Watsonx Sustainability Advisor App

The Sustainability Reporting Manager in IBM Envizi spends a lot of time in preparing the responses to the various questions from the disclosure reports.

We leveraged the generative AI capability of watsonx.ai to create the Watsonx Sustainability Advisor, a tool that serves as a dynamic guide to respond to any given reporting framework question.

Watson-x Sustainability Advisor can help the sustainability reporting manager to get the responses to the various questions very quickly.

It leverages watsonx Assistant to get the questions from the sustainability reporting manager.

For the given question, the Watson Assistant send this question to Watson Discovery, which does a semantic search by scanning through various sustainability reports and retrieves the relevant document passages back to the Watson Assistant.

Then watson Assistant sends the question along with the document passages that are retrieved by the Watson Discovery as an instruction to Watsonx.ai

Watsonx.ai reads the prompt and uses the LLM model to summarize the answer for the question back to the watson Assistant.

The watson Assistant shows the summary response to the user as an answer to their question.

1. Source code details

1.1 Watson Assistant UI and API

The app folder of this repo contains the python app code, which uses Watson Discovery, and Watsonx.ai to respond to the queries from the UI.

The web-wa folder contains the ReactJS UI code and it opens up the Watson-Assistant in the screen. Then the Watson-Assistant would calls Watson Discovery, and Watsonx.ai to respond to the queries of the user.

Solution Architecture

This solution follows RAG architecture model. The tech stacks used here are the following

• Watsonx.ai
• Watson Assistant
• Watson Discovery
• LangChain

1.2 React UI and API

There is the React UI based version of Watsonx Sustainability Advisor is also available.

The web folder contains the ReactJS UI code, which calls the above python app to query llm.

Solution Architecture

This solution follows RAG architecture model. The tech stacks used here are the following

• Watsonx.ai
• Watson Discovery
• LangChain

2.Installation

2.1. Install python 3.9 version

1. Run the below command to install python 3.9 version.

brew install [email protected]

2. Run the below command to ensure python command uses homebrew’s version of python3

echo "alias python3=/opt/homebrew/bin/python3.9" >> ~/.zshrc
echo "alias python=/opt/homebrew/bin/python3.9" >> ~/.zshrc

2.2. Install dependent libraries

1. Run the below commands to install the dependent libraries

python -m pip install python-dotenv
python -m pip install ibm-generative-ai
python -m pip install "ibm-generative-ai[langchain]"
python -m pip install pypdf
python -m pip install InstructorEmbedding
python -m pip install 'transformers[torch]'
python -m pip install sentence-transformers
python -m pip install Flask flask-restful flask_httpauth
python -m pip install cachetools
python -m pip install unstructured
python -m pip install chromadb
python -m pip install faiss-cpu

python -m pip install flask_cors
python -m pip install ibm-cos-sdk
python -m pip install rogue
python -m pip install ibm-watson

python -m pip install libffi-dev

2.3 Create .env file for the app

Create the .env file in the root folder of the app with the below content.

LOGLEVEL=DEBUG

### Watsonx.ai
GENAI_KEY=<REPLACE IT WITH YOUR IBM GENAI APIKEY>
GENAI_API=https://us-south.ml.cloud.ibm.com/ml/v1-beta/generation/text?version=2023-05-29
PROJECT_ID="2677a1e4-xxxxxxxx"

### Watson Discovery
WD_API_KEY=<REPLACE IT WITH YOUR WATSON DESCOVERY API KEY>
WD_SERVICE_URL=https://api.us-south.discovery.watson.cloud.ibm.com/instances/a83d0a4b-a6d9-41db-972a-5b70d5fc2c91
WD_PROJECT_ID=80c28471-xxxxxxx

### Dataset
DATA_PATH=./datasets

3. Starting the App

1. Goto the root folder of this app.

2. Run the below command to start the app in virtual environment called venv .

python -m venv myvenv
source myvenv/bin/activate
python app/main.py

4. To deactivate the virtual environment

1. Goto the root folder of this app.

2. Run the below command to deactivate the virtual environment.

deactivate