🧠 Mini Embed Service

A lightweight, containerized embedding API that runs locally using sentence-transformers. Designed to power semantic search and retrieval-augmented generation (RAG) pipelines without relying on OpenAI or external APIs.

Docker Image:

greenygh0st/mini-embed

---

🚀 Features

• Runs locally via Flask + Docker
• Uses the nomic-ai/nomic-embed-text-v1 model (768-dim embeddings)
• Accepts text over HTTP and returns vector embeddings
• CPU-friendly — no GPU required
• Built-in request logging to STDOUT

---

📦 Requirements

• Docker
• Python (for local testing without Docker)

---

🔧 Quick Start

▶️ Build and run with Docker

docker build -t mini-embed .
docker run -d -p 5000:5000 --name mini-embed mini-embed

📡 Test the API

curl -X POST http://localhost:5000/embed \
     -H "Content-Type: application/json" \
     -d '{"text": "How do I reset my password?"}'

---

🔄 API

POST /embed

Request Body:

{
  "text": "Your input text goes here"
}

Response:

{
  "embedding": [0.123, -0.456, ...]
}

---

🔍 Use Cases

• Semantic search with PGVector
• Local RAG pipelines
• Embedding indexing for internal documents
• Offline language understanding

---

🛠️ Configuration

By default, the server listens on:

Host: 0.0.0.0
Port: 5000

Edit embed_server.py to change the model or port.

---

🧪 Development (without Docker)

pip install flask sentence-transformers
python embed_server.py

---

📝 License

MIT or similar — use freely, modify as needed.

---

✨ Inspired By

• sentence-transformers
• OpenAI Embeddings — but free and local