Create app.py

app.py

@@ -0,0 +1,104 @@
+import os
+import streamlit as st
+from huggingface_hub import login
+from datasets import load_dataset
+from llama_cpp import Llama
+from huggingface_hub import hf_hub_download
+import chromadb
+from sentence_transformers import SentenceTransformer
+
+# Load Hugging Face token
+hf_token = os.getenv("HF_TOKEN")  
+login(hf_token)
+
+# Load dataset
+dataset = load_dataset("Maryem2025/final_dataset")
+
+# Initialize Llama model
+llm = Llama(
+    model_path=hf_hub_download(
+        repo_id="TheBloke/CapybaraHermes-2.5-Mistral-7B-GGUF",
+        filename="capybarahermes-2.5-mistral-7b.Q2_K.gguf",
+    ),
+    n_ctx=2048,
+)
+
+# Initialize ChromaDB
+class VectorStore:
+    def __init__(self, collection_name):
+        self.embedding_model = SentenceTransformer('sentence-transformers/multi-qa-MiniLM-L6-cos-v1')
+        self.chroma_client = chromadb.Client()
+        if collection_name in self.chroma_client.list_collections():
+            self.chroma_client.delete_collection(collection_name)
+        self.collection = self.chroma_client.create_collection(name=collection_name)
+
+    def populate_vectors(self, dataset):
+        titles = dataset['train']['title'][:2000]
+        servings = dataset['train']['servings'][:2000]
+        total_times = dataset['train']['total_time'][:2000]
+        courses = dataset['train']['course'][:2000]
+        sections = dataset['train']['sections'][:2000]
+        instructions = dataset['train']['instructions'][:2000]
+        cuisines = dataset['train']['cuisine'][:2000]
+        calories = dataset['train']['calories'][:2000]
+
+        texts = [
+            f"Title: {title}. Servings: {serving}. Total Time: {total_time} minutes. "
+            f"Course: {course}. Sections: {section}. Instructions: {instruction}. "
+            f"Cuisine: {cuisine}. Calories: {calorie}."
+            for title, serving, total_time, course, section, instruction, cuisine, calorie 
+            in zip(titles, servings, total_times, courses, sections, instructions, cuisines, calories)
+        ]
+
+        for i, item in enumerate(texts):
+            embeddings = self.embedding_model.encode(item).tolist()
+            self.collection.add(embeddings=[embeddings], documents=[item], ids=[str(i)])
+
+    def search_context(self, query, n_results=1):
+        query_embedding = self.embedding_model.encode([query]).tolist()
+        results = self.collection.query(query_embeddings=query_embedding, n_results=n_results)
+        return results['documents']
+
+# Initialize and populate vector store
+vector_store = VectorStore("embedding_vector")
+vector_store.populate_vectors(dataset)
+
+# Define function for generating text
+def generate_text(message):
+    context_results = vector_store.search_context(message, n_results=1)
+    context = context_results[0] if context_results else ""
+
+    prompt_template = (
+        f"SYSTEM: You are a recipe generating bot.\n"
+        f"SYSTEM: {context}\n"
+        f"USER: {message}\n"
+        f"ASSISTANT:\n"
+    )
+
+    output = llm(
+        prompt_template,
+        temperature=0.3,
+        top_p=0.95,
+        top_k=40,
+        repeat_penalty=1.1,
+        max_tokens=600,
+    )
+
+    input_string = output['choices'][0]['text'].strip()
+    cleaned_text = input_string.strip("[]'").replace('\\n', '\n')
+    return cleaned_text
+
+# Streamlit UI
+st.title("JOSHI's Kitchen 🍽️")
+st.write("Generate recipes using AI powered by Hugging Face and ChromaDB!")
+
+user_input = st.text_area("Enter ingredients or ask for a recipe:", "")
+
+if st.button("Generate Recipe"):
+    if user_input:
+        with st.spinner("Generating recipe... 🍲"):
+            response = generate_text(user_input)
+        st.subheader("Generated Recipe:")
+        st.write(response)
+    else:
+        st.warning("Please enter a message.")