메뉴
Rocky CLI quick reference. Use when you need to recall a `rocky <command>` invocation, JSON output shape, or exit-code behavior. For config authoring, use the `rocky-config` skill at the monorepo root. For the Rust output-struct cascade, use `rocky-codegen`.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | rocky |
| description | Rocky CLI quick reference. Use when you need to recall a `rocky <command>` invocation, JSON output shape, or exit-code behavior. For config authoring, use the `rocky-config` skill at the monorepo root. For the Rust output-struct cascade, use `rocky-codegen`. |
Concise cheat sheet for the rocky binary. For full detail:
rocky-config skill (monorepo root .claude/skills/)rocky-codegen skillengine/CLAUDE.md| Command | Purpose |
|---|---|
rocky init [path] | Scaffold a new project |
rocky validate -c rocky.toml | Validate config + models (no API calls) |
rocky compile --models models/ [--contracts contracts/] | Type-check models, validate contracts |
rocky test --models models/ [--contracts contracts/] | Run model tests against in-memory DuckDB (auto-loads data/seed.sql) |
rocky ci --models models/ [--contracts contracts/] | Combined compile + test |
rocky discover -c rocky.toml | List sources/connectors and tables |
rocky plan -c rocky.toml [--filter <k=v>] | Preview SQL (dry-run) |
rocky run -c rocky.toml [--filter <k=v>] | Execute the pipeline |
rocky run -c rocky.toml --resume-latest | Resume a failed run from the last checkpoint |
rocky run -c rocky.toml --partition KEY | Time-interval materialization (per-partition) |
rocky run -c rocky.toml --from KEY --to KEY | Partition range |
rocky run -c rocky.toml --latest / --missing | Partition discovery modes |
rocky state -c rocky.toml | Show watermarks |
rocky history [--model <name>] | Recent run history + trends |
rocky metrics <model> | Quality snapshots, alerts, column trends |
rocky lineage <model> [--column <col>] | Model/column lineage |
rocky lineage-diff [<base_ref>] | Per-changed-column downstream impact between two git refs (PR-comment Markdown) |
rocky doctor -c rocky.toml | Aggregate health checks (config, state, adapters, pipelines, state_sync) |
rocky drift -c rocky.toml | Schema drift report |
rocky optimize -c rocky.toml | Cost model + materialization-strategy recommendations |
rocky compare | Shadow vs production comparison |
rocky compact <fqn> / rocky compact --catalog <name> | OPTIMIZE/VACUUM SQL plan (single table or every Rocky-managed table in a catalog) |
rocky archive --older-than <span> (<fqn> / --catalog <name>) | DELETE + VACUUM SQL plan (single table or catalog-wide) |
rocky profile-storage | Column encoding recommendations |
rocky import-dbt --dbt-project <path> | dbt → Rocky migration |
rocky validate-migration | Validate a dbt → Rocky migration |
rocky hooks list / rocky hooks test <event> | Hook management |
rocky ai "<intent>" | AI-assisted model generation (needs ANTHROPIC_API_KEY) |
rocky ai-sync / ai-explain / ai-test | AI intent layer sub-commands |
rocky test-adapter <name> | Adapter conformance suite |
rocky lsp | Language Server Protocol server over stdio (for the VS Code extension) |
rocky mcp | Model Context Protocol server over stdio (for AI agents) |
rocky serve | HTTP API mode |
rocky export-schemas <dir> | Export JSON schemas for all *Output types (used by just codegen) |
| Flag | Effect |
|---|---|
--output json | Emit typed JSON (every command has a JsonSchema-backed struct — see rocky-codegen) |
--output table | Human-readable table (default) |
-c <path>, --config <path> | Config file (default: rocky.toml in cwd) |
--state-path <path> | Override embedded state store location |
--filter <k=v> | Scope to matching pipelines/sources/groups (e.g. --filter source=shopify) |
--resume / --resume-latest | Resume a checkpointed run |
--output-file <path> | Write JSON output to file instead of stdout |
| Code | Meaning |
|---|---|
| 0 | Success |
| 1 | Hard failure (config error, unreachable adapter, panic) |
| 2 | Partial success — pipeline ran but some tables failed. Valid JSON still emitted on stdout. The dagster integration explicitly handles this (allow_partial=True). |
Every --output json command emits a struct with version + command fields plus a command-specific payload. Full list in engine/CLAUDE.md → "JSON Output Schema" (ls schemas/*.schema.json | wc -l for the count). Highlights:
discover → { connectors: [{ id, client, components, tables, excluded_tables }] }
plan → { statements: [{ purpose, target, sql }] }
run → { tables_copied, materializations, check_results, drift, permissions, anomalies }
state → { watermarks: [{ table, last_value, updated_at }] }
doctor → { config, state, adapters, pipelines, state_sync }
compile → { models: [{ name, columns, contracts }], diagnostics, timings }
lineage → { nodes, edges } (— or ColumnLineageOutput when `--column` set)
history → { runs: [...] } (— or ModelHistoryOutput when `--model` set)
metrics → { snapshots, alerts, column_trend }
To change any of these, see the rocky-codegen skill — the Rust struct is the source of truth and regenerated bindings ship with every change.
| Issue | Fix |
|---|---|
source.catalog not set | Set the relevant adapter env var (e.g. FIVETRAN_SOURCE_CATALOG) |
schema does not start with prefix | Check [pipeline.<name>.source.schema_pattern].prefix matches your schema naming |
no connectors found | Check destination_id, API credentials, and that connectors are in "connected" state |
circular dependency | Check depends_on in model sidecar .toml files |
unsafe SQL fragment | Metadata column values must be NULL, numbers, or single-quoted strings |
drift detected, dropping target | Source column type changed unsafely — target recreated automatically (see drift.rs:is_safe_type_widening() allowlist) |
tag engine-v… already exists | You're trying to release — see the rocky-release skill |
codegen drift in CI | You edited output.rs without running just codegen — see the rocky-codegen skill |
cd engine
cargo build # Debug
cargo build --release # Release (used by `just codegen` + `regen-fixtures`)
cargo run -- discover -c rocky.toml --output json
cargo run -- plan -c rocky.toml --filter source=shopify
cargo run -- run -c rocky.toml --filter source=shopify
# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/rocky-data/rocky/main/engine/install.sh | bash
# Windows
iwr -useb https://raw.githubusercontent.com/rocky-data/rocky/main/engine/install.ps1 | iex
Both scripts filter GitHub Releases by the engine-v* tag prefix.
from dagster_rocky import RockyResource, load_rocky_assets
from dagster_rocky import RockyComponent
# ConfigurableResource (subprocess-based)
rocky = RockyResource(config_path="rocky.toml")
assets = load_rocky_assets(rocky)
# State-backed component (cached discover/compile)
# In defs/rocky/defs.yaml:
# type: dagster_rocky.RockyComponent
# attributes:
# config_path: rocky.toml
Three execution modes for rocky run in RockyResource:
run(...) — buffered subprocess.run, no Dagster contextrun_streaming(context, ...) — Popen with stderr streaming to context.logrun_pipes(context, ...) — full Dagster Pipes via PipesSubprocessClientSee integrations/dagster/CLAUDE.md for the layer architecture.
rocky-config (monorepo root) — Full rocky.toml authoring referencerocky-codegen — JSON output struct cascade (Rust → Pydantic + TS)rocky-new-cli-command — Adding a new rocky <verb>rocky-dsl-change — Changing .rocky DSL syntaxrocky-release — Tag-namespaced release workflowdatabricks / fivetran (this same engine-local skills dir) — REST API details for the two main adaptersDependency hygiene for the Rocky engine workspace — how to add/update deps via [workspace.dependencies], MSRV 1.88 policy, cargo-audit / cargo-deny / cargo-machete usage, and how the weekly security audit surfaces advisories.
Rocky CLI JSON-output schema cascade. Use when editing any `*Output` struct in `engine/crates/rocky-cli/src/output.rs` (or `commands/doctor.rs`), adding a new CLI command schema, or when a change could affect `schemas/*.schema.json`, `sdk/python/src/rocky_sdk/types_generated/`, or `editors/vscode/src/types/generated/`. Also use when CI reports `codegen-drift`.
Canonical `rocky.toml` authoring reference. Use when writing or reviewing a Rocky pipeline config — covers the 4 pipeline types (replication, transformation, quality, snapshot), adapter variants (duckdb/databricks/snowflake/fivetran), minimal-config defaults, env-var substitution, governance, checks, hooks, and the ${VAR:-default} syntax.
Top-level router for Rocky development tasks. Use this FIRST for any request to change Rocky — it points at the right subskill based on what's being touched. Also use for quick monorepo orientation (build, test, lint, install, where-is-what).
Rocky DSL (`.rocky` file) cross-subproject cascade. Use when adding/changing a Rocky DSL keyword, operator, or pipeline step — changes must land in the engine parser, compiler, VS Code TextMate grammar, and snippets together in one PR.
End-to-end checklist for adding a new `rocky <command>` subcommand across the engine, JSON schema export, Dagster Pydantic types, Dagster resource wiring, and VS Code extension command. Use whenever a new top-level CLI verb needs to be exposed.