deployment-taskflow / README_SPACES.md
Tripagra
Add Hugging Face Spaces deployment configuration
a8772c4
|
Raw
History Blame Contribute Delete
1.27 kB
metadata
title: Smart Escalation API
emoji: 🤖
colorFrom: blue
colorTo: green
sdk: docker
pinned: false
license: mit

Smart Escalation API

AI-powered L1 customer support system with intelligent escalation.

Features

  • RAG-based retrieval using sentence-transformers and FAISS
  • Intelligent escalation to human agents
  • Google Gemini LLM integration
  • FastAPI backend with automatic documentation

API Endpoints

  • GET / - API information
  • GET /health - Health check
  • POST /ask - Submit customer questions

Environment Variables

Configure these in your Space settings:

  • GOOGLE_API_KEY - Your Google Gemini API key (required)
  • ARTICLES_DIR - Path to help articles (default: data/articles)
  • EMBEDDING_MODEL - Sentence transformer model (default: all-MiniLM-L6-v2)
  • RELEVANCE_THRESHOLD - Minimum similarity score (default: 0.5)
  • TOP_K_CHUNKS - Number of chunks to retrieve (default: 3)
  • LLM_MODEL - Gemini model name (default: gemini-1.5-flash)
  • CORS_ORIGINS - Allowed CORS origins

Usage

curl -X POST https://YOUR_USERNAME-escalation-api.hf.space/ask \
  -H "Content-Type: application/json" \
  -d '{"question": "How do I reset my password?"}'

Documentation

Visit /docs for interactive API documentation.