 |
-
st.subheader("💡 Suggestions")
|
185 |
-
st.markdown("""
|
186 |
-
Suggestions to improve the app's UI and user interface are also welcome. Feel free to reach out to the developer with your suggestions.\n
|
187 |
-
""")
|
188 |
-
st.subheader("👨💻 Developer Info")
|
189 |
-
st.markdown("""
|
190 |
-
**Developer**: Jatin Mehra\n
|
191 |
-
**Email**: [email protected]\n
|
192 |
-
**Mobile**: 9910364780\n
|
193 |
-
""")
|
194 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
195 |
if __name__ == "__main__":
|
196 |
-
|
|
|
|
1 |
import os
|
2 |
+
import dotenv
|
|
|
|
|
|
|
|
|
|
|
3 |
import pickle
|
4 |
+
import uuid
|
5 |
+
from fastapi import FastAPI, UploadFile, File, Form, HTTPException, BackgroundTasks, Request
|
6 |
+
from fastapi.responses import JSONResponse
|
7 |
+
from fastapi.middleware.cors import CORSMiddleware
|
8 |
+
from fastapi.staticfiles import StaticFiles
|
9 |
+
from pydantic import BaseModel
|
10 |
+
import uvicorn
|
11 |
+
from preprocessing import (
|
12 |
+
model_selection,
|
13 |
+
process_pdf_file,
|
14 |
+
chunk_text,
|
15 |
+
create_embeddings,
|
16 |
+
build_faiss_index,
|
17 |
+
retrieve_similar_chunks,
|
18 |
+
agentic_rag,
|
19 |
+
tools,
|
20 |
+
memory
|
21 |
+
)
|
22 |
+
from sentence_transformers import SentenceTransformer
|
23 |
+
import shutil
|
24 |
+
import traceback
|
25 |
+
|
26 |
+
# Load environment variables
|
27 |
+
dotenv.load_dotenv()
|
28 |
|
29 |
+
# Initialize FastAPI app
|
30 |
+
app = FastAPI(title="PDF Insight Beta", description="Agentic RAG for PDF documents")
|
31 |
+
|
32 |
+
# Add CORS middleware
|
33 |
+
app.add_middleware(
|
34 |
+
CORSMiddleware,
|
35 |
+
allow_origins=["*"],
|
36 |
+
allow_credentials=True,
|
37 |
+
allow_methods=["*"],
|
38 |
+
allow_headers=["*"],
|
39 |
)
|
40 |
|
41 |
+
# Create upload directory if it doesn't exist
|
42 |
+
UPLOAD_DIR = "uploads"
|
43 |
+
if not os.path.exists(UPLOAD_DIR):
|
44 |
+
os.makedirs(UPLOAD_DIR)
|
45 |
+
|
46 |
+
# Store active sessions
|
47 |
+
sessions = {}
|
48 |
+
|
49 |
+
# Define model for chat request
|
50 |
+
class ChatRequest(BaseModel):
|
51 |
+
session_id: str
|
52 |
+
query: str
|
53 |
+
use_search: bool = False
|
54 |
+
model_name: str = "meta-llama/llama-4-scout-17b-16e-instruct"
|
55 |
+
|
56 |
+
class SessionRequest(BaseModel):
|
57 |
+
session_id: str
|
58 |
+
|
59 |
+
# Function to save session data
|
60 |
+
def save_session(session_id, data):
|
61 |
+
sessions[session_id] = data
|
62 |
+
|
63 |
+
# Create a copy of data that is safe to pickle
|
64 |
+
pickle_safe_data = {
|
65 |
+
"file_path": data.get("file_path"),
|
66 |
+
"file_name": data.get("file_name"),
|
67 |
+
"chunks": data.get("chunks"),
|
68 |
+
"chat_history": data.get("chat_history", [])
|
69 |
+
}
|
70 |
+
|
71 |
+
# Persist to disk
|
72 |
+
with open(f"{UPLOAD_DIR}/{session_id}_session.pkl", "wb") as f:
|
73 |
+
pickle.dump(pickle_safe_data, f)
|
74 |
+
|
75 |
+
# Function to load session data
|
76 |
+
def load_session(session_id, model_name="meta-llama/llama-4-scout-17b-16e-instruct"):
|
77 |
+
try:
|
78 |
+
# Check if session is already in memory
|
79 |
+
if session_id in sessions:
|
80 |
+
return sessions[session_id], True
|
81 |
+
|
82 |
+
# Try to load from disk
|
83 |
+
file_path = f"{UPLOAD_DIR}/{session_id}_session.pkl"
|
84 |
+
if os.path.exists(file_path):
|
85 |
+
with open(file_path, "rb") as f:
|
86 |
+
data = pickle.load(f)
|
87 |
+
|
88 |
+
# Recreate non-pickled objects
|
89 |
+
if data.get("chunks") and data.get("file_path") and os.path.exists(data["file_path"]):
|
90 |
+
# Recreate model, embeddings and index
|
91 |
+
model = SentenceTransformer('all-MiniLM-L6-v2')
|
92 |
+
embeddings = create_embeddings(data["chunks"], model)
|
93 |
+
index = build_faiss_index(embeddings)
|
94 |
+
|
95 |
+
# Recreate LLM
|
96 |
+
llm = model_selection(model_name)
|
97 |
+
|
98 |
+
# Reconstruct full session data
|
99 |
+
data["model"] = model
|
100 |
+
data["index"] = index
|
101 |
+
data["llm"] = llm
|
102 |
+
|
103 |
+
# Store in memory
|
104 |
+
sessions[session_id] = data
|
105 |
+
return data, True
|
106 |
+
|
107 |
+
return None, False
|
108 |
+
except Exception as e:
|
109 |
+
print(f"Error loading session: {str(e)}")
|
110 |
+
return None, False
|
111 |
+
|
112 |
+
# Mount static files (we'll create these later)
|
113 |
+
app.mount("/static", StaticFiles(directory="static"), name="static")
|
114 |
+
|
115 |
+
# Route for the home page
|
116 |
+
@app.get("/")
|
117 |
+
async def read_root():
|
118 |
+
return {"status": "ok", "message": "PDF Insight Beta API is running"}
|
119 |
+
|
120 |
+
# Route to upload a PDF file
|
121 |
+
@app.post("/upload-pdf")
|
122 |
+
async def upload_pdf(
|
123 |
+
file: UploadFile = File(...),
|
124 |
+
model_name: str = Form("meta-llama/llama-4-scout-17b-16e-instruct")
|
125 |
+
):
|
126 |
+
# Generate a unique session ID
|
127 |
+
session_id = str(uuid.uuid4())
|
128 |
+
file_path = None
|
129 |
+
|
130 |
+
try:
|
131 |
+
# Save the uploaded file
|
132 |
+
file_path = f"{UPLOAD_DIR}/{session_id}_{file.filename}"
|
133 |
+
with open(file_path, "wb") as buffer:
|
134 |
+
shutil.copyfileobj(file.file, buffer)
|
135 |
+
|
136 |
+
# Check if API keys are set
|
137 |
+
if not os.getenv("GROQ_API_KEY"):
|
138 |
+
raise ValueError("GROQ_API_KEY is not set in the environment variables")
|
139 |
+
|
140 |
+
# Process the PDF
|
141 |
+
text = process_pdf_file(file_path)
|
142 |
+
chunks = chunk_text(text, max_length=1500)
|
143 |
+
|
144 |
+
# Create embeddings
|
145 |
+
model = SentenceTransformer('all-MiniLM-L6-v2')
|
146 |
+
embeddings = create_embeddings(chunks, model)
|
147 |
+
index = build_faiss_index(embeddings)
|
148 |
+
|
149 |
+
# Initialize LLM
|
150 |
+
llm = model_selection(model_name)
|
151 |
+
|
152 |
+
# Save session data
|
153 |
+
session_data = {
|
154 |
+
"file_path": file_path,
|
155 |
+
"file_name": file.filename,
|
156 |
+
"chunks": chunks,
|
157 |
+
"model": model,
|
158 |
+
"index": index,
|
159 |
+
"llm": llm,
|
160 |
+
"chat_history": []
|
161 |
}
|
162 |
+
save_session(session_id, session_data)
|
163 |
+
|
164 |
+
return {"status": "success", "session_id": session_id, "message": f"Processed {file.filename}"}
|
165 |
+
|
166 |
+
except Exception as e:
|
167 |
+
# Clean up on error
|
168 |
+
if file_path and os.path.exists(file_path):
|
169 |
+
os.remove(file_path)
|
170 |
+
|
171 |
+
error_msg = str(e)
|
172 |
+
stack_trace = traceback.format_exc()
|
173 |
+
print(f"Error processing PDF: {error_msg}")
|
174 |
+
print(f"Stack trace: {stack_trace}")
|
175 |
+
|
176 |
+
return JSONResponse(
|
177 |
+
status_code=400,
|
178 |
+
content={
|
179 |
+
"status": "error",
|
180 |
+
"detail": error_msg,
|
181 |
+
"type": type(e).__name__
|
182 |
+
}
|
183 |
+
)
|
184 |
+
|
185 |
+
# Route to chat with the document
|
186 |
+
@app.post("/chat")
|
187 |
+
async def chat(request: ChatRequest):
|
188 |
+
# Try to load session if not in memory
|
189 |
+
session, found = load_session(request.session_id, model_name=request.model_name)
|
190 |
+
if not found:
|
191 |
+
raise HTTPException(status_code=404, detail="Session not found. Please upload a document first.")
|
192 |
+
|
193 |
+
try:
|
194 |
+
# Retrieve similar chunks
|
195 |
+
similar_chunks = retrieve_similar_chunks(
|
196 |
+
request.query,
|
197 |
+
session["index"],
|
198 |
+
session["chunks"],
|
199 |
+
session["model"],
|
200 |
+
k=3
|
201 |
+
)
|
202 |
+
context = "\n".join([chunk for chunk, _ in similar_chunks])
|
203 |
+
|
204 |
+
# Generate response using agentic_rag
|
205 |
+
response = agentic_rag(
|
206 |
+
session["llm"],
|
207 |
+
tools,
|
208 |
+
query=request.query,
|
209 |
+
context=context,
|
210 |
+
Use_Tavily=request.use_search
|
211 |
+
)
|
212 |
+
|
213 |
+
# Update chat history
|
214 |
+
session["chat_history"].append({"user": request.query, "assistant": response["output"]})
|
215 |
+
save_session(request.session_id, session)
|
216 |
+
|
217 |
+
return {
|
218 |
+
"status": "success",
|
219 |
+
"answer": response["output"],
|
220 |
+
"context_used": [{"text": chunk, "score": float(score)} for chunk, score in similar_chunks]
|
221 |
}
|
222 |
+
|
223 |
+
except Exception as e:
|
224 |
+
raise HTTPException(status_code=500, detail=f"Error processing query: {str(e)}")
|
225 |
+
|
226 |
+
# Route to get chat history
|
227 |
+
@app.post("/chat-history")
|
228 |
+
async def get_chat_history(request: SessionRequest):
|
229 |
+
# Try to load session if not in memory
|
230 |
+
session, found = load_session(request.session_id)
|
231 |
+
if not found:
|
232 |
+
raise HTTPException(status_code=404, detail="Session not found")
|
233 |
+
|
234 |
+
return {
|
235 |
+
"status": "success",
|
236 |
+
"history": session.get("chat_history", [])
|
237 |
+
}
|
238 |
+
|
239 |
+
# Route to clear chat history
|
240 |
+
@app.post("/clear-history")
|
241 |
+
async def clear_history(request: SessionRequest):
|
242 |
+
# Try to load session if not in memory
|
243 |
+
session, found = load_session(request.session_id)
|
244 |
+
if not found:
|
245 |
+
raise HTTPException(status_code=404, detail="Session not found")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
246 |
|
247 |
+
session["chat_history"] = []
|
248 |
+
save_session(request.session_id, session)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
249 |
|
250 |
+
return {"status": "success", "message": "Chat history cleared"}
|
251 |
+
|
252 |
+
# Route to list available models
|
253 |
+
@app.get("/models")
|
254 |
+
async def get_models():
|
255 |
+
# You can expand this list as needed
|
256 |
+
models = [
|
257 |
+
{"id": "meta-llama/llama-4-scout-17b-16e-instruct", "name": "Llama 4 Scout 17B"},
|
258 |
+
{"id": "llama-3.1-8b-instant", "name": "Llama 3.1 8B Instant"},
|
259 |
+
{"id": "llama-3.3-70b-versatile", "name": "Llama 3.3 70B Versatile"},
|
260 |
+
]
|
261 |
+
return {"models": models}
|
262 |
+
|
263 |
+
# Run the application if this file is executed directly
|
264 |
if __name__ == "__main__":
|
265 |
+
uvicorn.run("app:app", host="0.0.0.0", port=8000, reload=True)
|
266 |
+
|