Find similar startups

This is a demo of startupradar.co's API combined with OpenAI embeddings to find similar startups.

The code works as follows:

1. load the list of startup domains in domains.txt
2. fetch descriptions of these startups with startupradar's API
3. create embeddings for all startups with OpenAI
4. compute cosine similarities between all pairs
5. output a similarity matrix as similarity_matrix.csv

The formatted output looks like this:

and the provided sample can be found in a public Google Sheet.

Installation and Usage

Install the dependencies into a virtual environment.

pip install -r requirements.txt

Create a config.py file and add the credentials for startupradar and OpenAI:

STARTUPRADAR_API_KEY = "your-key-here"

OPENAI_API_KEY = "your-key-here"
OPENAI_ENGINE = "text-similarity-davinci-001"

An API key for OpenAI can be created online. Please note that embedding a lot of startups can result in significant charges. Make sure to set budgets upfront!

Run with

python cli.py run