No credit cardStart free

Shipping log

Changelog

v1.4.0

June 2026

Live dashboard, no more refresh

The dashboard now updates itself. When something changes, the affected list refreshes on its own. No reload, no polling spinner.

  • Live everywhere: create a connector, chart, page, transform, saved query or search and it shows up immediately, whether the change came from the UI, the rq CLI, or another open tab
  • Sync status in real time: a connection flips from syncing to ready the moment the sync worker finishes, instead of waiting on a 5-second poll
  • One stream per workspace: built on Postgres LISTEN/NOTIFY over a single Server-Sent Events connection, so the browser stops hammering the API with refetch loops

v1.3.2

May 2026

Codex / OpenAI parity with Claude

Codex now has the same rawquery CLI integration path as Claude.

  • 100% Codex / OpenAI parity with Claude: rq codex-init installs a native Codex skill in .agents/skills/rq/
  • `rq update` now refreshes the Codex skill too when it already exists locally, same behavior as the Claude path
  • One CLI, both agent stacks: rawquery workspace introspection, SQL help, connections, transforms, charts, pages, and search workflows are now wired for both Claude and Codex

v1.3.1

April 2026

Join fan-out detection

Text-to-SQL commonly writes joins that silently multiply rows and inflate SUM and COUNT. rawquery now flags the pattern before the wrong number lands in a dashboard.

  • Fan-out warnings on every query: when 2+ tables are joined and aggregated at the top level, a non-blocking heads-up surfaces in the CLI, the dashboard, and the API response
  • `rq` Claude skill updated to pre-aggregate 1-to-many tables in CTEs, so LLM-written queries avoid the trap by default
  • `rq jobs list`: finally list your async jobs from the CLI, with --status and --limit filters
  • Async concurrency cap: per-workspace limit on pending and running jobs (3 free, 10 team, 20 business), so a burst of long queries can't fill the queue
  • Query surface hardening: tightened the SQL parser allowlist around DuckDB internal functions and settings catalogs after an internal adversarial audit, and added a per-IP throttle on public chart / page / search routes
  • `rq queries show` no longer errors on saved queries with numeric default parameters

v1.3.0

April 2026

Full-text search, Parquet ingest

Search any text column in your lakehouse, and push real data in without a JSON middleman.

  • Full-text search: rq fts create to build an index, query with the fts_search('index', 'query') table function, fuzzy matching for typos and phrase queries
  • Parquet push: rq push --file x.parquet streams up to 2 GB, preserves types exactly, no JSON intermediate. The right path for warehouse migrations or df.to_parquet() handoffs

v1.2.0

April 2026

Embeddings, rebuilt

We reworked the embedding pipeline end to end. Faster throughput, bounded memory at any table size, and pre-flight estimates that match reality.

  • Rebuilt embedding engine, materially faster throughput
  • Separate billing for searches and rows embedded, pay for what you actually use
  • Team 2M rows/month + €0.50/1M extra. Business 20M/month + €0.30/1M extra
  • Pre-flight estimate before every embed. Rows, duration, and cost, measured rather than guessed
  • Bulk jobs scale to any table size without memory surprises
  • Output appears progressively as a bulk job runs. Query partial results while it's still going

v1.1.3

April 2026

Async jobs in the Query Editor

Async queries now have a home in the editor. Submit, watch, and pull back results without leaving the page.

  • New Run async button alongside Run. One click for queries that don't fit in the interactive window
  • Jobs tab next to Results and History: status, duration, result preview, cancel, copy SQL back
  • CLI update notifier: rq checks once a day and tells you when a newer version is out, so you're not stuck on a stale binary

v1.1.2

April 2026

Heavy queries that finish

Async jobs are now properly async. Submit a long-running query, come back later, get your answer.

  • `rq jobs submit` handles heavy queries that don't fit in the interactive window
  • `rq jobs wait` returns your results when they're ready
  • Results are kept for 24 hours so you can fetch them on your own time

v1.1

April 2026

Automatic Iceberg compaction

A background worker merges small parquet files into big ones. Your lakehouse stays fast as it grows. No cron, no dbt job, no babysitting.

  • Dedicated worker with bin-packing, retry, orphan garbage collection
  • rq compact CLI for on-demand runs
  • Queries pause briefly during compaction with a clear message, not an error
  • Memory safety across the stack: Docker limits, DuckDB guard, OOM errors return actionable messages

v1.0

April 2026

Production launch

First stable release.

  • Custom connectors: connect any HTTP or GraphQL API with a declarative JSON spec. OAuth2, custom auth, pagination, schema evolution. No code, no SDK, no waiting on us.
  • Wire protocol over TLS: connect Metabase, Looker, Tableau, Grafana, or any Postgres-compatible client. JDBC-compatible.
  • OAuth 2.1 server for MCP: plug Claude.ai, ChatGPT, or any modern AI client directly into your workspace.
  • Self-hosted embeddings: EU-hosted, no third-party dependency.
  • SSE-KMS encryption at rest: AES-256 on every Iceberg parquet file.
  • Iceberg table export: take your data and leave, any time, zero friction.
  • Social login via GitHub and Google (OIDC)

v0.9.9

March 2026

Vector search

Semantic search on your lakehouse. DuckDB VSS extension, HNSW indexing, ACORN predicate filtering.

  • rq embed generates embeddings for any text column
  • Search CRUD, public search pages at /s/{token} with password gates
  • Billing integrated, same pricing logic as queries, visible in usage dashboard

v0.9.5

March 2026

Security hardening and schema versioning

Enterprise-grade foundations before a wider audience.

  • Workspace isolation hardened across wire protocol, legacy query paths, and public tokens
  • Authentication hardening (P0 + P1 fixes), public /security page
  • Alembic migrations: schema versioned, upgrades idempotent at startup
  • MCP server live at mcp.rawquery.dev with Claude Code integration docs

v0.9

February 2026

Live mode

Query external databases without copying a single row.

  • Live mode: attach external Postgres or MySQL read-only via DuckDB scanners. Query immediately, no sync required.
  • Transforms can source from live connections; charts can source from transforms directly
  • Stress-test hardening: sync worker scaled to 3 replicas via FOR UPDATE SKIP LOCKED
  • Automated pg_dump backups with healthcheck alerting

v0.8

January 2026

Paddle billing and team workspaces

Pricing, permissions, and identity. The boring stuff done right.

  • Paddle billing: Merchant of Record, EU-first, no payment data on our servers
  • RBAC (OWNER / ADMIN / MEMBER / VIEWER) with workspace member management
  • Parameterized saved queries: write once, run with parameters, share the URL
  • Password reset, email verification, profile management
  • Per-minute rate limiting, Postgres-backed, consistent across workers

v0.5

January 2026

Charts and public pages

Every view is a URL. Send the link, the colleague sees the answer.

  • Charts: Chart.js preview in the editor, publish with a public URL, refresh on open
  • Pages: grid of charts, publish as a dashboard, password gate optional
  • Interactive DAG for transforms: pipelines vs standalones, scalable node layout
  • Schema evolution for HTTP connectors: columns added automatically on the fly

v0.4

December 2025

Custom connectors and rq push

The connector that connects to anything. Plus push your own data in one command.

  • HTTPConnector: declarative JSON spec for any REST or GraphQL API
  • OAuth2, custom auth, pagination, flatten + lift transformations
  • `rq push`: ad-hoc data ingestion from the CLI, creates Iceberg tables on the fly
  • rq connect command for scripted onboarding

v0.1

December 2025

First light

The stack. SQL in, answers out, no infra to run.

  • Wire protocol: a Postgres-compatible server on port 5433. Any PG client works, psql included.
  • Monaco SQL editor: schema browser, query history, export to CSV / JSON / Parquet
  • Transforms: SQL models with DAG execution, cron scheduling
  • Execute API: programmatic SQL via API key, rate-limited, structured errors
  • Worker pool with auto-scaling for async queries
  • Lakehouse catalog: browse tables, preview data, see schemas
  • Iceberg on S3 via Nessie catalog, one branch per workspace