ReviewSense

ReviewSense leverages retrieval-augmented generation with LoRA fine-tuning to extract actionable insights from large-scale product review datasets. Built with LangGraph for agentic workflows and Pinecone for vector search, it provides intelligent product analysis through category-specific sentiment classification.

Running the Script

• Note that the reviewsense_pipeline.py file takes less than 5 minutes to run, so there is no unit test file
• The Script contains several Unit Test Questions.
• GPU usage is recommended for faster inference. Cloud deployment is also recommended.
• For running the advanced RAG Agent with Lora Fine Tuned Model, use the reviewsense_frontend.py, where you can input questions and model at your choice, or if you want to compare them
• Just use [docker run -it reviewsense python reviewsense_frontend.py] command and the front end system will run

Data

• The data is provided as final_review_chunked_df.csv
• Download the data as in the name as final_review_chunked_df.csv
• If you change the name of the dataset, remember to change it as well in the dockerfile as well as in the code to avoid running into any issue.

API KEY

• The projects require you to have an OPENAI API KEY to access OpenAI's GPT-3.5-turbo API and a Pinecone API Key. Thus, it is recommended to create an (.env) file in your root directory.
• Add you API Key In this plain text form
• OPENAI_API_KEY=your_api_key_here and PINECONE_API_KEY=your_api_key_here
• Remember to build your .env in the folder that you transfer to docker to ensure no error.
• Use load_dotenv() to read in these key.

Dockerfile

• Remember that the data is included in the dockerfile as the line
• COPY final_review_chunked_df.csv /app/
• Note that there is a .env file in the dockerfile that is expected to have. You can remove that if you find it is not there
• If you change the name of the dataset, remember to change it as well in the dockerfile to avoid running into any issue.

Docker Instructions

Building the Docker Image

Build your Docker image:

docker build -t reviewsense .

Verify the Docker Image Was Built

Check if the image was successfully created by running:

docker images

This command will list all available images. Look for your image name in the output.

Run the Docker Image

To start a container from your image:

docker run -it reviewsense /bin/bash

This opens an interactive session inside the container.

If you want to run the whole code (reviewsense_pipeline.py)

docker run -it reviewsense

If you want to run the Front End (reviewsense_frontend.py)

docker run -it reviewsense python reviewsense_frontend.py

Delete Docker Image (When No Longer Needed)

To remove an image you no longer need:

1. Find the Image ID

   docker images

2. Delete the Image

   docker rmi image_id

   If the image is in use, you may need to force delete it:

   docker rmi -f image_id