README.md

metadata

title: Gemischtes Hack Embeddings
emoji: 🎙️
colorFrom: gray
colorTo: yellow
sdk: docker
app_port: 7860
pinned: false

Embedding Server for Gemischtes Hack

FastAPI server hosting intfloat/multilingual-e5-small embeddings model on Hugging Face Spaces.

Setup

1. Create a new HF Space: https://huggingface.co/new-space

   • Name: gemischtes-hack-embed
   • License: MIT
   • SDK: Docker

2. Clone this Space to your machine (or manually upload files)

3. The Docker container will:

   • Install dependencies from requirements.txt
   • Load the multilingual-e5-small model
   • Expose FastAPI on port 7860

4. Once deployed, the Space URL will be available at: https://{your-username}-gemischtes-hack-embed.hf.space

API

POST /embed

Generate embeddings for text.

curl -X POST https://{your-username}-gemischtes-hack-embed.hf.space/embed \
  -H "Content-Type: application/json" \
  -d '{"text": "Was ist Gemischtes Hack?"}'

Response:

{
  "embedding": [0.123, -0.456, ..., 0.789]  // 384-dim vector
}

GET /health

Check server status.

GET /

View API info.

Notes

• First request takes ~10-30 seconds (model loading + HF Spaces cold start)
• Subsequent requests take ~500ms
• Space auto-sleeps after 48 hours of inactivity
• Max 2 vCPU / 16 GB RAM (free tier)

Integration

Update web/src/lib/embed.ts:

const HF_SPACE_URL = "https://{your-username}-gemischtes-hack-embed.hf.space";

async function embedLocal(text: string): Promise<number[]> {
  const response = await fetch(`${HF_SPACE_URL}/embed`, {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({ text }),
  });

  if (!response.ok) {
    throw new Error(`Embed error: ${response.status}`);
  }

  const data = await response.json();
  return data.embedding;
}