The dashboard

Every server has an admin dashboard at /_/ (e.g. http://127.0.0.1:8090/_/), embedded in the binary. Create an administrator with the /_/setup#token=… link that serve prints at startup while there is no administrator, or with todo-server admin create you@example.com, then sign in.

• Data: every table (system tables grouped below), with a live, virtualized document grid (smooth at 10k+ rows; pages load as you scroll; new rows slide in, changed cells flash, deleted rows fade out, up to 5,000 loaded rows), filters on index fields, a search box for tables with a full-text search index (live, most relevant first), a quick filter over the loaded rows, JSONL/CSV export, a side panel with a JSON editor (insert, edit, delete; validated against the table's Rust type), and clickable ids that open the referenced document. The schema view lists indexes, tokenizers (with their language), vector index kinds (with HNSW parameters) and the table's analytics columns with their SQL types. Secrets (password and token hashes) are never shown, and system tables are read-only except the name, image and emailVerified of users.
• Functions: every function with its TS signature, a runner (JSON args, optionally as a user through a real, short-lived, audited session), latency (p50/p95) and calls/errors charts over the last hour to 7 days, from the execution log, and the function's own live log.
• Logs: a live tail of executions (filter by path or one function, kind, outcome; expand for the error, log lines and args); new executions slide in, tinted by outcome.
• Users: users, linked providers, sessions (revoked live; impersonation sessions marked), delete, and "open app as user": a short-lived, audited session of the user opened in a new tab at an allowed redirect URL.
• Files: stored files with thumbnails and previews (images, video, audio, text), download, delete, and upload by drag and drop; works with private storage (signed URLs). Id<"_storage"> values in the data grid link here.
• Workflows: workflows by state, an open one's steps (args, outcome, duration, job), and cancel.
• Scheduler: crons with their schedule, time zone, last and next run, and "run now" (system crons last); scheduled jobs live, by state, with cancel (running actions too), run now and retry.
• Migrations: every #[migration] in run order with its dependencies, live progress, errors linking to the failing document, the schema check, run / cancel / restart and dry runs with before/after examples; migrating tables are highlighted.
• Backups: create (optionally with aux.db and files), download, delete and restore backups (a stored one or an archive from your computer: checked completely, then applied at the next start; cancel, or "Restart now" under systemd/Docker); Configure… sets the schedule (interval, how many to keep, files and aux.db) and the destination (the data directory, another directory, or an S3-compatible bucket with its own credentials, with a "Test destination" check), applied at once; with --replicate-to, the replication status (target, lag, last segment and snapshot, generations) and a point-in-time restore picker (a time or a commit ts); export a table (JSONL/CSV) and import JSONL or CSV (insert, upsert or replace).
• Analytics (with --olap): the copy's state and lag, its tables and columns, a SQL editor (one SELECT, the sandbox of ctx.olap, audited) with the results as a table, and the whole result as a CSV or Parquet download; rebuild the copy.
• Environment: environment variables (ctx.env): add, edit, reveal, copy, remove, paste a .env; live, and queries that read a variable rerun as soon as it changes.
• Settings: OAuth providers (GitHub, Google, Discord, Apple, OIDC), SMTP (with a test email), redirect URLs, sign-ups, token lifetimes and CORS origins, applied without a restart. Fields fixed by the code, a flag or the environment are read-only; secrets can be replaced but are never shown.
• Audit log: document edits, user and session changes, settings and environment variable changes, impersonation, function runs, restores and analytics queries, live, filterable by administrator, action, table, document and time, with field-level diffs.
• Overview: commit ts, commits/s, connections, shared queries and reruns, the scheduler queue, database sizes, busiest and failing functions, the query cache hit rate and the analytics copy's state.

Ctrl+K (⌘K) opens a command palette (pages, tables, theme, sign out); Alt+1…9 and Alt+0 open the first ten pages in sidebar order.

It talks to the server over the same WebSocket protocol as apps (admin-only _system/* functions), so every view is live, plus /api/admin/* for logs, stats, backups, replication, analytics, export/import and impersonation (docs/CONTRACT.md §4). The built assets are committed, so cargo build needs no JS toolchain; after changing packages/dashboard (or the client packages it bundles), run bun run build:dashboard and commit crates/rxferrix/dashboard-dist (cargo test fails while it is stale). For UI work, cd packages/dashboard && bun run dev serves it on :5174 with /api proxied to :8090. Build without it with default-features = false on the rxferrix dependency.