commands

`remote`

github edit

remote

Reads the configured Cloudflare remote archive through the Worker API. The Worker is deployed separately from discrawl in openclaw/crawl-remote with Wrangler; discrawl stores an endpoint/archive and never deploys the service itself.

#Usage

discrawl remote status
discrawl remote archives
discrawl remote whoami
discrawl whoami

#Reports

  • remote status returns the crawlkit control status for the configured archive
  • remote archives lists archives visible to the authenticated identity
  • remote whoami and whoami report the GitHub/org identity associated with the token

#Notes

Remote commands require [remote] config with mode = "cloud", endpoint, archive, and a token in remote.token_env (default: DISCRAWL_REMOTE_TOKEN). They do not open or create the local SQLite archive.

#See also

#Scheduled publication

The existing backup workflow can also publish its successful source state to D1 and R2. It saves the runtime cache before Cloud publication. A Cloud failure therefore does not discard the Discord sync or the published Git snapshot.

Enable DISCRAWL_CLOUD_PUBLISH_ENABLED=1 only after the backend supports discrawl.current-state.v1 and publisher credentials are configured. The first run checks the archive inventory before adoption. Configure these GitHub Actions secrets:

  • DISCRAWL_CLOUD_ENDPOINT: the Worker URL.
  • DISCRAWL_CLOUD_ARCHIVE: the existing archive ID.
  • DISCRAWL_CLOUD_ACCESS_CLIENT_ID and DISCRAWL_CLOUD_ACCESS_CLIENT_SECRET:

    • service credentials, if the endpoint also requires Cloudflare Access.

The job uses its existing DISCORD_BACKUP_TOKEN for Cloud login. The Worker still checks the token's organization and team membership; an unauthorized token fails before any Cloud data changes. No local CLI credential is copied into CI.

The workflow creates one filtered SQLite export with discrawl cloud publish --export-only PATH. The output must not already exist. This command needs no remote credentials and excludes direct messages, deleted records, raw payloads and history. The scheduled publisher uses that same file for D1 batches and the R2 gzip snapshot. The existing cloud publish flags remain supported and also use one fixed export for both destinations.

D1 keeps one current set of rows. Unchanged key ranges are skipped; changed ranges include updates and deletions. Failed uploads resume from server progress. Query commands stay available as batches update in place. Readers can see newer batches before the whole refresh finishes. The last completed SQLite snapshot remains downloadable. R2 retains the current and previous snapshots, plus an unfinished upload; it leaves other archives and legacy objects alone. Public job logs contain aggregate counts, not archive IDs or message data.

Before first adoption, the publisher checks remote keys against the runtime, including deleted source rows. If a remote key is absent, publication stops for owner review. Do not bypass this check or reset the remote tables.

#Reconciling an older hosted archive

If first adoption finds identities missing from the source, keep Cloud publication paused. The backup workflow has an optional reconcile_cloud input for a manual run on reviewed main. Its private DISCRAWL_CLOUD_RECONCILE_PLAN Actions secret contains a gzip/base64 JSON plan with the target archive, member identities and message identities, including their guild/channel and original creation time. The plan must match the configured archive and collector guild.

The helper checks current guild/channel access, then fetches those exact records from Discord. It restores live records through the normal store converters. Only HTTP 404 with Discord's specific Unknown Member or Unknown Message code creates a deletion record. Permission errors, other failures and mismatched identities stop before observations are applied. Existing source rows, including deletion records, are preserved. Media downloads and embeddings are disabled.

Git publication and cache persistence follow reconciliation in the same job. The normal Cloud adoption guard remains required. Remove the private plan secret after recovery succeeds; scheduled runs never enable reconciliation themselves.