Build and Deploy an AI Semantic Search Engine with Python FastAPI and Postgres pgvector

Build a personal knowledge-base search engine where users paste paragraphs and search by meaning, not just exact words.

Target searches: deploy fastapi python app, postgres vector database hosting, host ai backend no serverless, fastapi alternatives render heroku.

The Stack

Python with FastAPI
Postgres with the pgvector extension
Gemini API for embeddings
Adios runtime with secrets and managed database wiring

Why This Feels Like a Superpower

AI backends often import heavier Python libraries and initialize API clients at startup. On Adios, your FastAPI app runs as a normal long-lived service, so your backend can stay warm instead of rebuilding state on every request.

1. Build the API

Ask the Adios AI agent:

Build a FastAPI semantic search API. Add endpoints to insert documents and
search them using embeddings. Use Gemini for embeddings, Postgres pgvector for
similarity search, and /healthz for health checks.

2. Add `adios.yaml`

name: semantic-search
region: de
replicas: 1

build_cmd: python -m pip install -r requirements.txt
start_cmd: python -m uvicorn app.main:app --host 0.0.0.0 --port $PORT

runtime:
  name: [email protected]
  port: 8000
  health_path: /healthz
  memory_mb: 1024

secrets:
  DATABASE_URL: secret://DATABASE_URL
  GEMINI_API_KEY: secret://GEMINI_API_KEY

requires:
  - vector-db

Create a Postgres pgvector resource named vector-db, or set DATABASE_URL manually if you already have one.

3. Deploy

adios secrets set GEMINI_API_KEY
adios up