drt-init
/drt-init
Prowadzi przez inicjalizację nowego projektu drt i konfigurację połączenia z hurtownią danych dla osób zaczynających pracę z tym narzędziem.
Ocena i bezpieczeństwo
- Pliki w paczce
- 1
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 27.10.2026, bez odwołań do wycofanych modeli i API.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
Co robi i kiedy
Prowadzi przez inicjalizację nowego projektu drt i konfigurację połączenia z hurtownią danych dla osób zaczynających pracę z tym narzędziem.
Oryginalny opis (EN)
Guide a user through initializing a new drt project. Use this skill when a user is setting up drt for the first time, asks how to get started with drt, or needs help configuring their data warehouse connection.
Jak działa
- Steps
- Tips
Podgląd SKILL.md
---
name: drt-init
description: >
Guide a user through initializing a new drt project. Use this skill when
a user is setting up drt for the first time, asks how to get started with
drt, or needs help configuring their data warehouse connection.
---
Guide the user through initializing a new drt project.
## Steps
1. Confirm drt is installed. Pick the install line for the user's source:
```bash
pip install drt-core # core (DuckDB / SQLite / REST API source — no extras)
pip install drt-core[bigquery] # BigQuery source
pip install drt-core[postgres] # Postgres / Redshift source
pip install drt-core[mysql] # MySQL source
pip install drt-core[clickhouse] # ClickHouse source
pip install drt-core[snowflake] # Snowflake source
pip install drt-core[databricks] # Databricks source
pip install drt-core[sqlserver] # SQL Server source
pip install drt-core[deltalake] # Delta Lake source
pip install drt-core[iceberg] # Iceberg source
```
(Or `uv add drt-core[...]` if the user prefers uv.)
2. Pick the right starter shape. Four flows produce the same sync YAML shape:
**A) Template scaffolds (fastest, ~3 commands)** — best when the user just wants something running:
```bash
mkdir my-drt-project && cd my-drt-project
drt init --template duckdb_to_rest # DuckDB → REST API (no accounts needed, uses httpbin.org)
```
Other templates:
```bash
drt init --template list # see all available templates
drt init --template postgres_to_slack # operational alerts
drt init --template duckdb_to_hubspot # CRM-style upsert
```
Each template prints next-steps for the env vars / source data it needs.
**B) Interactive wizard (guided)** — best when the user wants to be walked through a real warehouse setup:
```bash
mkdir my-drt-project && cd my-drt-project
drt init
```
Prompts for project name and one of the interactive wizard's source types (**bigquery / duckdb / sqlite / postgres / redshift / clickhouse**), then asks for source-specific connection fields and auth method. The other built-in source types (**snowflake / mysql / databricks / sqlserver / deltalake / iceberg / rest_api**) remain available through templates or a hand-authored `profiles.yml` entry; `drt profile add` also covers MySQL and Snowflake.
**C) dbt manifest scaffold** — best when a dbt project already has built models:
```bash
drt init --from-dbt target/manifest.json
```
This interactively lists dbt models and writes selected
`syncs/sync_<model>.yml` files without overwriting existing files.
**D) Apache Ossie semantic-model scaffold** — best when the source metadata
is already portable Ossie YAML/JSON:
```bash
drt init --from-ossie semantic_model.yaml --destination postgres
drt init --from-ossie semantic_model.yaml --dataset customer --dataset account --dialect SNOWFLAKE
```
By default this writes one upsert sync per dataset with a `primary_key` and
prints why unkeyed datasets are skipped. Supported destination stubs are
`postgres` and `hubspot`; generated TODOs still need credentials, target
table/object, and property mappings. A requested field dialect falls
back to `ANSI_SQL`; metrics are listed but not generated because entity grain
is unclear.
The template and wizard flows write a full project shell; metadata scaffolds
write sync YAML into `syncs/` (or Ossie's `--output-dir`):
```
my-drt-project/
├── drt_project.yml
└── syncs/<name>.yml # first sync, runnable
```
plus credentials to `~/.drt/profiles.yml`.
3. Set up auth for the chosen source (only the relevant bullet):
- **BigQuery**: `gcloud auth application-default login` or `export GOOGLE_APPLICATION_CREDENTIALS=/path/to/sa.json`
- **Postgres / Redshift / MySQL / ClickHouse / Snowflake / SQL Server**: set the password env var the profile references (e.g. `export DRT_PG_PASSWORD=...`)
- **Databricks**: set `DATABRICKS_TOKEN` (personal access token from Workspace → User Settings)
- **DuckDB / SQLite**: nothing — file path only
- **REST API source**: set the bearer token env var the profile references (e.g. `export REST_API_TOKEN=...`)
For deployed environments, an `*_env` field may contain an `aws-sm://`,
`gcp-sm://`, or `vault://` provider URI instead of an env-var name; install
the matching optional extra. For Postgres, Snowflake, Databricks, and
BigQuery, keep `managed_schema: _drt` (default) or choose an isolated
schema/dataset if snapshot-diff incremental or warehouse-backed state will
be used. This is drt bookkeeping space, not the query's default schema.
Both `drt init` flows already write a profile to `~/.drt/profiles.yml`. To add
another later (a second warehouse, a prod/dev split) without re-running init:
```bash
drt profile add <name> # interactive — prompts for source type + connection fields
drt profile list # names + source types already configured
drt profile show <name> # inspect one, inline secrets masked
```
4. Validate the setup — **credential first**, since that is the step most likely to fail:
```bash
drt profile test <name> # actually connects, using the source's own connection check
drt doctor # env-level triage — catches missing env vars, malformed profile, etc.
drt validate # YAML schema check
drt list # confirm syncs are discovered
drt plugins list # installed entry-point extensions + isolated load failures
```
`drt doctor` and `drt validate` inspect configuration; only `drt profile test`
proves the warehouse accepts the credential. Run it before the first sync so a
bad password surfaces here rather than mid-run.
5. Offer to create a first sync using the `/drt-create-sync` skill.
## Tips
- `drt_project.yml` selects which profile from `~/.drt/profiles.yml` to use; override per-run with `--profile <name>` or `DRT_PROFILE`.
- Put each sync in a separate `syncs/<name>.yml` file (sync discovery is glob-based).
- `drt sources --detailed` and `drt destinations --detailed` print every connector's required env vars and a sample YAML stanza — useful when hand-authoring beyond the templates.
- `drt run --dry-run` runs through the engine without writing data; `--dry-run --diff` previews record-level changes for queryable destinations. `drt run --limit 10` (#774) really loads at most 10 rows for a safe first send (watermark won't advance; refused for `mode: mirror` / `replace`).
- A project that needs shared, SQL-queryable state can set `state: {backend: warehouse, connection_profile: <profile>}`. Postgres, Snowflake, Databricks, and BigQuery store `_drt_runs`, `_drt_history`, and `_drt_dlq` under that profile's `managed_schema`; this is separate from `sync.watermark.storage`.
- `drt build` (#777) runs each sync and then its `tests:` in one pass — the `dbt build` shape; `drt build --select tag:crm --fail-fast` stops at the first failure.
- `drt test --unit` runs top-level `unit_tests:` fixture rows through computed fields, mappings, masking, and metadata without source/destination credentials.
- `drt status` shows recent run results; `drt status --output json` is the CI-friendly form.
- Every `drt run` that resolves syncs also writes `target/drt/run_results.json`; `--target-path` relocates it for CI artifact upload.
- PR CI can generate a schema-v3 manifest and run only changed definitions with `--select state:modified --state <manifest.json>` (also valid on build/test/validate).
- `drt docs generate` (v0.7.11+) renders a static HTML catalog of your project — every sync, its model, destination, and a lineage DAG — to `target/docs/`. Labels are docs-safe by default (endpoints/emails/phones stripped, #696); add `--format mermaid` or `--format json` for other outputs.
- For non-US BigQuery datasets, set `location` in `profiles.yml` (e.g. `"EU"`, `"asia-northeast1"`).
- Want to try drt without installing locally? Open the repo in GitHub Codespaces — the devcontainer ships drt + a seeded DuckDB warehouse + the `duckdb_to_rest` template pre-scaffolded.
Wersje
W bazie nie ma innych skilli o nazwie drt-init.
Migawka bazy z 13.11.2026 · wersja paczki 6e408b6abd02 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100drt-create-sync
Generuje plik YAML konfiguracji synchronizacji drt, łączącej tabelę hurtowni danych z zewnętrzną usługą w potoku Reverse ETL. - 100drt-debug
Diagnozuje i naprawia nieudaną synchronizację drt: błędy uwierzytelniania, limity zapytań, błędy szablonów i inne problemy konfiguracji. - 100get-started
Prowadzi przez instalację gitagent i stworzenie pierwszego agenta ze szkieletem projektu, konfiguracją i walidacją dla osób zaczynających z narzędziem. - 100initializing-memory
Prowadzi przez inicjalizację lub reorganizację pamięci agenta i tworzenie skutecznych plików pamięci, przy poleceniu /init. - 100warehouse-init
Inicjuje odkrywanie schematu hurtowni danych i generuje plik .astro/warehouse.md z metadanymi wszystkich tabel do szybkich wyszukiwań; odświeża się po zmianie schematu. - 100resolve-edit
Pomaga w cięciu, przycinaniu, tempie i przebudowie osi czasu w DaVinci Resolve, także offline z plikami DRT.