CoolFace
Apppublic

parupati/iran-us-aviation-rag

sourceHugging Faceupdated 7mo agoView on Hugging Face
0likes
App README

Hugging Face Spaces — Deployment Guide

This folder contains everything needed to deploy the Aviation Disruption RAG API as a Docker-based Hugging Face Space.

Live Space: https://huggingface.co/spaces/parupati/iran-us-aviation-rag

API Docs: https://parupati-iran-us-aviation-rag.hf.space/docs

Prerequisites

  • —A Hugging Face account
  • —huggingface_hub Python package (pip install huggingface_hub)
  • —Logged in via huggingface-cli login
  • —An OpenAI API key

Files

hf-space/
├── api.py              # FastAPI app with /query, /portfolio-chat, /health
├── Dockerfile          # Docker build config
├── requirements.txt    # Python dependencies
├── portfolio_info.md   # Portfolio data for the chat endpoint
├── src/
│   ├── ingest.py       # CSV ingestion + vector store builder
│   └── rag.py          # RAG retrieval + LLM chain
└── data/               # 6 CSV dataset files

Initial Setup (One-Time)

1. Create the Space

python
from huggingface_hub import HfApi

api = HfApi()
api.create_repo(
    "iran-us-aviation-rag",
    repo_type="space",
    space_sdk="docker",
    exist_ok=True,
)

2. Upload Files

python
from huggingface_hub import HfApi

api = HfApi()
api.upload_folder(
    folder_path="hf-space",
    repo_id="parupati/iran-us-aviation-rag",
    repo_type="space",
    ignore_patterns=[".git*"],
)

3. Set the OpenAI API Key as a Secret

python
from huggingface_hub import HfApi

api = HfApi()
api.add_space_secret(
    repo_id="parupati/iran-us-aviation-rag",
    key="OPENAI_API_KEY",
    value="sk-your-openai-key-here",
)

Or set it manually at: Settings → Repository Secrets → New Secret

Deploying Updates

Upload changed files

python
from huggingface_hub import HfApi

api = HfApi()
api.upload_folder(
    folder_path="hf-space",
    repo_id="parupati/iran-us-aviation-rag",
    repo_type="space",
    ignore_patterns=[".git*"],
)

Restart the Space (reuses cached Docker image)

Use when you only changed a Space secret or want a quick restart. Does NOT pick up code/file changes.

python
from huggingface_hub import HfApi

api = HfApi()
api.restart_space("parupati/iran-us-aviation-rag")

Factory Rebuild (rebuilds Docker image from scratch)

Use when you changed Dockerfile, requirements.txt, api.py, source files, or data files.

python
from huggingface_hub import HfApi

api = HfApi()
api.restart_space("parupati/iran-us-aviation-rag", factory_reboot=True)

Building the Docker Image Locally

bash
cd hf-space

# Build
docker build -t aviation-rag .

# Run (replace with your actual key)
docker run -p 7860:7860 -e OPENAI_API_KEY=sk-your-key aviation-rag

# Test
curl http://localhost:7860/health

Monitoring

  • —Build logs: https://huggingface.co/spaces/parupati/iran-us-aviation-rag/logs/build
  • —Runtime logs: https://huggingface.co/spaces/parupati/iran-us-aviation-rag/logs/container
  • —Space settings: https://huggingface.co/spaces/parupati/iran-us-aviation-rag/settings

Common Operations Reference

TaskCommand
Upload filesapi.upload_folder(folder_path="hf-space", repo_id="...", repo_type="space")
Set a secretapi.add_space_secret(repo_id="...", key="...", value="...")
Simple restartapi.restart_space("...")
Full rebuildapi.restart_space("...", factory_reboot=True)
Check statusapi.get_space_runtime("...")
Pause Spaceapi.pause_space("...")
Resume Spaceapi.resume_space("...")

Notes

  • —Free HF Spaces sleep after ~48 hours of inactivity. First request after sleep takes 30-60 seconds (cold start).
  • —The vector store is built during docker build (in the Dockerfile's RUN python src/ingest.py), so it's baked into the image.
  • —portfolio_info.md is loaded at startup and kept in memory for the /portfolio-chat endpoint.
  • —CORS is set to allow_origins=["*"] to allow calls from any frontend domain.