Halkali merkez neighborhood fatih st ozgur apt no 46 , Kucukcekmece , Istanbul , 34303 , TR
Deploy LiteLLM Proxy with PostgreSQL and Redis, unify providers behind one OpenAI-compatible API, and configure virtual keys, budgets, rate limits, TLS and log privacy.
Deploy LiteLLM Proxy with PostgreSQL and Redis, unify providers behind one OpenAI-compatible API, and configure virtual keys, budgets, rate limits, TLS and log privacy. Every command is presented as an operational step; replace example hostnames, passwords and secrets before execution.
Turkey (Turkish)
Germany (German)
Worldwide (English)
openssl rand -hex 32
openssl rand -base64 48LITELLM_MASTER_KEY=sk-YONETICI_ICIN_GUCLU_ANAHTAR
LITELLM_SALT_KEY=DEGISTIRILMEMESI_GEREKEN_RASTGELE_DEGER
POSTGRES_PASSWORD=GUCLU_DB_PAROLASI
DATABASE_URL=postgresql://litellm:GUCLU_DB_PAROLASI@postgres:5432/litellm
REDIS_HOST=redis
OPENAI_API_KEY=sk-...
GEMINI_API_KEY=...
ANTHROPIC_API_KEY=...model_list:
- model_name: gpt-4o-mini
litellm_params:
model: openai/gpt-4o-mini
api_key: os.environ/OPENAI_API_KEY
- model_name: gemini-flash
litellm_params:
model: gemini/gemini-2.5-flash
api_key: os.environ/GEMINI_API_KEY
- model_name: claude-sonnet
litellm_params:
model: anthropic/claude-sonnet-4-5
api_key: os.environ/ANTHROPIC_API_KEY
- model_name: yerel-qwen
litellm_params:
model: ollama/qwen3:8b
api_base: http://ollama:11434
general_settings:
master_key: os.environ/LITELLM_MASTER_KEY
database_url: os.environ/DATABASE_URLservices:
postgres:
image: postgres:16-alpine
restart: unless-stopped
environment:
POSTGRES_USER: litellm
POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
POSTGRES_DB: litellm
volumes:
- postgres_veri:/var/lib/postgresql/data
redis:
image: redis:7-alpine
restart: unless-stopped
volumes:
- redis_veri:/data
litellm:
image: ghcr.io/berriai/litellm-database:main-stable
restart: unless-stopped
env_file: .env
command: ["--config", "/app/config.yaml", "--port", "4000"]
ports:
- 127.0.0.1:4000:4000
volumes:
- ./config.yaml:/app/config.yaml:ro
depends_on:
- postgres
- redis
volumes:
postgres_veri:
redis_veri:curl -sS http://127.0.0.1:4000/health/liveliness
curl -sS http://127.0.0.1:4000/v1/models -H "Authorization: Bearer sk-YONETICI_ICIN_GUCLU_ANAHTAR"curl https://llm.example.com/v1/chat/completions \
-H "Authorization: Bearer sk-SANAL_ANAHTAR" \
-H "Content-Type: application/json" \
-d '{"model":"yerel-qwen","messages":[{"role":"user","content":"Merhaba"}]}'sudo docker compose exec -T postgres pg_dump -U litellm -d litellm -Fc > litellm-$(date +%F).dump
sha256sum litellm-$(date +%F).dumpsudo apt update
sudo apt install -y nginx certbot python3-certbot-nginx
sudo tee /etc/nginx/sites-available/llm.example.com > /dev/null <<'NGINX'
server {
listen 80;
listen [::]:80;
server_name llm.example.com;
client_max_body_size 10G;
proxy_read_timeout 3600;
proxy_send_timeout 3600;
location / {
proxy_pass http://127.0.0.1:4000;
proxy_http_version 1.1;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection "upgrade";
}
}
NGINX
sudo ln -s /etc/nginx/sites-available/llm.example.com /etc/nginx/sites-enabled/llm.example.com
sudo nginx -t
sudo systemctl reload nginxsudo certbot --nginx -d llm.example.com --redirect --agree-tos --no-eff-email -m [email protected]
sudo certbot renew --dry-run
curl -I https://llm.example.com
openssl s_client -connect llm.example.com:443 -servername llm.example.com </dev/null 2>/dev/null | openssl x509 -noout -subject -issuer -dates| Belirti / Symptom | Kök neden ve çözüm / Root cause and fix |
|---|---|
| Service or container does not start | Inspect compose configuration, dependency health, file permissions and the first fatal log line instead of repeatedly restarting. |
| Domain opens but HTTPS fails | Check A/AAAA records, ports 80/443, proxy mode, certificate challenge and conflicting reverse proxies. |
| Application cannot reach its database or queue | Use the internal service hostname, verify credentials and health checks, and keep database ports off the public interface. |
| Works initially but fails under load | Measure memory peaks, disk latency, connection pools and worker concurrency; increase capacity only after identifying the bottleneck. |
Yes after secrets are replaced, public access is restricted and a complete restore test has succeeded.
Minimums only prove the software can start. Size CPU, memory, IOPS and storage from representative workload measurements.
Yes. Use Full (strict) TLS and configure origin certificates, WebSockets and client IP forwarding where required.
Do not automatically apply major releases. Review migrations, back up data and retain the previous tested image.
Back up databases, uploaded files, persistent volumes, configuration, encryption keys and the deployed version together.
An NVMe VPS fits normal web workloads. Inference, accelerated OCR and video transcoding may require a compatible GPU.
Deploy LiteLLM Proxy with PostgreSQL and Redis, unify providers behind one OpenAI-compatible API, and configure virtual keys, budgets, rate limits, TLS and log privacy.