NBuild.io NIPs Chatbot RAG Model

NBuild.io chatbot is using RAG-based question answering over NIPs docs. DeepSeek-V3.

Install Python and Tools

Since many Python versions may be installed, you must check that Python 3.13, pip and pipenv are installed correctly.

python3 --version
python3 -m ensurepip
python3 -m pip --version
python3 -m pip install pipenv
python3 -m pipenv --version

Formatter: Google Yapf

Before you start coding, you must install Yapf to active code formatting.

Visual Studio Code: Google Yapf Extension

python3 -m pip install yapf

You should check which Python installation is active in Visual Studio Code or Cursor.

Official Nostr Sources used by RAG

• GitHub nostr-protocol/nips

More to mention soon.

You can open a ticket here if you want to include specific data in the RAG-based chatbot.

Start RAG-based Chatbot API locally

python3 -m pip install pipenv
pipenv shell
REPLICATE_API_TOKEN="<token from replicate.com account>" TOKENIZERS_PARALLELISM=false uvicorn nips_chat_api.chat_api:app --app-dir="src" --reload

Using CURL to get a Q answered

curl -X "POST" "http://127.0.0.1:8000/predict" \
     -H 'Content-Type: application/json; charset=utf-8' \
     -d $'{
  "user_q": "Describe what Nostr is?"
}'

Start RAG-based Chatbot CLI locally

You could also chat with NBuild.io Chatbot via a terminal. We prefer this when improving the RAG-based chatbot and its Nostr and NIPs knowledge.

REPLICATE_API_TOKEN="<token from replicate.com account>" TOKENIZERS_PARALLELISM=false python3 src/nips_chat/chat_cli.py

Build and Run Docker Image

You should install Docker Desktop on Mac and Google Cloud CLI before you proceed.

You can deploy our NBuild.io RAG-based Model to other Docker platforms as well. But we recommend to start with Google Cloud Run because of the low operating costs and best reliability.

gcloud login
gcloud config set project [id]
gcloud services enable run.googleapis.com

Build Docker Image on Google Cloud

docker build -t gcr.io/[id]/nbuild-nips-chat-model .
docker push gcr.io/[id]/nbuild-nips-chat-model

Deploy to Cloud Run on Google Cloud

gcloud run deploy nbuild-nips-chat-model \
    --image gcr.io/[id]/nbuild-nips-chat-model \
    --platform managed \
    --region us-central1 \
    --port 8080 \
    --cpu 1 \
    --memory 4Gi \
    --min-instances 1 \
    --max-instances 2 \
    --set-env-vars REPLICATE_API_TOKEN="<token from replicate.com account>",TOKENIZERS_PARALLELISM=false \
    --allow-unauthenticated