Ask Edgar

Edgar is your AI investment assistant. Ask Edgar English-language questions about public company financial data.

For example:

What were the top 5 companies with the highest EBITDA?
Show me the R&D expenses of Apple over the last 8 years.
How much revenue did Tesla have in 2018?

Demo: https://www.loom.com/share/8e8735e38fb64e3ab6e53ddd9129cf1b

Setup

Requirements / Initial setup

Install miniconda: https://docs.conda.io/en/latest/miniconda.html
Set up the conda and pip environments.

conda env create -f tools/environment.yml

Enter the conda environment

conda activate edgar

To exit, run conda deactivate.

Set up environment

Load the following variables into your environment:

POSTGRES_ADDRESS
POSTGRES_PORT
POSTGRES_USERNAME
POSTGRES_PASSWORD
POSTGRES_DBNAME
FMP_KEY
OPENAI_KEY

Set up database

Run the following only if you are creating a new database instance! Otherwise, stop and move on to the next section.

Run the jupyter notebook to set up and inspect the database.

cd tools
jupyter notebook setup_db.ipynb

Then run populate_db.py to fill the postgres database with data.

python populate_db.py

Set up the AWS Elastic Beanstalk project

If the project does not yet exist, follow these instructions: https://docs.aws.amazon.com/elasticbeanstalk/latest/dg/create-deploy-python-flask.html

The following will require you to have Beanstalk CLI installed: https://github.com/aws/aws-elastic-beanstalk-cli-setup

cd server
eb init -p python-3.6 flask-edgar --region us-west-1
eb create flask-env
# Set up the environment with the same variables listed above.
eb setenv POSTGRES_ADDRESS=...

Run the code

Run the Python Flask server

Build the static React frontend. Also install npm dependencies.

cd client
npm install
npm run build
cd ..

Then enter the server directory and run the server locally:

cd server
python application.py

Run in React development mode

Run the client code in development mode.

From project root:

cd client
npm start

Test the API

Query the api endpoint for the given user query.

# Local
curl -X GET http://127.0.0.1:5000/api?query=${USER_QUERY}
# Deployed
curl -X GET http://flask-env.eba-dwnhvhak.us-west-1.elasticbeanstalk.com/api?query=${USER_QUERY}

Deploy

To deploy to production (S3 and AWS Elastic Beanstalk), run:

cd client
npm run build
aws s3 sync build/ s3://ask-edgar --acl public-read

cd ../server
eb deploy flask-env

Make sure to make the files public: https://s3.console.aws.amazon.com/s3/buckets/ask-edgar?region=us-west-1&tab=objects

TODO: Deploy the web app correctly at some point.

asta-li/edgar