Skip to content

deploy-render-aiven-qdrant.md

Latest commit

 

History

History
144 lines (100 loc) · 3.76 KB

deploy-render-aiven-qdrant.md

File metadata and controls

144 lines (100 loc) · 3.76 KB

Free Cloud Deployment: Render + Aiven + Qdrant Cloud

This guide migrates the backend away from Railway while keeping the Vercel frontend.

Target Architecture

  • Frontend: Vercel
  • Backend: Render Web Service, Docker build from backend/Dockerfile
  • MySQL: Aiven MySQL Free
  • Redis/Valkey: Aiven Valkey Free or Upstash Redis
  • Vector store: Qdrant Cloud Free

1. Create Aiven MySQL

Create a MySQL service on Aiven, then copy:

  • Host
  • Port
  • Database name
  • User
  • Password

Render needs these environment variables:

SPRING_DATASOURCE_URL=jdbc:mysql://<AIVEN_MYSQL_HOST>:<PORT>/<DATABASE>?sslMode=REQUIRED&allowPublicKeyRetrieval=true&useUnicode=true&characterEncoding=UTF-8&serverTimezone=Asia/Shanghai
SPRING_DATASOURCE_USERNAME=<AIVEN_MYSQL_USER>
SPRING_DATASOURCE_PASSWORD=<AIVEN_MYSQL_PASSWORD>

For a fresh database, initialize schema with:

db/ai_learning_assistant.sql
db/migrations/2026-04-30_create_ai_config_table.sql

If data is migrated from an old Railway database, import the dump first, then run only missing migration files. At minimum, make sure ai_config exists.

2. Create Redis/Valkey

For Aiven Valkey or Upstash Redis, copy host, port, and password.

Render variables:

SPRING_REDIS_HOST=<REDIS_HOST>
SPRING_REDIS_PORT=<REDIS_TLS_PORT>
SPRING_REDIS_PASSWORD=<REDIS_PASSWORD>
SPRING_REDIS_DATABASE=0
SPRING_REDIS_SSL_ENABLED=true

If the provider gives a non-TLS port, set SPRING_REDIS_SSL_ENABLED=false.

3. Create Qdrant Cloud

Create a free Qdrant cluster, then copy:

  • Cluster URL
  • API Key

Render variables:

APP_QDRANT_ENABLED=true
APP_QDRANT_BASE_URL=https://<QDRANT_CLUSTER_HOST>
APP_QDRANT_API_KEY=<QDRANT_API_KEY>
APP_QDRANT_COLLECTION_NAME=material_segment_vectors

The backend creates the collection lazily when embedding/retrieval runs.

4. Deploy Backend On Render

Recommended manual setup:

Service Type: Web Service
Repository: Nyx-Amanises/AI-Intelligent-Learning-Assistant-System
Runtime: Docker
Root Directory: backend
Dockerfile Path: Dockerfile
Plan: Free

Or use the repository render.yaml Blueprint.

Required Render variables:

APP_CORS_ALLOWED_ORIGINS=https://ai-intelligent-learning-assistant-s.vercel.app
APP_FILE_UPLOAD_DIR=/tmp/ai-learning-uploads
APP_AI_CONFIG_FILE=/tmp/ai-config.json
APP_AI_ENABLED=true
APP_AI_MOCK_MODE=true
APP_JWT_SECRET=<random-long-secret>

SPRING_DATASOURCE_URL=<see section 1>
SPRING_DATASOURCE_USERNAME=<see section 1>
SPRING_DATASOURCE_PASSWORD=<see section 1>

SPRING_REDIS_HOST=<see section 2>
SPRING_REDIS_PORT=<see section 2>
SPRING_REDIS_PASSWORD=<see section 2>
SPRING_REDIS_SSL_ENABLED=true

APP_QDRANT_ENABLED=true
APP_QDRANT_BASE_URL=<see section 3>
APP_QDRANT_API_KEY=<see section 3>
APP_QDRANT_COLLECTION_NAME=material_segment_vectors

AI keys do not need to be set in Render if the admin user configures the shared key in the app.

5. Update Vercel

After Render deploys successfully, copy the Render backend URL, then set this in Vercel:

VITE_API_BASE_URL=https://<RENDER_BACKEND_DOMAIN>/api

Redeploy Vercel after changing the variable.

6. Post-deploy Checks

  1. Open the backend health smoke check:
https://<RENDER_BACKEND_DOMAIN>/api/ai/config

Unauthenticated requests should return an auth error, which still proves the backend is reachable.

  1. Open the frontend, log in as admin.
  2. Go to AI config, switch to shared config, save the shared API key.
  3. Upload a small document and confirm parse -> embedding -> RAG retrieval.

Notes

Render free web services sleep after inactivity. First request after sleep can be slow.

Render free web service storage is not persistent. This project should treat uploads as temporary input files and rely on parsed database records plus Qdrant vectors for long-lived data.