3. Start the services¶
Step 3 of 5 — setup steps 2 and 3, The semantic layer and The sandbox. At the end of it, both services are running and the portal has checked each one.
The portal starts nothing itself. It gives you a compose file for each service, you run it, and you tell the portal the address. There are two services, and they share nothing:
| The semantic layer | The sandbox | |
|---|---|---|
| What it is | The model store, Cube and Cube Store | Where the agent runs Python |
| Holds | Your warehouse credentials | No credential at all |
| Address on this machine | http://localhost:8000 |
http://localhost:8600 |
The semantic layer¶
- Click Download docker-compose.yml. The file contains your warehouse credentials — keep it private, and do not commit it.
-
Put it in a folder of its own, and in that folder run:
docker compose up -d -
Type its address —
http://localhost:8000on the same machine — and click Check and continue.
The portal checks that something answers at the address, that it is the model store, and that it accepts the token in the file. A wrong address is refused with a reason; see Troubleshooting.
- 2The semantic layer
Cube, the model store that holds your data model, and Cube Store. Everything the agent answers comes through it.
⤓ Download docker-compose.ymlIt holds your warehouse password and this deployment's secrets.Save it anywhere, then in that directory — nothing else needs to be there:
docker compose up -dCheck and continue
Setup step 2, after Check and continue. A wrong address shows a red reason here instead.
Apple Silicon
If docker compose up fails with no matching manifest, the Cube Store line is for the
wrong architecture. The file's comment marks the line to change. See
The two deployments.
The sandbox¶
The same three steps, with its own file, in its own folder. Its address is
http://localhost:8600.
When both are checked, setup step 4 opens and the portal starts reading your warehouse's tables. This is the first time anything connects to the warehouse — a wrong password shows up here, not earlier.
What is inside each compose file, and why there are two, is in The two deployments.
Next: 4. Build the data model