Polyer

LiteLLM

Deploy LiteLLM on Polyer to run an OpenAI-compatible proxy in front of many LLM providers, with virtual keys and spend tracking. The template runs a public LiteLLM service with a private PostgreSQL 16 database and a private Redis cache, both on volumes. Review the provider API keys and the preconfigured model list.

AI applicationsCoolify open catalog

What you can use it for

  • Give your applications one OpenAI-format endpoint for models from OpenAI, Anthropic and other providers.
  • Issue a separate key to each team or application and track their spending in one place.

What this template creates

litellmghcr.io/berriai/litellm-database:main-stable
Public endpoint
postgrespostgres:16-alpine
Private service Persistent volume
redisredis:7-alpine
Private service Persistent volume

Configuration to review

These inputs are exposed by the template. The deployment form identifies required values and validates them before creating services.

InputEnvironment key
Anthropic api baseANTHROPIC_API_BASE
Anthropic api keyANTHROPIC_API_KEY
Openai api baseOPENAI_API_BASE
Openai api keyOPENAI_API_KEY
Voyage api baseVOYAGE_API_BASE
Voyage api keyVOYAGE_API_KEY

How to deploy

  1. Select Configure deployment and sign in or create an account.
  2. Choose a new or existing project and select an available region.
  3. Review services, required inputs, storage and compatibility warnings. Submit the deployment only when the configuration is ready.
  4. Inspect deployment logs, complete the application's setup and test its endpoints before production use.

Operating this application

  • Sign in to the admin UI with the generated UI username and password, keep the generated master key for administration, and give each application its own virtual key.
  • Enter only the provider keys you use and review the preconfigured model list, which may name models a provider has since retired. If you enable STORE_MODEL_IN_DB, set LITELLM_SALT_KEY before adding models in the UI and do not change it afterwards.
  • Back up the PostgreSQL database, which stores virtual keys, settings and spend data, and check release notes or pin a version before redeploying, since the main-stable tag moves. Provider inference charges are billed separately from hosting.

Hosting costs

CPU, memory, storage and network usage contribute to your bill. The selected region affects resource prices. A multi-service template incurs usage for each service; external APIs and software licenses are separate.

Review prices and estimate resource costs

Source and validation

This deployment configuration comes from Coolify's Apache-2.0 licensed open template collection and is adapted through Polyer's compose importer.

Static compatibility analysis is not a live application verification. Test your chosen configuration and recovery workflow before using it for production.

Editorial content reviewed: 2026-10-10

Upstream documentation Template source and license

Choose your next step

Related templates