docs: add docker runtime preview and ghcr image instructions

Author: ppiova

04-compose/README.md

@@ -51,6 +51,22 @@ Example response:
 { "model": "ai/gemma3", "reply": "Use a small base image and a multi-stage build to keep the final image lean." }
 ```
 
+## Pre-built image
+
+[![ghcr.io](https://img.shields.io/badge/ghcr.io-compose--api-2496ED?logo=docker&logoColor=white)](https://github.com/ppiova/docker-model-runner-lab/pkgs/container/docker-model-runner-lab%2Fcompose-api)
+
+A ready-to-run image is published to GitHub Container Registry by the
+[publish workflow](../.github/workflows/publish.yml). Run it without cloning the repo:
+
+```bash
+docker run --rm -p 8080:8080 \
+  -e OPENAI_BASE_URL=http://model-runner.docker.internal/engines/v1 \
+  -e MODEL=ai/gemma3 \
+  ghcr.io/ppiova/docker-model-runner-lab/compose-api:latest
+```
+
+Docker Model Runner must be enabled so the container can reach `model-runner.docker.internal`.
+
 ## Stop
 
 ```bash

05-blazor-chat/README.md

@@ -43,6 +43,23 @@ docker compose up --build
 Compose injects `OPENAI_BASE_URL` and `MODEL` into the container, so the app talks to the model
 at `http://model-runner.docker.internal` with no extra configuration.
 
+## Pre-built image
+
+[![ghcr.io](https://img.shields.io/badge/ghcr.io-blazor--chat-2496ED?logo=docker&logoColor=white)](https://github.com/ppiova/docker-model-runner-lab/pkgs/container/docker-model-runner-lab%2Fblazor-chat)
+
+A ready-to-run image is published to GitHub Container Registry by the
+[publish workflow](../.github/workflows/publish.yml). Run it without cloning the repo:
+
+```bash
+docker run --rm -p 8080:8080 \
+  -e OPENAI_BASE_URL=http://model-runner.docker.internal/engines/v1 \
+  -e MODEL=ai/gemma3 \
+  ghcr.io/ppiova/docker-model-runner-lab/blazor-chat:latest
+```
+
+Then open `http://localhost:8080`. Docker Model Runner must be enabled so the container can reach
+`model-runner.docker.internal`.
+
 ## Stop
 
 ```bash

README.md

@@ -64,6 +64,21 @@ All examples default to `ai/gemma3` and accept a `MODEL` environment variable to
 
 > Every module was tested end to end with .NET 10 and `ai/gemma3` on Docker Desktop, running on the llama.cpp backend.
 
+## Running in Docker
+
+The model and the app run as containers side by side. The model is provisioned by Docker Model
+Runner and the apps reach it at `model-runner.docker.internal`, all from a single `docker compose up`.
+
+![docker model status, list and compose ps showing the model and app running](assets/docker-runtime.svg)
+
+The web apps are also published as ready-to-run images on GitHub Container Registry, built by the
+[publish workflow](.github/workflows/publish.yml):
+
+| Image | Module | Pull |
+| ----- | ------ | ---- |
+| [![compose-api](https://img.shields.io/badge/ghcr.io-compose--api-2496ED?logo=docker&logoColor=white)](https://github.com/ppiova/docker-model-runner-lab/pkgs/container/docker-model-runner-lab%2Fcompose-api) | [04-compose](04-compose) | `docker pull ghcr.io/ppiova/docker-model-runner-lab/compose-api:latest` |
+| [![blazor-chat](https://img.shields.io/badge/ghcr.io-blazor--chat-2496ED?logo=docker&logoColor=white)](https://github.com/ppiova/docker-model-runner-lab/pkgs/container/docker-model-runner-lab%2Fblazor-chat) | [05-blazor-chat](05-blazor-chat) | `docker pull ghcr.io/ppiova/docker-model-runner-lab/blazor-chat:latest` |
+
 ## Endpoints
 
 | Where your code runs | Base URL |