Skip to content

Releases: arezki1990/dbt-zed

zdbt 0.4.0

Choose a tag to compare

@arezki1990 arezki1990 released this 14 Sep 15:59

Fourth feature release. 93 commits since 0.3.0.

Extract-Load

zdbt now moves data as well as modelling it. A new EL section next to the dbt panels builds extract-load pipelines and runs them locally or on a server you own — and it works in projects without dbt too.

  • Pipeline canvas — connections, streams and one Cast & Map step per stream drawn as nodes; preview data straight from a node; multi-select tables and drag-drop to create streams; rename, remove and reorder streams from the node menu.
  • Connections — DuckDB, Postgres, files and Oracle (Oracle's pure-Rust thin driver, with a thick-driver fallback that reaches 10g and 11g). Connection editor with profiles (dev / recette / prod). Specs are plain YAML under el/, edited through normal buffers so every change is undoable and diffable.
  • Loads — incremental loads with typed cursors, MERGE commits and a state store; per-stream sync settings; runs from the IDE or the headless zdbt el CLI.
  • Remote daemon — add a server from a wizard that installs zdbt-el-serve over SSH, watch its status live, and deploy pipelines explicitly with a chosen profile: nothing runs on a remote until you ship it, and the deploy popup shows what will change first. Remote run history with next-run times, per-run detail with logs, SQL highlighting and cast-failure samples.
  • Data explorer and query console per connection; table listings, queries and previews can run on a remote's worker.
  • Secrets never surface: specs reference ${VAR} names only; values reach workers through the environment.
  • The engine, worker and daemon live in their own repository; database drivers compile only into the worker, never into the IDE.

Windows

  • The app and installer are now named zdbt — no more "Zed Dev".

Fixes

  • ctrl-enter in a dbt model runs dbt show again instead of opening the inline assistant.
  • Fixed a workspace double-lease crash triggered by console refreshes.

Builds are unsigned. On macOS 15+, use System Settings → Privacy & Security → Open Anyway (the older right-click → Open bypass was removed by Apple). Verify downloads with shasum -a 256 --ignore-missing -c SHA256SUMS.txt.

zdbt 0.3.0

Choose a tag to compare

@arezki1990 arezki1990 released this 08 Sep 07:54

Third feature release. 3 commits since 0.2.0.

Database explorer

  • New left-dock Database panel — browse your warehouse as a tree: databases → schemas → relations → columns. Built entirely from the dbt artifacts (manifest.json + catalog.json), so it works offline with no database drivers and no live connection.
  • Shows models, seeds, snapshots and sources with materialization, owner, row counts and sizes from the catalog; column types in table order with descriptions (manifest docs overlaid on catalog types).
  • Click a relation to preview its data in the results grid — dbt models run through dbt show --select (and fill the Compiled and Lineage tabs), everything else previews via an inline select *.
  • Right-click: Preview data, Open model file, Copy fully-qualified name, Copy name.
  • Filter box searches relations and columns; matches auto-expand to show where they live.
  • Works from manifest.json alone; run dbt docs generate to populate columns, row counts and sizes.
  • Toggle with dbt: toggle database focus from the command palette.
  • The tree reloads automatically when the artifacts change on disk.

Release integrity

  • SHA256SUMS.txt is now generated and verified by CI on every release, from the exact assets published.

Builds are unsigned. On macOS 15+, use System Settings → Privacy & Security → Open Anyway (the older right-click → Open bypass was removed by Apple). Verify downloads with shasum -a 256 --ignore-missing -c SHA256SUMS.txt.

zdbt 0.2.0

Choose a tag to compare

@arezki1990 arezki1990 released this 04 Sep 14:13

Second feature release. 43 commits since 0.1.1.

Lineage

  • AST-based column lineage — compiled SQL is parsed with a real SQL parser (Snowflake dialect, generic fallback) instead of matching names. CTEs resolve as real scopes, joins resolve per alias, subqueries and unions are followed, and SELECT * expands. Measured at 1709/2216 catalog columns (77%) on the reference project; anything it cannot follow falls back to the previous name-matching heuristic.
  • Column-focus DAG — right-click a column to open a graph of only that column's path, with the expression at each hop.
  • Transformation badges and tooltips per node, rename-aware column edges, and cmd+scroll zoom anchored on the browsed model.
  • The canvas now auto-centres on the model you are browsing, and stops fighting manual panning.
  • Model details sidebar — description, relation, row count, size, tags and column types from the catalog.

Agents

  • Built-in MCP server exposing six dbt tools (dbt_list_models, dbt_model_info, dbt_lineage, dbt_column_lineage, dbt_show, dbt_compile). It auto-registers with Zed's agent when a worktree contains dbt_project.yml — no configuration. The same binary serves any MCP client via --dbt-mcp.

Results grid

  • Rebuilt from scratch: working horizontal scroll, always-visible scrollbars, resizable columns, virtualized rows, zebra striping, and click-a-cell to inspect the full untruncated value.

Connection

  • New Connection tab showing what a run would actually use: resolved dbt binary and its source, profiles.yml path, profile, active target and which layer set it. No value from profiles.yml is ever read into the UI — only parameter names — so credentials cannot leak.

dbt Core

  • dbt Core is now supported alongside Fusion, with a Python virtualenv installer and native file/directory pickers in settings.

Fixes

  • Fixed a crash on cmd-enter caused by reading the workspace during its own update.
  • dbt: show model data is now bound on Linux and Windows (ctrl-enter, ctrl-k d) — previously macOS-only.

Builds are unsigned. On macOS 15+, use System Settings → Privacy & Security → Open Anyway (the older right-click → Open bypass was removed by Apple). Verify downloads with shasum -a 256 --ignore-missing -c SHA256SUMS.txt.

zdbt 0.1.1

Choose a tag to compare

@arezki1990 arezki1990 released this 03 Sep 08:29

Fixes from the first day in the wild, plus zdbt can now install dbt for you.

New

  • dbt Fusion auto-install — when no dbt binary is configured or on PATH, zdbt downloads the official dbt Fusion CLI from dbt Labs' CDN and uses it for queries, compilation, lineage, and the language server. Two new settings on the dbt page: dbt.auto_install (default on) and dbt.fusion_version (latest, dev, canary, or a pinned version). Your own dbt.binary setting and PATH installs always take precedence.

Fixed

  • "no result rows found" on models that don't exist — when dbt show --select <name> matched no nodes (e.g. running a file that isn't a parsed model), Fusion exits successfully with the explanation only in stderr, which the panel hid. The panel now reports the selection warning directly with a hint. (Reported from a Windows install running Fusion preview.218.)

Assets

Asset Platform
zdbt-macos-aarch64.dmg macOS (Apple Silicon)
zdbt-linux-x86_64.tar.gz Linux (x86_64)
zdbt-windows-x86_64-setup.exe Windows (x86_64)

Assets are uploaded by CI as each platform build finishes. Verify downloads against SHA256SUMS.txt (uploaded last). Builds are unsigned, dev-channel (install alongside official Zed, never auto-update over it). Licensed GPL-3.0; not affiliated with Zed Industries or dbt Labs.

zdbt 0.1.0

Choose a tag to compare

@arezki1990 arezki1990 released this 02 Sep 14:03

First release of zdbt — a dbt IDE built as a fork of Zed v1.17.2, with native (GPUI, no webview) dbt support.

Highlights

  • dbt SQL language — Jinja-aware SQL highlighting, default for .sql files
  • Language server — dbt Fusion dbt lsp preferred (with a workaround for its didChangeConfiguration crash), community Go LSP auto-download fallback
  • cmd+enter query results — run the current model or a selected SQL chunk via dbt show, results in a sortable/searchable data grid with resizable columns, column picker, pinned row numbers, and CSV export
  • Compiled SQL view in a read-only highlighted editor
  • Interactive lineage canvas — layered auto-layout, materialization colors, per-node collapse handles, node drag / graph pan / semantic zoom, column-level lineage with click-to-highlight, collapsible tree sidebar; browse-driven both ways (open file ⇄ click node)
  • Project automation — nested dbt_project.yml discovery, in-repo profiles (local_profiles/, profiles/, .dbt/), .env loading up to the repo root, auto dbt parse + dbt compile --write-catalog
  • Settings page — a full "dbt" page in Zed's settings UI (11 settings), plus a gear shortcut in the results panel

Assets

Asset Platform
zdbt-macos-aarch64.dmg macOS (Apple Silicon)
zdbt-linux-x86_64.tar.gz Linux (x86_64)
zdbt-windows-x86_64-setup.exe Windows (x86_64)

Notes

  • Builds are unsigned and use the dev release channel: they install alongside official Zed and never auto-update over it. On macOS, right-click → Open the first time.
  • You need dbt installed yourself (dbt Fusion recommended); it is not bundled.
  • Licensed GPL-3.0 (see LICENSE-GPL and the README's licensing section). This project is not affiliated with or endorsed by Zed Industries or dbt Labs.