jjleng/sensei: Yet another open source Perplexity

[ShellLM]

• jjleng/sensei: Yet another open source Perplexity

jjleng/sensei: Yet another open source Perplexity

Snippet

"Sensei Search is an AI-powered tool designed to deliver the relevant search results.

Demo

http://sensei-frontend.default.52.24.120.109.sslip.io/

Screenshots

[Screenshots]

Tech Stack

Sensei Search is built using the following technologies:

• Frontend: Next.js, Tailwind CSS
• Backend: FastAPI, OpenAI client
• LLMs: Mistral-7b and Command-R
• Search: SearxNG
• Memory: Redis
• Deployment: AWS, Paka

How to Run Sensei Search

You can run Sensei Search either locally on your machine or in the cloud.

Running Locally

Follow these steps to run Sensei Search locally:

Prepare the backend environment:

cd sensei_root_folder/backend/
mv .env.development.example .env.development
Edit .env.development as needed. The example environment assumes you run models through Ollama. Make sure you have reasonably good GPUs to run the command-r model.

No need to do anything for the frontend.

Run the app with the following command:

cd sensei_root_folder/
docker compose up

Open your browser and go to http://localhost:3000

Running in the Cloud

We deploy the app to AWS using paka. Please note that the models require GPU instances to run.

Before you start, make sure you have:

• An AWS account
• Requested GPU quota in your AWS account

The configuration for the cluster is located in the cluster.yaml file. You'll need to replace the HF_TOKEN value in cluster.yaml with your own Hugging Face token. This is necessary because the mistral-7b and command-r models require your account to have accepted their terms and conditions.

Follow these steps to run Sensei Search in the cloud:

Install paka:

pip install paka

Provision the cluster in AWS:

make provision-prod

Deploy the backend:

make deploy-backend

Deploy the frontend:

make deploy-frontend

Get the URL of the frontend:

paka function list

Open the URL in your browser."

Suggested labels

None

[ShellLM]

Related content

#761 similarity score: 0.86
#678 similarity score: 0.85
#136 similarity score: 0.85
#396 similarity score: 0.85
#774 similarity score: 0.84
#386 similarity score: 0.84