storm-server

What is this project?

Context

This project is a server based implementation of storm which provides a way to perform web searches to simulate the process of creating a paper of sorts that can be used in academia or on Wikipedia.

Technologies

FastAPI - Server framework of choice, very simple to set up and getting systems set up.

Helm - Templating system to help create kubernetes manfiests

Docker - Technology that creates containers that can be deployed to servers or ran within docker-compose to create a isolated network environment.

DevContainers - Allows a docker image to be loaded, acting as an ethermal environment.

Running

Docker Compose

To run using Docker compose, make sure both Docker and Docker Compose are both installed and properly running.

Create an env using this structure if using the OpenAI + you.com route:

OPENAI_API_KEY=<OPENAI_KEY>
YDC_API_KEY=<YOUCOM_KEY>
MAX_THREAD_NUM=3
MAX_CONV_TURN=3
MAX_PERSPECTIVE=3
SEARCH_TOP_K=3
OUTPUT_DIR=/opt/results
PDF_STORAGE=/opt/pdf
SUPABASE_URL=<SUPABASE_URL>
SUPABASE_KEY=<SUPABASE_SUBSCRIPTION_KEY>
SUPABASE_BUCKET=<SUPABASE_PUBLIC_BUCKET>

Create an env using this structure if using the Anthropic + Serper route:

OUTPUT_DIR=/opt/results
PDF_STORAGE=/opt/pdf
SUPABASE_URL=<SUPABASE_URL>
SUPABASE_KEY=<SUPABASE_SUBSCRIPTION_KEY>
SUPABASE_BUCKET=<SUPABASE_PUBLIC_BUCKET>
ANTHROPIC_API_KEY=<ANTHROPIC_API_KEY>
ANTHROPIC_SIMULATOR_LM=claude-3-haiku-20240307
ANTHROPIC_ASKER_LM=claude-3-sonnet-20240229
ANTHROPIC_GEN_LM=claude-3-opus-20240229
ANTHROPIC_ARTICLE_GEN_LM=claude-3-opus-20240229
ANTHROPIC_POLISH_LM=claude-3-opus-20240229
SERPER_API_KEY=<SERPER_API_KEY>

Run the following command:

docker compose -f dev_docker_compose.yaml up --build

This will spin up the server, and will listen to your localhost instance on port 8000.

Navigate to localhost:8000/docs to see the routes.

Run /create-wiki-article to create an article based off a topic, currently this uses chatgpt 3.5-turbo but will be updated in the future.

Run /create-pdf using the content portion of the previous api call to create a pdf that will be updated to supabase storage.

Credits

This project wouldn't be possible without the storm team who created and maintain the python package.

K8s

Using Visual Studio Code, open up the project. Ensure that docker is installed.

A pop up should pop up letting you know that the folder containers a .devcontainer, select to open up Visual Studio Code using the DevContainer. Let it extract the docker image.

Once it has been set up, click on the bash terminal to open up an ohmyzash instance. From here, run the following command:

make create-cluster

Let it finish setting up the cluster, and importing the images + helm chart. Once done, you can run kubectl get pods to see the api server running successfully.

Citations

@inproceedings{shao2024assisting,
      title={{Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models}}, 
      author={Yijia Shao and Yucheng Jiang and Theodore A. Kanell and Peter Xu and Omar Khattab and Monica S. Lam},
      year={2024},
      booktitle={Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers)}
}