app.py

# Import necessary libraries
import os
import pandas as pd
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np
import gradio as gr
from groq import Groq

# Set up Groq API
GROQ_API = "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"  # Replace with your Groq API key
client = Groq(api_key=GROQ_API)

# # Load environmental dataset (upload to Colab)
# from google.colab import files
# uploaded = files.upload()  # Upload 'environmental_impact_assessment_dataset.csv'

# Load the dataset
df = pd.read_csv('/content/environmental_impact_assessment_dataset.csv')  # Replace with the uploaded file name

# Combine relevant text columns for embeddings
text_column = df['Project Type'] + ' ' + df['Mitigation Measures']  # Adjust based on your dataset columns

# Use SentenceTransformers to generate text embeddings
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')  # Lightweight embedding model
embeddings = embedding_model.encode(text_column.tolist())

# Convert embeddings to numpy array
embeddings_np = np.array(embeddings).astype(np.float32)

# Build FAISS index for document retrieval
index = faiss.IndexFlatL2(embeddings_np.shape[1])  # L2 distance for similarity search
index.add(embeddings_np)  # Add the document embeddings to the FAISS index

# Function to retrieve the most relevant document from the dataset
def retrieve_relevant_document(query):
    # Generate query embedding
    query_embedding = embedding_model.encode([query])
    query_embedding_np = np.array(query_embedding).astype(np.float32)

    # Perform similarity search in FAISS
    _, indices = index.search(query_embedding_np, k=1)  # Top 1 match
    retrieved_text = text_column.iloc[indices[0][0]]  # Retrieve corresponding text

    return retrieved_text

# Function to generate an EIA report using Groq's API
def generate_report(user_input):
    # Check if input is empty
    if not user_input.strip():
        return "Please provide project details to generate the Environmental Impact Assessment report."

    # Retrieve relevant information using FAISS
    relevant_document = retrieve_relevant_document(user_input)

    # Use Groq API to generate a report based on the retrieved document
    chat_completion = client.chat.completions.create(
        messages=[
            {"role": "user", 
             "content": f"Generate an environmental impact assessment report based on the following details:\n\n{relevant_document}\n\nUser Query: {user_input}"}
        ],
        model="llama3-8b-8192",  # Groq model
    )

    # Return the Groq-generated content
    return chat_completion.choices[0].message.content

# Gradio interface for user interaction
def gradio_interface(project_details):
    return generate_report(project_details)

# Launch Gradio app
iface = gr.Interface(
    fn=gradio_interface,
    inputs="text",  # Input: text box for project details
    outputs="text",  # Output: text box for the generated report
    live=False  # Set to False for non-live mode
)

iface.launch()