turbo remote cache

A remote cache server backed by S3.

This server is designed to run within a CI environment where it already has access to a S3 bucket. So no authentication yet.

server

reference: https://github.com/vercel/turbo/blob/main/cli/internal/client/client.go

Example request:

https://api.vercel.com/v8/artifacts/09b4848294e347d8?teamID=team_lMDgmODIeVfSbCQNQPDkX8cF

authentication over bearer token.

configuration

• AWS access is either assumed or exists already. it will perform s3:GetObject and s3:PutObject actions.
• configure s3 bucket name via environment variable STORAGE_PATH:

  STORAGE_PATH=my-turbo-cache-storage
  AWS_REGION=us-east-1

Run yarn build then yarn start.

If you want to change the TCP port, update bin/start_server.ts.

data model

the cache will be saved in S3:

• bucket name is configured via environment variable STORAGE_PATH.
• the object key is {teamSlug}/{hash}.

upload cache

PUT /v8/artifacts/:hash?slug=turbo-team
Authorization: Bearer {{token}}
header: x-artifact-client-ci

response:

{ok: true}

download cache

GET /v8/artifacts/:hash?slug=turbo-team
Authorization: Bearer {{token}}

response:

200
Content-Length: n
Content-Type: application/octet-stream
x-artifact-tag: xxx

{ok: true}

error response:

404

{ok: false}

RecordAnalyticsEvents

POST /v8/artifacts/events
Authorization: Bearer {{token}}

see details here

client

Setup environment variable:

TURBO_API=http://localhost:41300
TURBO_TOKEN=foo
TURBO_TEAM=a-name

When using this remote cache implementation, a bogus TURBO_TOKEN works. It suggests using the repo name as value of TURBO_TEAM to separate cache from repo to repo.