# Data Deploys the opencode.ai/data site from `packages/data` as three Workers with `cf`: | Worker | Config | What it is | | ---------------------------- | ---------------------- | ----------------------------------------------------------------------------- | | `opencode-data-` | `cloudflare.config.ts` | The SolidStart site, on `/data*` and `//data*` of the stage's host | | `opencode-data-og-` | `og.config.ts` | Open Graph images, reached only through the site's `OG_WORKER` binding | | `opencode-data-sync-` | `sync.config.ts` | The `opencode-data-sync-` Workflow and the cron handler that starts it | `script/build.ts` builds all three into one Build Output tree and `script/deploy.ts` deploys them with `cf deploy --prebuilt`, the OG Worker first. The `data` job in `.github/workflows/deploy.yml` runs `deploy:dev` or `deploy:production` and passes the Worker secrets from AWS Secrets Manager (`opencode//DATA_DATABASE_HOST`, `DATA_DATABASE_USERNAME`, `DATA_DATABASE_PASSWORD`, `R2_SQL_AUTH_TOKEN`, `DATA_HIDDEN_MODELS`) with `--secrets-file`. Each Worker gets only the secrets its config declares. The database is the `opencode-stats` PlanetScale database: the `dev` branch for dev and `production` for production, through the `data-worker-` password. It is not managed as code. Every config, `script/build.ts` and `script/deploy.ts` take the stage from `--mode`, which must be `dev` or `production` (`stage.ts`). Anything else, including no mode, throws instead of falling back to production, so pass `--mode` to `cf` commands that read the configs too (`cf deploy --dry-run --mode dev`). ```bash bun run dev # the site on your machine (packages/data/app) bun run build:dev # or build:production bun run typecheck ```