Browser UI: Import / Restore a graph from an RDB dump (local upload + S3 URL)

[shubham21155102]

Summary

The FalkorDB Browser today supports exporting a graph (/api/graph/[graph]/export) but has no symmetric import / restore path. Operators who already have an RDB dump — produced by BGSAVE, the Browser's own Export, KubeBlocks backup, or a teammate — have to drop out of the UI and into shell + Docker volume mounts to restore it.

This issue proposes a first-class "Restore graph" flow in the Browser, supporting two sources:

1. Local file upload — pick a .rdb / .dump from disk in the Browser
2. Remote URL — paste an s3://…, https://…, or presigned URL

Motivation / why this is genuinely missing

• Existing users running self-hosted FalkorDB (Docker, bare metal, KubeBlocks) keep RDB snapshots in object storage (S3, GCS, R2, MinIO). Restoring one today requires:
  • stopping the container,
  • aws s3 cp to the data dir,
  • editing REDIS_ARGS / --dbfilename,
  • restarting,
  • hoping the file matches the graph key (see [crash] importing a dump using a different graph key causes the redis server to crash FalkorDB#2048 — wrong key crashes the server).
• The Browser already owns the "graph lifecycle" UX (create, duplicate, delete, export). Restore is the obvious missing verb.
• Docs (Browser, Migration) currently send users to Docker volume tricks for a flow that should be a button.
• Related: KubeBlocks already wires S3 BackupRepo on the ops side (docs) — there's no in-Browser equivalent.

Proposed UX

On the graph list / graph management screen, next to Export Data, add Restore Graph:

• Modal with two tabs: Upload file | From URL
• Target graph name (default: filename stem; warn if it collides with an existing graph — ref [crash] importing a dump using a different graph key causes the redis server to crash FalkorDB#2048)
• Source field:
  • Upload: file picker, multipart upload, size cap configurable via env
  • URL: text input accepting https://, s3:// (with optional AWS creds form / presigned URL), and gs:// / azure:// as follow-ups
• Progress + final status; on success, refresh the graph list and select the restored graph
• Hard guardrails: confirm overwrite if target graph exists; reject mismatched graph keys early to avoid the [crash] importing a dump using a different graph key causes the redis server to crash FalkorDB#2048 crash

Proposed implementation sketch

• New API route: app/api/graph/[graph]/restore/route.ts
  • POST multipart for upload
  • POST JSON { "source": "s3://bucket/key", "credentials": {...} } for URL
• Server-side:
  • Stream the dump to a temp file (or in-memory for small dumps)
  • Use RESTORE for individual graph keys when feasible, or invoke DEBUG RELOAD / a new GRAPH.RESTORE primitive (paired core request to follow, will cross-link)
  • For S3: support both static IAM keys (form) and presigned URLs (no creds needed); never persist credentials
• Tests: round-trip — Export → Restore → query parity
• Docs: add a "Restore via Browser" section to docs.falkordb.com/browser

Out of scope (separate issues if wanted)

• Scheduled / recurring restores
• Multi-graph bulk restore
• Non-RDB formats (CSV/Parquet — already covered by FalkorDBImporter)

Willing to contribute

Happy to open a draft PR against falkordb-browser for the upload + presigned-URL path first, then layer s3:// with credentials once API shape is agreed. Would appreciate maintainer input on:

1. Preferred restore primitive on the core side (RESTORE per key vs. a new GRAPH.RESTORE accepting a URL) — will file paired issue in FalkorDB/FalkorDB
2. Whether S3 creds should live in the Browser session or be passed per-request only

[shubham21155102]

Paired core-primitive request filed: FalkorDB/FalkorDB#2078 — proposes a GRAPH.RESTORE <key> FROM <source> command (file path / https / s3) that this Browser feature would call into. Happy to align the API shape across both.

[overcut-ai]

Completed Working on "Automated Ticket Triage "

✅ Ticket triage completed for #1802. Applied labels: feature. No duplicate was found in targeted top-result checks, so duplicate was not applied (no qualifying similar issue/candidate link available). needs-info was not applied because the issue contains sufficient detail and actionable scope.

✅ Workflow completed successfully.

---

👉 View complete log

[shubham21155102]

Draft PR up: #1803 — implements the upload + presigned-URL slice. s3:// with credentials, gs://, azure:// deferred to follow-ups so this stays small. Marked draft pending maintainer feedback on the API shape (esp. the open question around a core GRAPH.RESTORE … FROM … primitive — paired in FalkorDB/FalkorDB#2078).

[barakb]

@dudizimber please provide your feedback

[dudizimber]

@dudizimber please provide your feedback

This must be held off until FalkorDB/FalkorDB#2078 is implemented/closed.
Ideally we'd use GRAPH.RESTORE <key> FROM <source> to match the LOAD CSV syntax.