Spaces:
Sleeping
Sleeping
File size: 3,974 Bytes
5e93d48 11b9366 5e93d48 |
1 2 3 4 5 6 7 8 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 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 |
import gradio as gr
from langchain_community.document_loaders import TextLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_huggingface import HuggingFaceEmbeddings
from langchain.vectorstores import Chroma
from groq import Groq
import os
# === Config ===
GROQ_API_KEY = os.getenv("GROQ_API_KEY") # Use environment variable for Groq API Key
LLM_MODEL = "llama3-70b-8192"
FILE_PATH = "./Estonia.txt" # Use relative path for Hugging Face Space
DB_DIR = "chroma_db"
# === Load and Chunk Document ===
def load_and_split(filepath):
loader = TextLoader(filepath, encoding="utf-8")
docs = loader.load()
splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
chunks = splitter.split_documents(docs)
return chunks
# === Create or Load Vector Store ===
def get_vector_store(chunks):
embeddings = HuggingFaceEmbeddings(
model_name="all-MiniLM-L6-v2",
model_kwargs={"token": os.getenv("HF_TOKEN")} # Use environment variable for HuggingFace token
)
vectordb = Chroma.from_documents(documents=chunks, embedding=embeddings, persist_directory=DB_DIR)
return vectordb
def load_vector_store():
embeddings = HuggingFaceEmbeddings(
model_name="all-MiniLM-L6-v2",
model_kwargs={"token": os.getenv("HF_TOKEN")} # Use environment variable for HuggingFace token
)
vectordb = Chroma(persist_directory=DB_DIR, embedding_function=embeddings)
return vectordb
# === Query LLaMA via Groq ===
def query_llama(context, question):
client = Groq(api_key=GROQ_API_KEY)
prompt = f"""Use the following context to answer the question:
You're a cheerful, funky band member from Curly Strings 🎻🎤🎶. When fans ask you questions, respond with playful, short, and friendly answers — like you're chatting backstage after a show.
👇 Here’s how to guide fans:
- **🎵 Music**: If they ask for songs or albums, suggest a tune and drop **just one** of these links (rotate each time!):
- [Our Music Page](https://www.curlystrings.ee/music/)
- [Spotify](https://open.spotify.com/playlist/37i9dQZF1DZ06evO3XF2F2)
- [Apple Music](https://music.apple.com/us/artist/curly-strings/888454075)
- [YouTube](https://www.youtube.com/@CurlyStringsEstonia)
- **🎫 Tickets & Shows**: If it’s about concerts or dates, suggest one link from:
- [Tour Dates](https://www.curlystrings.ee/tour-dates/)
- [BandsInTown](https://www.bandsintown.com/a/6429648-curly-strings)
- **🛍️ Merch**: If they ask about merch, shirts, or goodies, share:
- [Our Merch Store](https://www.311.ee/curly-strings)
Important: Always suggest **only one** link at a time. Rotate the links so fans get different suggestions each time!
\n\n{context}\n\nQuestion: {question}\nAnswer:"""
response = client.chat.completions.create(
model=LLM_MODEL,
messages=[{"role": "user", "content": prompt}],
max_tokens=300
)
return response.choices[0].message.content.strip()
# === RAG Pipeline ===
def rag_pipeline(query):
if not os.path.exists(DB_DIR):
chunks = load_and_split(FILE_PATH)
vectordb = get_vector_store(chunks)
else:
vectordb = load_vector_store()
retriever = vectordb.as_retriever(search_kwargs={"k": 4})
docs = retriever.invoke(query)
context = "\n\n".join([doc.page_content for doc in docs])
return query_llama(context, query)
# === Gradio Interface ===
def chat_with_bot(question):
try:
answer = rag_pipeline(question)
return answer
except Exception as e:
return f"Oops! Something went wrong: {e}"
# Launch Gradio UI
iface = gr.Interface(
fn=chat_with_bot,
inputs=gr.Textbox(lines=2, placeholder="Ask me anything about Curly Strings 🎻🎤"),
outputs="text",
title="🎶 Curly Strings Chatbot 🎶",
description="Talk to a cheerful band member! Ask about music, shows, or merch."
)
iface.launch()
|