README
¶
lilysaas
- what are the core database schema "things"
- what is the plan for taking payments?
- what are the key actions?
entities
- auth
- use keycloak
- model
- json file to start with
- a list of things that lilypad can run
- job
- inference
- training (fine tunings)
- starts with an existing model
- produces a new model
plan
This is to have a hello world that you can login to.
- docker compose stack
- keycloak
- frontend (react & vite)
- api (go)
- postgres
dev
You need the following installed:
You also need the lilypad repo cloned alongside this one.
smart contract interface
When the smart contract in lilypad changes - checkout latest main of lilypad and then:
./stack generate-golang-bindings
This will re-create the api/pkg/contract/Modicum.go
running alongside lilypad in local dev
Make sure that you have cloned lilypad at the same level as this repo.
export RUN_LILYPAD=1
./stack start
This will boot a fresh lilypad and lilysaas stack and link everything together using tmux.
To run each part manually here are the guides:
lilypad
First start lilypad and then bacalhau:
cd lilypad
./stack boot
Then we start bacalhau:
NOTE you will require the correct version of bacalhau - run through this guide:
./stack bacalhau-serve
Now we need 3 other terminals each with LOG_LEVEL=debug
./stack solver --server-url http://172.17.0.1:8080
so it reports an address accessible from inside docker (the default docker bridge ip - this will probably only work on linux? on mac maybe you can use host.docker.internal)
./stack mediator
./stack resource-provider
Now we switch to lilysaas and get it booted alongside lilypad.
lilysaas
We need to create an top level .env file like so
export WEB3_PRIVATE_KEY=XXX
You can get the WEB3_PRIVATE_KEY value with this command:
cat ../lilypad/.env | grep JOB_CREATOR_PRIVATE_KEY
Copy the value of this to be the WEB3_PRIVATE_KEY value in the .env file - the other values should be as shown.
docker-compose up -d
Then we exec into the api container and run the api server:
docker-compose exec api bash
go run . serve
port forwarding for local web development on a remote GPU machine
If you want to run the stack on a GPU machine and access it via browser on another machine, the following might be useful:
sudo ssh -L 80:localhost:80 <user>@<gpu-node>
and in another terminal:
ssh -L 8081:localhost:8081 <gpu-node>
Then open http://localhost in the browser
of course, VS code remote via SSH is useful in this scenario as well
Directories