import gradio as gr
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_openai import ChatOpenAI
from langchain.chains import RetrievalQA
from langchain.vectorstores import FAISS


def recipe_generator(ingredients):
  """
  This function takes a list of ingredients as input and returns a recipe using the LangChain library and OpenAI API.
  """
  # Load the LangChain model and vector store
  embedd = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large")
  vector_store = FAISS.load_local("faiss_index", embedd,allow_dangerous_deserialization=True)

  # Create the LangChain chain
  llm = ChatOpenAI(
      openai_api_key="sk-proj-JjAAcDuAsxPkm3zg9Iz2T3BlbkFJ5txMWIx2TS6T24rPYhjN",
      model="gpt-3.5-turbo",
      temperature=0.3
  )
  qa_chain = RetrievalQA.from_chain_type(
      llm,
      retriever=vector_store.as_retriever(),
      return_source_documents=True
  )

  # Generate the recipe
  question = f"Make a recipe using the following ingredients: {ingredients}"
  result = qa_chain({"query": question})
  return result["result"]

# Create the Gradio interface
interface = gr.Interface(
    fn=recipe_generator,
    inputs=[
        gr.Textbox(
            show_label=False,
            placeholder="Enter your ingredients separated by commas"
        ),
    ],
    outputs=[
        "textbox",
    ],
    examples=[
        ["chicken, rice, vegetables"],
        ["pasta, tomato sauce, cheese"],
    ],
    title="Recipe Generator",
    description="Enter a list of ingredients and I will generate a recipe for you."
)

# Launch the Gradio app
interface.launch()