Show how to deploy Semantic Search as an API with FastAPI or Flask?

 

Deploy Semantic Search as an API with FastAPI

Let's build a Semantic Search API using FastAPI + FAISS + Sentence Transformers.

---

Install Required Libraries

pip install fastapi uvicorn sentence-transformers faiss-cpu

---

1. Folder Structure

semantic_search_api/
│
├─ main.py              # FastAPI app
├─ model.py             # Embedding + FAISS setup
└─ requirements.txt      # Dependencies

---

2. Model Setup (model.py)

Initialize Sentence Transformers and FAISS index.

from sentence_transformers import SentenceTransformer
import faiss
import numpy as np

# Load Sentence Transformer Model
model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')

# Sample Documents
documents = [
    "I love playing football.",
    "Artificial Intelligence is the future.",
    "Machine learning powers AI.",
    "The weather is sunny today.",
    "Deep learning improves neural networks.",
    "It's raining outside."
]

# Generate Embeddings
document_embeddings = model.encode(documents)

# Normalize embeddings for cosine similarity
faiss.normalize_L2(document_embeddings)

# Create FAISS Index
embedding_dimension = document_embeddings.shape[1]
index = faiss.IndexFlatIP(embedding_dimension)  # Inner Product (Cosine Similarity)
index.add(document_embeddings)

---

3. API Setup (main.py)

Create the FastAPI app.

from fastapi import FastAPI
from pydantic import BaseModel
from model import model, index, documents
import faiss
import numpy as np

app = FastAPI()

class Query(BaseModel):
    text: str
    top_k: int = 3

@app.get("/")
def read_root():
    return {"message": "Semantic Search API is running 🚀"}

@app.post("/search")
def search(query: Query):
    # Generate Query Embedding
    query_embedding = model.encode([query.text])

    # Normalize for Cosine Similarity
    faiss.normalize_L2(query_embedding)

    # Search FAISS Index
    distances, indices = index.search(query_embedding, query.top_k)

    results = []
    for i in range(query.top_k):
        results.append({
            "document": documents[indices[0][i]],
            "distance": float(distances[0][i])
        })

    return {"query": query.text, "results": results}

---

4. Run the API

uvicorn main:app --reload

---

API Endpoints

Method	Endpoint	Description
GET	/	Health check
POST	/search	Perform Semantic Search

---

5. Test with Postman or Curl

Request:

POST /search
Content-Type: application/json

{
  "text": "Future of AI technology",
  "top_k": 3
}

Response:

{
  "query": "Future of AI technology",
  "results": [
    {
      "document": "Artificial Intelligence is the future.",
      "distance": 0.9823
    },
    {
      "document": "Machine learning powers AI.",
      "distance": 0.9745
    },
    {
      "document": "Deep learning improves neural networks.",
      "distance": 0.9634
    }
  ]
}

---

Bonus Tip:

If you need vector persistence between restarts, you can save and load FAISS indexes:

faiss.write_index(index, "faiss_index.bin")
index = faiss.read_index("faiss_index.bin")

---

Conclusion 🎯

FastAPI + FAISS + Sentence Transformers gives you a blazing-fast Semantic Search API.