atlas-service

Search anything.

For a tutorial on how this service works, see the Deploy Whisper and Sentence Transformer Model to HuggingFace notebook or open it in a Colab:

Quickstart

TODO: Getting weird import errors when trying to import sibling directories.

pipenv run python src/app.py ModuleNotFoundError: No module named 'atlas'

Temporary workaround: use sls wsgi serve --config serverless.dev.yml.

cp serverless.yml serverless.dev.yml
Set the following environment variables:

  environment:
    HUGGING_FACE_URL: ""
    HUGGING_FACE_API_KEY: ""
    PINECONE_API_KEY: ""

sls wsgi serve --config serverless.dev.yml Make sure to put the credentials in serverless.dev.yml and not serverless.yml

pip install pipenv
pipenv install
pipenv run python src/app.py

If pipenv doesn't work, try prefixing all your commands with python -m pipenv

python -m pipenv install
python -m pipenv run python src/app.py

Test the service as a Lambda function locally

To simulate how the lambda sevice will run locally:

sls wsgi serve

Using Gitpod

To set the environment variables in gitpod, edit the following and paste into your terminal.

eval $(gp env -e HUGGING_FACE_URL=your_value)
eval $(gp env -e HUGGING_FACE_API_KEY=your_value)
eval $(gp env -e PINECONE_API_KEY=your_value)

Using the API

You can test as a GET request by putting setting the ?url=<url> and &q=<search term>

Example: https://5000-atilatech-atlasservice-9ogzr6pvij2.ws-us80.gitpod.io/search?url=https://youtu.be/lOI0Gs_t6cQ&q=neural%20network

Since your url might be different you can just copy-paste the following and append it to your API URL.

/search?q=neural%20network&url=https://youtu.be/lOI0Gs_t6cQ

To search all videos

/search?q=throw under the bus

You can also send POST requests.

curl --location --request POST 'http://127.0.0.1:5000/search' \
--header 'Content-Type: application/json' \
--data-raw '{
    "url": "https://www.youtube.com/watch?v=ciKdF97JWpU",
    "q": "throw under bus"
}'

Remove the URL property to search all videos

curl --location --request POST 'http://127.0.0.1:5000/search' \
--header 'Content-Type: application/json' \
--data-raw '{
    "url": "https://www.youtube.com/watch?v=ciKdF97JWpU",
    "q": "throw under bus"
}'

Deployment

The service is deployed as an AWS Lambda function
View Function logs in Cloudwatch

atilatech / atlas-service