Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,31 +1,40 @@
|
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
|
| 3 |
-
# ✅
|
| 4 |
-
from langchain.text_splitter import CharacterTextSplitter
|
| 5 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
| 6 |
from langchain_community.vectorstores import FAISS
|
| 7 |
-
from langchain.chains import ConversationalRetrievalChain
|
| 8 |
from langchain_community.llms import HuggingFaceHub
|
| 9 |
from langchain_community.document_loaders import PyPDFLoader
|
| 10 |
|
| 11 |
-
# 1️⃣ Load your PDF
|
| 12 |
-
|
| 13 |
-
|
|
|
|
| 14 |
documents = loader.load()
|
| 15 |
|
| 16 |
-
# 2️⃣ Split into chunks
|
| 17 |
text_splitter = CharacterTextSplitter(chunk_size=800, chunk_overlap=100)
|
| 18 |
texts = text_splitter.split_documents(documents)
|
| 19 |
|
| 20 |
-
# 3️⃣ Create embeddings + vector
|
| 21 |
embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
|
| 22 |
db = FAISS.from_documents(texts, embeddings)
|
| 23 |
|
| 24 |
-
# 4️⃣ Build retriever-based chatbot
|
| 25 |
retriever = db.as_retriever(search_kwargs={"k": 3})
|
| 26 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
qa = ConversationalRetrievalChain.from_llm(
|
| 28 |
-
|
| 29 |
retriever=retriever
|
| 30 |
)
|
| 31 |
|
|
@@ -33,13 +42,54 @@ chat_history = []
|
|
| 33 |
|
| 34 |
def respond(message, history):
|
| 35 |
global chat_history
|
|
|
|
| 36 |
result = qa({"question": message, "chat_history": chat_history})
|
| 37 |
chat_history.append((message, result["answer"]))
|
| 38 |
-
return result["answer"]
|
| 39 |
|
| 40 |
-
# 5️⃣
|
| 41 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
|
| 43 |
-
|
| 44 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
import gradio as gr
|
| 3 |
|
| 4 |
+
# ✅ LangChain imports
|
| 5 |
+
from langchain.text_splitter import CharacterTextSplitter
|
| 6 |
from langchain_community.embeddings import HuggingFaceEmbeddings
|
| 7 |
from langchain_community.vectorstores import FAISS
|
| 8 |
+
from langchain.chains import ConversationalRetrievalChain
|
| 9 |
from langchain_community.llms import HuggingFaceHub
|
| 10 |
from langchain_community.document_loaders import PyPDFLoader
|
| 11 |
|
| 12 |
+
# --- 1️⃣ Load your PDF ---
|
| 13 |
+
current_dir = os.path.dirname(__file__)
|
| 14 |
+
pdf_path = os.path.join(current_dir, "chimera.pdf")
|
| 15 |
+
loader = PyPDFLoader(pdf_path)
|
| 16 |
documents = loader.load()
|
| 17 |
|
| 18 |
+
# --- 2️⃣ Split into chunks ---
|
| 19 |
text_splitter = CharacterTextSplitter(chunk_size=800, chunk_overlap=100)
|
| 20 |
texts = text_splitter.split_documents(documents)
|
| 21 |
|
| 22 |
+
# --- 3️⃣ Create embeddings + FAISS vector store ---
|
| 23 |
embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
|
| 24 |
db = FAISS.from_documents(texts, embeddings)
|
| 25 |
|
| 26 |
+
# --- 4️⃣ Build retriever-based chatbot ---
|
| 27 |
retriever = db.as_retriever(search_kwargs={"k": 3})
|
| 28 |
|
| 29 |
+
# --- Hugging Face Hub LLM setup with secret token ---
|
| 30 |
+
llm = HuggingFaceHub(
|
| 31 |
+
repo_id="google/flan-t5-base", # smaller for faster response
|
| 32 |
+
model_kwargs={"temperature":0},
|
| 33 |
+
huggingfacehub_api_token=os.environ.get("HUGGINGFACEHUB_API_TOKEN")
|
| 34 |
+
)
|
| 35 |
+
|
| 36 |
qa = ConversationalRetrievalChain.from_llm(
|
| 37 |
+
llm,
|
| 38 |
retriever=retriever
|
| 39 |
)
|
| 40 |
|
|
|
|
| 42 |
|
| 43 |
def respond(message, history):
|
| 44 |
global chat_history
|
| 45 |
+
chat_history = chat_history[-6:] # keep last 3 exchanges
|
| 46 |
result = qa({"question": message, "chat_history": chat_history})
|
| 47 |
chat_history.append((message, result["answer"]))
|
| 48 |
+
return result["answer"], chat_history
|
| 49 |
|
| 50 |
+
# --- 5️⃣ Gradio Blocks UI with Entry Warning ---
|
| 51 |
+
with gr.Blocks() as demo:
|
| 52 |
+
with gr.Column():
|
| 53 |
+
# Warning message
|
| 54 |
+
warning_text = gr.HTML(
|
| 55 |
+
"""
|
| 56 |
+
<div style="background-color:black;color:white;padding:20px;font-family:monospace;font-size:18px;">
|
| 57 |
+
⚠ WARNING — INVESTIGATIVE SIMULATION ⚠<br><br>
|
| 58 |
+
You are about to enter <b>The Chimera Case</b>, a high-stakes investigation into Innovate Future Labs (IFL) and Project Chimera.<br>
|
| 59 |
+
The scenario contains allegations, leaked files, and disputed testimonies. Treat every claim as unverified until verified by evidence.<br>
|
| 60 |
+
Your decisions and observations will guide your understanding of the case.<br><br>
|
| 61 |
+
Are you ready to proceed?
|
| 62 |
+
</div>
|
| 63 |
+
"""
|
| 64 |
+
)
|
| 65 |
|
| 66 |
+
# Buttons
|
| 67 |
+
enter_btn = gr.Button("Enter the Case")
|
| 68 |
+
exit_btn = gr.Button("Exit")
|
| 69 |
+
|
| 70 |
+
# Chatbot (hidden initially)
|
| 71 |
+
chatbot = gr.Chatbot(visible=False)
|
| 72 |
+
user_input = gr.Textbox(placeholder="Type your message here...", visible=False)
|
| 73 |
+
submit_btn = gr.Button("Send", visible=False)
|
| 74 |
+
|
| 75 |
+
# Button interactions
|
| 76 |
+
def enter_case():
|
| 77 |
+
chatbot.visible = True
|
| 78 |
+
user_input.visible = True
|
| 79 |
+
submit_btn.visible = True
|
| 80 |
+
warning_text.update(value="") # hide warning
|
| 81 |
+
enter_btn.visible = False
|
| 82 |
+
exit_btn.visible = False
|
| 83 |
+
return gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)
|
| 84 |
|
| 85 |
+
def exit_case():
|
| 86 |
+
return gr.update(value="<h2>Session ended. You exited the simulation.</h2>"), None, None
|
| 87 |
+
|
| 88 |
+
enter_btn.click(enter_case)
|
| 89 |
+
exit_btn.click(exit_case)
|
| 90 |
+
|
| 91 |
+
submit_btn.click(respond, inputs=[user_input, chatbot], outputs=[chatbot, chatbot])
|
| 92 |
+
|
| 93 |
+
# --- 6️⃣ Launch ---
|
| 94 |
+
if __name__ == "__main__":
|
| 95 |
+
demo.launch(share=True, enable_queue=True)
|