Ollama and Open WebUI Compose Generator

Prepare a CPU-only local Compose configuration. The generator processes field values in your browser without submitting them; normal site analytics and font requests are separate. This tool generates text; it does not run a model, start containers, verify image availability or estimate performance.

Choose image tags or digests you have reviewed. The interface is bound to localhost. Keep the generated secret on your own machine, and back up the named volumes before updates.

Usage and supported inputs

Enter both image references, adjust numeric controls and select Generate configuration. Invalid input clears the previous result. Copy the text into compose.yaml; no download or deployment is performed.

Image inputs support a restricted repository plus tag or SHA-256 digest subset: lowercase repository characters, one explicit tag or digest, no registry ports and no combined tag-plus-digest. This filter does not verify full Docker reference grammar or registry availability.

Per-model parallel requests and maximum loaded models accept 1 through 16; keep-alive accepts 0 through 1440 minutes; the host port accepts 1 through 65535. These are generator bounds, not performance recommendations.

Before starting anything

Generate a local secret with openssl rand -hex 32 and set WEBUI_SECRET_KEY in your own environment. Never paste the secret here. Save the output as compose.yaml and run docker compose config --quiet locally. Validation checks configuration, not image existence, hardware fit or service readiness.

CPU-only is intentional. GPU passthrough depends on the host and driver stack; this generator does not promise GPU support. No host port is published for Ollama. The two services communicate through their Compose service names.

Sources and related reading

Ollama environment and per-model concurrency documentation; Docker Compose configuration reference; Open WebUI setup and secret guidance.

Read the local deployment guide or convert an existing Docker run command.