Workshop

11 modules · about 4 hours · every command says what it does, why, and what you should see.

You need Git and Docker with Compose v2.

Terminal

git clone https://github.com/jars-demo/local-ai-demo.git
cd local-ai-demo
cp .env.example .env
./scripts/pull-model.sh
docker compose up -d --build --wait

Part 1 · Core: run and understand the stack

For everyone. Containers, ports, volumes, HTTP, SQL, Docker DNS and a local model.

01Architecture and PrerequisitesThe four services, what "offline" really means, and getting your machine ready20m
02Images, Containers, Ports and VolumesStart the stack and see what each Docker building block does25m
03Frontend to Backend over HTTPFollow one click from the browser through nginx to FastAPI and back20m
04Backend to PostgreSQLConnection pools, migrations, parameterised SQL and full-text retrieval25m
05Docker DNS and Service NetworkingWhy service names work, why localhost does not, and the internal network20m
06Running and Querying the Local LLMTalk to the model directly, then through the app; retrieval before inference25m
07The Inference API and Model StorageOllama endpoints, timings, manifests, blobs and checksums on disk20m
08Multi-stage Builds and Image OptimizationFrom a 5.6 GB Ollama image to 142 MB, and why each stage exists25m

Part 2 · Advanced: ship it to an offline server

For operators. Bundle every artifact, deploy without internet, prove it, update it.