local-data-platform 0.1.1__tar.gz → 0.1.3__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (105) hide show
  1. local_data_platform-0.1.3/PKG-INFO +1065 -0
  2. local_data_platform-0.1.3/README.md +1007 -0
  3. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/pyproject.toml +14 -1
  4. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/__init__.py +3 -1
  5. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/catalog/provider.py +38 -4
  6. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/cli.py +152 -29
  7. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/config.py +50 -3
  8. local_data_platform-0.1.3/src/local_data_platform/dbt_layer/__init__.py +16 -0
  9. local_data_platform-0.1.3/src/local_data_platform/dbt_layer/build.py +349 -0
  10. local_data_platform-0.1.3/src/local_data_platform/dbt_layer/cli.py +217 -0
  11. local_data_platform-0.1.3/src/local_data_platform/dbt_layer/project.py +386 -0
  12. local_data_platform-0.1.3/src/local_data_platform/dbt_layer/publish.py +394 -0
  13. local_data_platform-0.1.3/src/local_data_platform/dbt_layer/sources.py +366 -0
  14. local_data_platform-0.1.3/src/local_data_platform/etl.py +275 -0
  15. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/events.py +105 -11
  16. local_data_platform-0.1.3/src/local_data_platform/filelock.py +99 -0
  17. local_data_platform-0.1.3/src/local_data_platform/format/excel/__init__.py +284 -0
  18. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/format/iceberg/__init__.py +111 -53
  19. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/format/iceberg/commit.py +118 -10
  20. local_data_platform-0.1.3/src/local_data_platform/integrations/__init__.py +6 -0
  21. local_data_platform-0.1.3/src/local_data_platform/integrations/airflow.py +249 -0
  22. local_data_platform-0.1.3/src/local_data_platform/lake.py +202 -0
  23. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/maintenance/cli.py +60 -13
  24. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/discovery.py +8 -2
  25. local_data_platform-0.1.3/src/local_data_platform/orchestration/__init__.py +73 -0
  26. local_data_platform-0.1.3/src/local_data_platform/orchestration/backfills.py +541 -0
  27. local_data_platform-0.1.3/src/local_data_platform/orchestration/cli.py +282 -0
  28. local_data_platform-0.1.3/src/local_data_platform/orchestration/dbt_task.py +395 -0
  29. local_data_platform-0.1.3/src/local_data_platform/orchestration/ledger.py +164 -0
  30. local_data_platform-0.1.3/src/local_data_platform/orchestration/recipes.py +134 -0
  31. local_data_platform-0.1.3/src/local_data_platform/orchestration/retries.py +182 -0
  32. local_data_platform-0.1.3/src/local_data_platform/orchestration/runner.py +394 -0
  33. local_data_platform-0.1.3/src/local_data_platform/orchestration/schedule.py +681 -0
  34. local_data_platform-0.1.3/src/local_data_platform/orchestration/schedule_cli.py +219 -0
  35. local_data_platform-0.1.3/src/local_data_platform/orchestration/tasks.py +581 -0
  36. local_data_platform-0.1.3/src/local_data_platform/orchestration/windows.py +646 -0
  37. local_data_platform-0.1.3/src/local_data_platform/orchestration/workflow.py +593 -0
  38. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/__init__.py +212 -25
  39. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/builders.py +36 -0
  40. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/builtin.py +10 -3
  41. local_data_platform-0.1.3/src/local_data_platform/pipeline/ingestion/duckdb_to_iceberg/__init__.py +35 -0
  42. local_data_platform-0.1.3/src/local_data_platform/pipeline/ingestion/excel_to_csv/__init__.py +31 -0
  43. local_data_platform-0.1.3/src/local_data_platform/pipeline/ingestion/excel_to_iceberg/__init__.py +31 -0
  44. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/ingestion/pyarrow/__init__.py +3 -2
  45. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/registry.py +8 -2
  46. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/quality/checks.py +6 -3
  47. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/spec.py +515 -38
  48. local_data_platform-0.1.3/src/local_data_platform/store/source/duckdb/__init__.py +155 -0
  49. local_data_platform-0.1.1/PKG-INFO +0 -680
  50. local_data_platform-0.1.1/README.md +0 -632
  51. local_data_platform-0.1.1/src/local_data_platform/etl.py +0 -141
  52. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/LICENSE +0 -0
  53. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/catalog/__init__.py +0 -0
  54. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/catalog/local/__init__.py +0 -0
  55. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/catalog/local/iceberg/__init__.py +0 -0
  56. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/datasets.py +0 -0
  57. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/demo.py +0 -0
  58. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/engine/__init__.py +0 -0
  59. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/engine/duckdb/__init__.py +0 -0
  60. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/engine/duckdb/row_filter.py +0 -0
  61. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/engine/router.py +0 -0
  62. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/engine/spark/__init__.py +0 -0
  63. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/exceptions.py +0 -0
  64. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/format/__init__.py +0 -0
  65. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/format/csv/__init__.py +0 -0
  66. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/format/parquet/__init__.py +0 -0
  67. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/fs.py +0 -0
  68. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/github/__init__.py +0 -0
  69. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/issue/__init__.py +0 -0
  70. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/logger.py +0 -0
  71. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/maintenance/__init__.py +0 -0
  72. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/maintenance/_common.py +0 -0
  73. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/maintenance/orphans.py +0 -0
  74. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/maintenance/snapshots.py +0 -0
  75. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/__init__.py +0 -0
  76. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/__main__.py +0 -0
  77. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/audit.py +0 -0
  78. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/cli.py +0 -0
  79. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/errors.py +0 -0
  80. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/guard.py +0 -0
  81. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/sandbox.py +0 -0
  82. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/server.py +0 -0
  83. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/mcp_server/tools.py +0 -0
  84. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/paths.py +0 -0
  85. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/egression/__init__.py +0 -0
  86. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/egression/csv_to_iceberg/__init__.py +0 -0
  87. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/egression/iceberg_to_csv/__init__.py +0 -0
  88. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/egression/iceberg_to_parquet/__init__.py +0 -0
  89. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/ingestion/__init__.py +0 -0
  90. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/ingestion/bigquery_to_csv/__init__.py +0 -0
  91. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/ingestion/csv_to_iceberg/__init__.py +0 -0
  92. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/pipeline/ingestion/parquet_to_iceberg/__init__.py +0 -0
  93. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/quality/__init__.py +0 -0
  94. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/quality/report.py +0 -0
  95. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/quality/runner.py +0 -0
  96. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/quality/temporal.py +0 -0
  97. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/__init__.py +0 -0
  98. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/source/__init__.py +0 -0
  99. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/source/gcp/__init__.py +0 -0
  100. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/source/gcp/bigquery/__init__.py +0 -0
  101. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/source/json/__init__.py +0 -0
  102. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/source/near/__init__.py +0 -0
  103. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/source/parquet/__init__.py +0 -0
  104. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/target/__init__.py +0 -0
  105. {local_data_platform-0.1.1 → local_data_platform-0.1.3}/src/local_data_platform/store/target/iceberg/__init__.py +0 -0
@@ -0,0 +1,1065 @@
1
+ Metadata-Version: 2.4
2
+ Name: local-data-platform
3
+ Version: 0.1.3
4
+ Summary: A Python API layer for running a complete Iceberg lakehouse data platform on your laptop.
5
+ License: MIT
6
+ License-File: LICENSE
7
+ Author: Tushar Choudhary
8
+ Requires-Python: >=3.12,<4
9
+ Classifier: License :: OSI Approved :: MIT License
10
+ Classifier: Programming Language :: Python :: 3
11
+ Classifier: Programming Language :: Python :: 3.12
12
+ Classifier: Programming Language :: Python :: 3.13
13
+ Classifier: Programming Language :: Python :: 3.14
14
+ Classifier: Programming Language :: Python :: 3.15
15
+ Provides-Extra: bigquery
16
+ Provides-Extra: dbt
17
+ Provides-Extra: dev
18
+ Provides-Extra: docs
19
+ Provides-Extra: duckdb
20
+ Provides-Extra: excel
21
+ Provides-Extra: glue
22
+ Provides-Extra: mcp
23
+ Provides-Extra: s3
24
+ Provides-Extra: spark
25
+ Requires-Dist: boto3 (>=1.34) ; extra == "s3"
26
+ Requires-Dist: build (>=1.2) ; extra == "dev"
27
+ Requires-Dist: cronsim (>=2.6,<3)
28
+ Requires-Dist: dbt-core (>=1.12,<2) ; extra == "dbt"
29
+ Requires-Dist: dbt-duckdb (>=1.11,<2) ; extra == "dbt"
30
+ Requires-Dist: defusedxml (>=0.7) ; extra == "dev"
31
+ Requires-Dist: defusedxml (>=0.7) ; extra == "excel"
32
+ Requires-Dist: duckdb (>=1.1) ; extra == "dbt"
33
+ Requires-Dist: duckdb (>=1.1) ; extra == "dev"
34
+ Requires-Dist: duckdb (>=1.1) ; extra == "duckdb"
35
+ Requires-Dist: duckdb (>=1.1) ; extra == "mcp"
36
+ Requires-Dist: flake8 (>=7) ; extra == "dev"
37
+ Requires-Dist: google-auth (>=2.20) ; extra == "bigquery"
38
+ Requires-Dist: google-cloud-bigquery (>=3.20) ; extra == "bigquery"
39
+ Requires-Dist: mcp (>=2.2,<3) ; extra == "dev"
40
+ Requires-Dist: mcp (>=2.2,<3) ; extra == "mcp"
41
+ Requires-Dist: mkdocs (>=1.6.0) ; extra == "docs"
42
+ Requires-Dist: mkdocs-material (==9.5.21) ; extra == "docs"
43
+ Requires-Dist: mkdocstrings[python] (>=0.25.0) ; extra == "docs"
44
+ Requires-Dist: moto[server] (>=5.0) ; extra == "dev"
45
+ Requires-Dist: openpyxl (>=3.1) ; extra == "dev"
46
+ Requires-Dist: openpyxl (>=3.1) ; extra == "excel"
47
+ Requires-Dist: pyarrow (>=17)
48
+ Requires-Dist: pyiceberg[glue] (>=0.11,<0.13) ; extra == "glue"
49
+ Requires-Dist: pyiceberg[pyarrow,pyiceberg-core,sql-sqlite] (>=0.11,<0.13)
50
+ Requires-Dist: pymdown-extensions (>=10.8.1) ; extra == "docs"
51
+ Requires-Dist: pyspark (>=4.0,<4.2) ; extra == "spark"
52
+ Requires-Dist: pytest (>=8) ; extra == "dev"
53
+ Requires-Dist: requests (>=2.31)
54
+ Project-URL: Documentation, https://local-data-platform.readthedocs.io
55
+ Project-URL: Repository, https://github.com/tusharchou/local-data-platform
56
+ Description-Content-Type: text/markdown
57
+
58
+ # Local Data Platform
59
+
60
+ **local-data-platform** (`ldp`) is a Python library for running an Apache Iceberg lakehouse on your
61
+ laptop, and for taking the same pipelines to a shared catalog and object storage when you need to.
62
+ You describe a dataset in a JSON config, and `ldp` loads it into an Iceberg table, runs data quality
63
+ checks before the write, and lets you query the table with DuckDB SQL and read earlier snapshots.
64
+ The default setup needs no server and no cloud account.
65
+
66
+ It's for learning how a lakehouse works, for building and testing pipelines locally before you pay
67
+ for cloud infrastructure, and for giving agents and notebooks safe, reproducible access to the
68
+ tables you build.
69
+
70
+ > **Status.** This is version 0.1.3, the Orchestration milestone. It adds windows and idempotency keys
71
+ > to `ldp run`, workflow files with `ldp workflow plan` and `ldp workflow run`, `ldp backfill`, retries,
72
+ > `ldp schedule` for cron and launchd, and an Airflow 3 operator (see [Orchestration](#orchestration)),
73
+ > with a [guide](docs/orchestration.md). Version 0.1.2 added `ldp dbt`, which runs a dbt project over
74
+ > your Iceberg tables and publishes the models you mark back to Iceberg (see
75
+ > [Transform with dbt](#transform-with-dbt)), the `DUCKDB` and `XLSX` source formats, time travel and
76
+ > folders of tables in `ldp query`, retries for concurrent appends, and a restaurant example with a
77
+ > [guide](docs/warehousing.md). Version 0.1.1 was the first release on PyPI since 0.1.0. It carries
78
+ > the hardening work (contract in [`docs/design/v0_1_1.md`](docs/design/v0_1_1.md)) and the platform
79
+ > work ("multi-writer, any catalog, object storage, agent-ready", contract in
80
+ > [`docs/design/v0_1_1_platform.md`](docs/design/v0_1_1_platform.md)), which works from Python and
81
+ > through the [platform `ldp` commands](#platform-commands). The version number follows the
82
+ > milestones in the [roadmap](docs/roadmap.md).
83
+
84
+ ## Quickstart
85
+
86
+ `ldp` needs Python 3.12 or newer. Install it from PyPI, preferably in a virtual environment:
87
+
88
+ ```bash
89
+ pip install "local-data-platform[duckdb]>=0.1.3"
90
+
91
+ ldp demo --workdir ldp_demo
92
+ ```
93
+
94
+ To work on the library itself, install it from a clone of this repo. `make install` creates
95
+ `.venv` and installs the package in editable mode with the `dev` and `docs` extras:
96
+
97
+ ```bash
98
+ git clone https://github.com/tusharchou/local-data-platform.git
99
+ cd local-data-platform
100
+ make install
101
+ source .venv/bin/activate
102
+
103
+ ldp demo --workdir ldp_demo
104
+ ```
105
+
106
+ `ldp demo` generates 1,000 synthetic taxi rides (always the same ones), then shows each feature
107
+ using only the public API:
108
+
109
+ 1. It loads a config from JSON.
110
+ 2. It appends twice, so you can see the duplicates, then overwrites twice to show that the load is
111
+ idempotent.
112
+ 3. It upserts on a key column.
113
+ 4. It writes a day-partitioned table.
114
+ 5. It runs quality checks that pass, then shows a bad batch being blocked before the write.
115
+ 6. It queries the table with DuckDB SQL.
116
+ 7. It reads an earlier snapshot (time travel).
117
+ 8. It exports the table back to CSV.
118
+
119
+ Everything it writes goes into the `--workdir` folder, so you can delete it afterwards. The
120
+ [quickstart guide](docs/quickstart.md) explains each step and how to write your own config.
121
+
122
+ More demos, each one a `make` target (see [Development](#development)):
123
+
124
+ | Target | What it shows |
125
+ |---|---|
126
+ | `make demo-robotics` | Humanoid-robot episodes and sensor frames in bronze, silver and gold Iceberg tables, temporal quality checks, a pinned training split and DuckDB research queries |
127
+ | `make demo-agent` | The read-only MCP server on the demo table, driven by a scripted agent client |
128
+ | `make demo-warehouse` | A restaurant chain from Excel and CSV exports to dbt-tested Iceberg tables with monthly customer churn, a late batch and time travel (needs the `dbt` and `excel` extras) |
129
+ | `make demo-orchestration` | The restaurant's nightly drops run as a workflow: a backfill that stops at a bad row and resumes after the fix, a third run that skips every day, and a scheduler tick (runs its dbt models task when the `dbt` extra is installed) |
130
+ | `make demo-spark` | The Scala Spark job reading the demo table and writing an aggregate that pyiceberg reads back (needs scala-cli) |
131
+ | `make demo-rest` | The demo config against an Iceberg REST catalog started locally on scala-cli's JVM: `ldp catalog test`, two runs, the snapshots and a query (needs scala-cli) |
132
+ | `make demo-all` | All of them; `demo-warehouse` is skipped when dbt or the Excel reader isn't installed and the two JVM demos when scala-cli isn't |
133
+
134
+ ### Extras
135
+
136
+ | Extra | Adds | You need it for |
137
+ |---|---|---|
138
+ | `duckdb` | DuckDB | SQL over Iceberg tables (`DuckDBEngine`, `ldp query`, the demo's SQL step) |
139
+ | `mcp` | The MCP SDK and DuckDB | The read-only MCP server for agents (`ldp mcp`) |
140
+ | `s3` | boto3 | SigV4-signed REST catalogs on AWS, such as S3 Tables. `s3://` IO itself needs nothing extra: pyarrow does it |
141
+ | `glue` | `pyiceberg[glue]` (boto3) | The AWS Glue catalog |
142
+ | `bigquery` | Google Cloud BigQuery client | BigQuery as a source |
143
+ | `spark` | PySpark | The experimental `SparkEngine`, which also needs a JDK 17+ |
144
+ | `excel` | openpyxl and defusedxml | Excel workbooks (`.xlsx`) as a source, for `ExcelToCSV` and `ExcelToIceberg` |
145
+ | `dbt` | dbt-core, the dbt-duckdb adapter and DuckDB | `ldp dbt`, which runs a dbt project over your Iceberg tables and publishes the models you mark back to Iceberg ([how it works](docs/design/warehousing.md)), and the tests marked `dbt` |
146
+ | `dev` | pytest, flake8, build, DuckDB, `moto[server]`, the MCP SDK, openpyxl | Working on the library |
147
+ | `docs` | MkDocs and plugins | Building the docs |
148
+
149
+ ## Use it from Python
150
+
151
+ A config names a source, a target and optional quality checks. Paths are relative to the config
152
+ file's folder. Save this as `rides.json`, next to a `data/rides.csv` with `ride_id`, `pickup_ts`
153
+ and `fare` columns:
154
+
155
+ ```json
156
+ {
157
+ "identifier": "rides",
158
+ "metadata": {
159
+ "source": {"name": "rides", "format": "CSV", "path": "data/rides.csv"},
160
+ "target": {
161
+ "name": "rides",
162
+ "format": "ICEBERG",
163
+ "catalog": {"identifier": "nyc", "warehouse_path": "warehouse"},
164
+ "write_mode": "upsert",
165
+ "join_cols": ["ride_id"],
166
+ "partition_by": [{"column": "pickup_ts", "transform": "day"}]
167
+ },
168
+ "quality": {
169
+ "on_failure": "fail",
170
+ "checks": [
171
+ {"check": "not_null", "columns": ["ride_id", "pickup_ts"]},
172
+ {"check": "unique", "columns": ["ride_id"]},
173
+ {"check": "range", "column": "fare", "min": 0}
174
+ ]
175
+ }
176
+ }
177
+ }
178
+ ```
179
+
180
+ Run it, query it and time-travel:
181
+
182
+ ```python
183
+ from local_data_platform import Config
184
+ from local_data_platform.engine.duckdb import DuckDBEngine
185
+ from local_data_platform.format.iceberg import Iceberg
186
+ from local_data_platform.pipeline.registry import create_pipeline
187
+
188
+ config = Config.from_json("rides.json")
189
+ result = create_pipeline(config).run() # picks CSVToIceberg from the config's formats
190
+ print(result.rows_read, result.rows_written, result.quality.passed)
191
+
192
+ table = Iceberg("rides", config.target["catalog"], base_dir=config.base_dir)
193
+
194
+ with DuckDBEngine() as duck:
195
+ duck.register_iceberg(table, "rides")
196
+ print(duck.query("SELECT count(*) AS n, avg(fare) AS avg_fare FROM rides").to_pylist())
197
+
198
+ first = table.snapshots()[0]["snapshot_id"]
199
+ print(table.get(snapshot_id=first).num_rows) # the table as of its first snapshot
200
+ ```
201
+
202
+ If a check fails and `on_failure` is `"fail"`, `run()` raises `DataQualityError` and writes
203
+ nothing; the failing checks are in `error.report`. With `"warn"` it logs them and writes anyway.
204
+
205
+ You can also build a pipeline in code, with your own transforms:
206
+
207
+ ```python
208
+ import pyarrow.compute as pc
209
+
210
+ from local_data_platform.format.csv import CSV
211
+ from local_data_platform.format.iceberg import Iceberg
212
+ from local_data_platform.pipeline import Pipeline
213
+ from local_data_platform.quality import NotNull, Unique
214
+
215
+ pipeline = Pipeline(
216
+ source=CSV("rides", "data/rides.csv"),
217
+ target=Iceberg("rides", {"identifier": "nyc", "warehouse_path": "warehouse"}, write_mode="overwrite"),
218
+ transforms=[lambda df: df.filter(pc.field("fare") >= 0)],
219
+ checks=[NotNull(["ride_id"]), Unique(["ride_id"])],
220
+ )
221
+ print(pipeline.run())
222
+ ```
223
+
224
+ ## Use it from the command line
225
+
226
+ ```bash
227
+ ldp --version
228
+ ldp pipelines # list the registered source/target routes
229
+ ldp run rides.json # run the config's pipeline
230
+ ldp run rides.json --mode overwrite # override the config's write_mode
231
+ ldp run rides.json --window 2026-08-01/2026-08-02 # staged and exactly once: a second run for the window skips
232
+ ldp run orders.json --window yesterday # a window by name: also last-week and last-month, in the config's time zone
233
+ ldp run rides.json --window 2026-08-01/2026-08-02 --run-id RUN_ID --attempt 2 # a retry of that run
234
+ ldp run rides.json --idempotency-key sales-2026-08-01 # a key of your own, for an orchestrator
235
+ ldp run rides.json --window 2026-08-01/2026-08-02 --json # the result as JSON on stdout, for scripts
236
+ ldp snapshots rides.json # list the target table's snapshots
237
+ ldp query rides.json "SELECT count(*) FROM rides" # SQL over the target, named after target.name
238
+ ldp query rides.json "SELECT * FROM rides" --max-rows 20 # print at most 20 rows (default 100)
239
+ ldp query rides.json "SELECT count(*) FROM rides" --snapshot 4437 # an older snapshot, from ldp snapshots
240
+ ldp query rides.json "SELECT count(*) FROM rides" --as-of "2026-07-31 23:59:59" # the newest snapshot at or before that time (UTC)
241
+ ldp query config/ "SELECT o.id FROM raw.orders o JOIN raw.customers c ON c.id = o.id" # a folder of configs: every table, joinable
242
+ ldp demo --workdir /tmp/ldp_demo # also --rows N (default 1000) and --seed N (default 42)
243
+ ```
244
+
245
+ With `--window` or `--idempotency-key` a run is staged and exactly once. It prints its window, its
246
+ idempotency key, its run id and attempt, its status (`published` or `skipped_duplicate`) and the snapshot that holds
247
+ its rows, and running it again for the same window finds that snapshot and writes nothing. `ldp plan CONFIG --window W`
248
+ prints the same key, and `--idempotency-key auto` asks for it. The key is looked for in the whole history that the table
249
+ still holds, before the source is read, so a window that ran weeks ago is skipped without being read too. A retry
250
+ reuses the run id with the next attempt number,
251
+ and the new attempt fences the older ones, so a staged branch that an earlier attempt left behind never reaches `main`.
252
+ `--attempt` needs `--run-id`, and both need a window or a key. Without these options `ldp run` writes directly, as it
253
+ always did. A window is `START/END`, or `yesterday`, `last-week` (the last full Monday-to-Monday week) or `last-month`,
254
+ counted in the time zone of the config, and `ldp run` prints the window it stands for. What a window reads, and how a
255
+ config says so, is in [Windows](#windows).
256
+
257
+ Every command follows one exit-code table.
258
+
259
+ | Exit | Meaning |
260
+ |---|---|
261
+ | 0 | success, including a run that published and a run that skipped a duplicate |
262
+ | 1 | an error: a bad config, a usage error, a failed write |
263
+ | 2 | reserved for an SLA breach that `ldp monitor` reports in 0.1.5 |
264
+ | 3 | a data gate held the run back, such as a quality check that blocked a write |
265
+
266
+ A scheduler retries a run that exits with 1, because the cause may pass, and never retries one that exits with 3,
267
+ because the data has to change first. Add `-v` to see a traceback. The commands
268
+ that read a table (`snapshots`, `query`, `commits`, `maintain`, `datasets pin`, `catalog test`) fail
269
+ without creating anything when the config's catalog is a SQLite file that doesn't exist yet (a
270
+ `local` catalog, or `sql` on SQLite). `ldp run`, `ldp snapshots` and `ldp query` also work with the
271
+ other [catalog types](#catalogs-and-object-storage); they have been run against `sql` (SQLite) and
272
+ `rest` catalogs.
273
+
274
+ ### Platform commands
275
+
276
+ `ldp --help` lists them, and `ldp <command> --help` shows each one's options. Each is also a Python
277
+ call:
278
+
279
+ | Command | What it does | From Python |
280
+ |---|---|---|
281
+ | `ldp catalog test SPEC` | Connect to a catalog (a spec file, or a config's catalog block) and list its namespaces and tables | `catalog.provider.create_catalog(spec).list_namespaces()` |
282
+ | `ldp schema [--kind KIND]` | Print the `ldp/v1` JSON Schema of a kind of file, the dataset config by default | `spec.json_schema()`, `spec.kind_schema(kind)` |
283
+ | `ldp workflow plan FILE [--window W] [--json]` | Check a workflow file, or a dataset config with a schedule, and every config it names, and print its tasks in the order they run with each pipeline task's idempotency key for the window | `orchestration.plan_workflow(path, window=...)` |
284
+ | `ldp backfill FILE --from START [--to END] [--dry-run] [--replace] [--run-backwards] [--max-windows N] [--stop-on-failure] [--json]` | Run a workflow, or a config with a window, for every window of a range, oldest first, each pipeline under the key of its window. A window that is done is skipped without being read, so running it again resumes where it stopped; `--dry-run` lists the windows with their keys and states, and `--replace` writes them again under a replace key; exit status 0, 1, or 3 for a data gate | `orchestration.backfill(path, start, end, ...)` |
285
+ | `ldp workflow run FILE --window W [--json]` | Run the tasks of a workflow one at a time for a window. A failed or blocked task stops the tasks that depend on it and the rest still run, and a second run skips what published or was built; exit status 0, 1, or 3 for a data gate | `orchestration.run_workflow(path, window=...)` |
286
+ | `ldp schedule tick DIR [--now TIME] [--max-windows N] [--json]` | Run every window that is due and not done, for every workflow and config under DIR that has a schedule, then exit. cron or launchd calls it every 5 minutes. A window that failed is carried on by the next tick as the same run with the next attempt, and a tick that starts while another runs does nothing; exit status 0, 1, or 3 for a data gate | `orchestration.schedule.tick(dir, now=...)` |
287
+ | `ldp schedule list DIR [--now TIME] [--max-windows N] [--json]` | List every scheduled workflow and config under DIR with its cron expression, time zone, next fire in UTC and in local time, last window and how it ended, the windows that are due, and the windows it missed with the `ldp backfill` command that fills them | `orchestration.schedule.list_scheduled(dir)` |
288
+ | `ldp schedule install --print {cron,launchd} DIR [--ldp PATH]` | Print the crontab line or the launchd plist that ticks DIR every 5 minutes. It never writes to a system location | `orchestration.recipes.recipe(kind, dir)` |
289
+ | `ldp plan CONFIG [--window WINDOW] [--json]` | Validate a config (or a folder of files, each as its kind) and print its spec hash, route, target, sinks, schedule, and the source path and idempotency key of a window | `spec.validate_spec(data)`, `spec.plan(config, window=...)` |
290
+ | `ldp runs CONFIG [--run RUN_ID] [--events] [--attempts]` | List the runs a config's sinks recorded, the last attempt of a run that was retried, or every attempt with `--attempts` | `events.summarize_runs(events.read_jsonl(path))` or `read_iceberg_events(catalog)` |
291
+ | `ldp commits CONFIG [--key KEY] [--branches]` | List the staged publishes on a table, newest first | `format.iceberg.commit.commit_log(iceberg.table())` and `staged_branches(...)` |
292
+ | `ldp datasets pin\|list\|export` | Pin, list and export dataset versions | `datasets.pin`, `list_versions`, `export` |
293
+ | `ldp maintain CONFIG [--expire DAYS] [--orphans] [--apply]` | Expire snapshots and find orphan files; a dry run unless `--apply` | `maintenance.expire_snapshots`, `find_orphans`, `remove_orphans` |
294
+ | `ldp mcp --config CONFIG ...` | Serve tables to an agent over MCP (stdio) | `python -m local_data_platform.mcp_server ...`, same options |
295
+ | `ldp dbt init\|build\|test\|docs\|exec PROJECT_DIR` | Run a dbt project over your Iceberg tables; `build` writes the models marked `ldp_publish` back to Iceberg (needs the `dbt` extra) | `dbt_layer.register_sources(...)`, `dbt_layer.publish.plan_publish(...)` and `publish(...)` |
296
+ | `ldp spark CONFIG` | Run the Scala Spark job on a config's table (needs scala-cli) | `engine.spark.ScalaSparkJob` |
297
+
298
+ ## Config schema
299
+
300
+ A config has an `identifier`, optional descriptive fields (`who`, `what`, `where`, `when`, `how`)
301
+ and a `metadata` block:
302
+
303
+ | Key | What it holds |
304
+ |---|---|
305
+ | `metadata.source` | `name`, `format` (`CSV`, `PARQUET`, `ICEBERG`, `JSON`, `DUCKDB` or `XLSX`), and a `path` (local, `s3://` or `gs://`, and it may have `{window_start:FORMAT}` placeholders, see [Windows](#windows)) or, for Iceberg, a `catalog`. A `DUCKDB` source also has one of `relation` (a table or view) and `sql` (one `SELECT` query), see [A DuckDB file as a source](#a-duckdb-file-as-a-source). An `XLSX` source may have `sheet` and `header_row`, see [An Excel sheet as a source](#an-excel-sheet-as-a-source) |
306
+ | `metadata.target` | The same, plus for Iceberg `write_mode` (`append`, `overwrite` or `upsert`), `join_cols` for upserts, and `partition_by` |
307
+ | `metadata.target.catalog` | `type` (`local` by default, `sql`, `rest` or `glue`) and that type's keys; see [Catalogs](#catalogs-and-object-storage) |
308
+ | `metadata.quality` | `on_failure` (`fail` or `warn`) and a list of `checks`: `row_count`, `not_null`, `unique`, `accepted_values`, `range`, `freshness`, `schema`, and the temporal checks `monotonic`, `max_skew`, `rate_below` and `max_gap` |
309
+ | `metadata.observability` | Where run events go, for example `{"sinks": ["iceberg", "jsonl"]}`. The default sends them nowhere |
310
+ | `metadata.schedule` | When the config runs: a 5-field cron expression, or `@hourly`, `@daily`, `@weekly` or `@monthly`. Without it, a `when` of `hourly`, `daily`, `weekly` or `monthly` means the alias. Editing it does not change the spec hash |
311
+ | `metadata.window` | `"none"`, or `size` (`day`, `hour` or `15m`), an optional `column` that selects the rows of a window, and an optional `timezone` (UTC by default); see [Windows](#windows) |
312
+ | `metadata.retries` | How a workflow tries the run again after a failure: `max_attempts` (1 by default, which is no retry), `delay_s` (30) and `backoff` (2.0); see [Workflow files](#workflow-files). Editing it does not change the spec hash |
313
+
314
+ A config may also say `"apiVersion": "ldp/v1"` and `"kind": "Pipeline"`. `kind` says what a JSON file is. `Pipeline`
315
+ is the default, so every config here is one, and a config has the same spec hash with and without it. `Workflow` is
316
+ the second kind. A kind is written exactly as shown, so `workflow` fails with a message that names `Workflow`, and
317
+ `Metrics`, `Report` and `Profiles` are reserved for later milestones. `ldp plan` validates each file of a folder as its
318
+ kind, and `ldp query`, `ldp mcp --config` and `ldp dbt` read the tables of the `Pipeline` files only, so a folder can
319
+ hold workflow files next to its configs. `ldp schema --kind Workflow` prints the schema of another kind.
320
+
321
+ Partition transforms are `identity`, `year`, `month`, `day`, `hour`, `bucket[N]` and
322
+ `truncate[W]`. On pyiceberg 0.10 and later, writing with any transform except `identity` needs the
323
+ `pyiceberg-core` package, which `ldp` depends on, so a normal install has it. If it's missing, the
324
+ demo's step 4 says so and partitions by `identity(pickup_date)` instead. The full schema, with an
325
+ example of every check, is in the [v0.1.1 hardening contract](docs/design/v0_1_1.md#config-schema).
326
+ `local_data_platform.spec.json_schema()` returns it as JSON Schema, and `ldp schema` prints it. The configs in
327
+ [`examples/`](examples/README.md) are real ones you can run.
328
+
329
+ ### Windows
330
+
331
+ A run can be responsible for one window of time, `[start, end)`, which is what `--window` gives it. A config says how a
332
+ window picks what the run reads, and when the config runs:
333
+
334
+ ```json
335
+ {
336
+ "identifier": "daily_orders",
337
+ "when": "daily",
338
+ "metadata": {
339
+ "schedule": "5 2 * * *",
340
+ "window": {"size": "day", "column": "ordered_at", "timezone": "Europe/Paris"},
341
+ "source": {"name": "orders", "format": "CSV", "path": "drops/orders_{window_start:%Y-%m-%d}.csv"},
342
+ "target": {"name": "orders", "format": "ICEBERG", "write_mode": "overwrite",
343
+ "catalog": {"identifier": "shop", "warehouse_path": "warehouse"}}
344
+ }
345
+ }
346
+ ```
347
+
348
+ - `{window_start:FORMAT}` and `{window_end:FORMAT}` in `metadata.source.path` name the file of the window, with a Python
349
+ `strftime` FORMAT shown in the time zone of the window. A run without a window on such a path fails and names the
350
+ placeholder.
351
+ - `window.column` keeps only the rows at or after the start of the window and before its end, before the transforms and
352
+ the quality checks. `run.extracted` reports the rows read and the rows kept.
353
+ - With `window.column` and the write mode `overwrite`, a run replaces only the rows of its window, so one changed day can
354
+ be redone and the other days stay. A published window is skipped when it runs again, so to redo a day that changed at
355
+ the source, replace it with `ldp backfill daily_orders.json --from 2026-08-02 --to 2026-08-03 --replace`, or run it with a
356
+ key of your own: `ldp run daily_orders.json --window 2026-08-02/2026-08-03 --idempotency-key orders-2026-08-02-v2`.
357
+ - A `freshness` check in a windowed run measures the age of the data at the end of the window.
358
+ - A config with a schedule that appends, and that has neither a placeholder nor a `window.column`, would load the same
359
+ rows in every window. `ldp plan` reports that as an error and `ldp run --window` logs a warning.
360
+
361
+ The [contract](docs/design/v0_1_1.md#windows-and-schedules-013) has the full rules, including how each column type is
362
+ compared.
363
+
364
+ ### Workflow files
365
+
366
+ A workflow file lists the pipelines, dbt models and maintenance that run for each window, and in what order:
367
+
368
+ ```json
369
+ {
370
+ "apiVersion": "ldp/v1",
371
+ "kind": "Workflow",
372
+ "identifier": "daily_orders",
373
+ "metadata": {"schedule": "5 2 * * *", "window": {"size": "day"}, "retries": {"max_attempts": 3, "delay_s": 30}},
374
+ "tasks": [
375
+ {"id": "orders", "pipeline": "config/orders.json"},
376
+ {"id": "customers", "pipeline": "config/customers.json"},
377
+ {"id": "gold", "dbt": {"project": "dbt", "select": "+daily_revenue"}, "depends_on": ["orders", "customers"]},
378
+ {"id": "export", "pipeline": "config/export.json", "depends_on": ["gold"]},
379
+ {"id": "tidy", "maintain": {"config": "config/orders.json", "expire": 7}, "depends_on": ["export"]}
380
+ ]
381
+ }
382
+ ```
383
+
384
+ ```bash
385
+ ldp workflow plan daily_orders.json --window 2026-08-01/2026-08-02 # the tasks in run order, with each key
386
+ ldp workflow run daily_orders.json --window 2026-08-01/2026-08-02 # run them for that day
387
+ ldp runs .ldp/events.jsonl # the workflow run and the run of each task
388
+ ```
389
+
390
+ - Each task has an `id`, an optional `depends_on`, and exactly one kind: `pipeline` (the path of a dataset config,
391
+ relative to the workflow file), `dbt` (a dbt project, built for the window, see below) or `maintain` (a config's Iceberg
392
+ table and the options of `ldp maintain`, a dry run unless `apply` is true). Later milestones add kinds without changing the
393
+ file format.
394
+ - Tasks run one at a time, in dependency order. A pipeline task runs with the window of the workflow, so its idempotency key
395
+ is the one `ldp plan CONFIG --window` prints.
396
+ - A task that fails, or that a quality check blocks, makes the tasks that depend on it `upstream_failed`, and they do not run.
397
+ The tasks that do not depend on it still run. The exit status is 1 when a task failed and 3 when the only problem is a
398
+ data gate.
399
+ - Run the same window again after fixing the cause: the pipeline tasks that published are skipped by their keys, without
400
+ reading their sources, and only the tasks that failed or did not run do their work.
401
+ - A task that fails with an error that another attempt may fix runs again by itself, as `retries` says, in the workflow
402
+ or in the task, and a key of the task overrides the same key of the workflow. These are retried: a dropped connection,
403
+ a timeout, a file that has not landed yet (any `OSError`) and the commit errors of pyiceberg. A data gate, a wrong
404
+ config and a fenced attempt never are. A retry is the next attempt of the same run, with the same run id and the same
405
+ idempotency key, so the new attempt fences the old one and a load that already landed is found and not written
406
+ twice. The wait is a random time from 0 to `delay_s * backoff ** (n - 1)` seconds after attempt n. `ldp runs` lists the
407
+ last attempt of each run, and `ldp runs LEDGER --attempts` lists them all.
408
+ - A `dbt` task does what [`ldp dbt build`](#transform-with-dbt) does, for the window: dbt gets `window_start` and
409
+ `window_end` as variables (`'{{ var("window_start") }}'::timestamp` in a model), and the models marked `ldp_publish` are
410
+ written to Iceberg. Failing data tests are a data gate, so the task is `blocked_quality` and the exit status is 3. Run the
411
+ same window again and the task is skipped when its last build succeeded with the same project and no task before it may
412
+ have written data. List the pipelines that load the tables a dbt project reads in its `depends_on`, and name the dbt task
413
+ in the `depends_on` of every task that reads the tables it publishes.
414
+ - Every event of a run goes to `.ldp/events.jsonl` next to the workflow file, with the `workflow`, `workflow_run_id` and
415
+ `task_id` of the task in its payload. A dataset config that has a `metadata.schedule` runs as a workflow of one task.
416
+
417
+ The [contract](docs/design/v0_1_1.md#workflow-files-013) has the full file format. `ldp schema --kind Workflow` prints its JSON
418
+ Schema.
419
+
420
+ ### Backfill a range of windows
421
+
422
+ `ldp backfill` runs a workflow, or a config that has a window, for every window of a range. It is how a new pipeline loads
423
+ its history, and how a load that stopped halfway finishes:
424
+
425
+ ```bash
426
+ ldp backfill daily_orders.json --from 2026-06-01 --to 2026-09-01 --dry-run # every window, with its key and its state
427
+ ldp backfill daily_orders.json --from 2026-06-01 --to 2026-09-01 # load them, oldest first
428
+ ldp backfill daily_orders.json --from 2026-06-01 --to 2026-09-01 # again: it skips every window that is done
429
+ ldp backfill daily_orders.json --from 2026-08-02 --to 2026-08-03 --replace # redo a day that changed at the source
430
+ ```
431
+
432
+ - The windows are the windows of `metadata.window.size` (`day`, `hour` or `15m`), or of the size the schedule means, a day for
433
+ `@daily` and an hour for `@hourly`. `--from` and `--to` are dates or datetimes in the time zone of the workflow. Each of them
434
+ has to start a window, and the window that starts at `--to` is not part of the range, so `--from 2026-06-01 --to
435
+ 2026-06-04` is the three days of June 1st, 2nd and 3rd. Without `--to` the range ends with the last complete window. A
436
+ range of more than `--max-windows` windows, 1000 by default, is refused, to catch a typo.
437
+ - Each window runs as `ldp workflow run` runs it, so each pipeline publishes under the key that `ldp plan CONFIG --window W`
438
+ prints, and a task with `retries` is retried. A window whose keys are on the table is skipped without being read, planned
439
+ or written to the ledger, so a finished backfill that runs again does nothing, and one that stopped at window 40 starts
440
+ again at window 40. A workflow with a task that a key cannot vouch for, such as a dbt task or a pipeline that writes a file,
441
+ also needs the last run of the window in the ledger to have succeeded for every task.
442
+ - A window that fails, or that a quality check blocks, does not stop the others, and it is listed with what went wrong,
443
+ such as the check that failed. The summary counts the windows that published, were skipped, were blocked and failed.
444
+ The exit status is 0 when no window failed or was blocked, 3 when a data gate is the only problem, and 1 otherwise.
445
+ `--stop-on-failure` stops at the first window that failed instead, and a window that a quality check blocked does not
446
+ stop it.
447
+ - `--dry-run` prints every window with its key and its state, which is `published` (done, so it is skipped), `partial`,
448
+ `missing`, `failed` or `blocked` (the last run of it in the ledger failed or was held back by a data gate), and runs
449
+ nothing, not even the creation of a table. `--run-backwards` runs the newest window first, and `--json` prints the result.
450
+ - `--replace` writes each window again under a replace key, the sha256 of its key, the word `replace` and the spec hash of
451
+ its config, with a window overwrite. The rows that a window has are replaced by the rows the source has now, a window that
452
+ has none is loaded, a window that was replaced under the same spec is skipped, and one that was replaced under an older
453
+ spec is replaced again. It needs a `metadata.window.column` in every pipeline that writes to Iceberg, and without one it
454
+ exits with 1 and changes nothing. A window that is replaced has to have rows in the source. An empty one is not written,
455
+ because the run fails with "No data to write", or is blocked by a `row_count` check, and the table keeps its rows.
456
+ - From Python, `from local_data_platform.orchestration import backfill` runs the same, and `backfill(path, start, end,
457
+ dry_run=True)` returns the windows with their keys and states. The [contract](docs/design/v0_1_1.md#backfills-013) has the
458
+ rest.
459
+
460
+ ### Run on a schedule
461
+
462
+ `ldp schedule tick` is a scheduler with no server and no daemon. cron or launchd calls it every 5 minutes, it runs every
463
+ window that is due and not done, and it exits, so a laptop that was asleep catches up at the first tick after it wakes:
464
+
465
+ ```bash
466
+ ldp schedule list ldp_demo/restaurant # what is scheduled, when it fires, what it missed
467
+ ldp schedule tick ldp_demo/restaurant --now 2026-08-05T02:05:00Z # run what is due at that time, then exit
468
+ ldp schedule install --print launchd ldp_demo/restaurant > ~/Library/LaunchAgents/restaurant.plist # macOS
469
+ (crontab -l 2>/dev/null; ldp schedule install --print cron ldp_demo/restaurant) | crontab - # Linux
470
+ ```
471
+
472
+ - A tick looks at every `.json` file under the folder that is a workflow file, or a dataset config, and has a schedule in
473
+ `metadata.schedule` or in a `when` such as `daily`. A config with a schedule is a workflow of one task. A file that has a
474
+ schedule and is not valid is printed with its problems and makes the tick exit with 1, and the other files still run.
475
+ - The schedule is a cron expression read in `metadata.window.timezone`, UTC without one. The window a fire owns is the
476
+ interval between the previous fire and this fire. When the workflow has a window size, in `metadata.window.size` or implied
477
+ by `@daily` and `@hourly`, it is the last complete window of that size at or before the fire, so the fire of `5 2 * * *` on
478
+ 5 August owns 4 August. A fixed time that a clock change skips fires right after the gap, and one that happens twice fires
479
+ once.
480
+ - Without `catchup`, a tick runs the window of the latest fire if it is not done, and `ldp schedule list` prints the
481
+ `ldp backfill` command for any windows before it that were missed. With `"catchup": true` and a `"start"` in the metadata of
482
+ the workflow, a tick runs every window since `start` that is not done, oldest first, up to `--max-windows` (1000) a tick.
483
+ - A window that failed, was blocked by a data gate, or never finished is run again by the next tick, as the next attempt of
484
+ the same run. The workflow and each task keep their run ids and the attempts go up by one, so the first commit of the new
485
+ attempt fences every older one and the window is loaded once however many ticks it takes. `retries` are the retries within
486
+ one tick.
487
+ - A tick takes the lock `DIR/.ldp/locks/schedule.lock` without waiting for it. A tick that starts while another runs does
488
+ nothing and exits with 0, and a crash leaves no lock. The exit status is 0 when nothing failed or was blocked, 3 when a data
489
+ gate is the only problem, and 1 otherwise.
490
+ - The cron line and the plist call `ldp` by its absolute path and write a log, `DIR/.ldp/schedule.log` or a file in
491
+ `~/Library/Logs`. cron does not run while a Mac sleeps, and launchd runs an interval that passed during sleep once when the
492
+ Mac wakes, so launchd is the recipe for macOS and cron the recipe for Linux.
493
+ - From Python, `from local_data_platform.orchestration.schedule import tick, list_scheduled` does the same. The
494
+ [contract](docs/design/v0_1_1.md#schedules-013) has the rest.
495
+
496
+ ### Run from Airflow
497
+
498
+ A team that already runs Apache Airflow 3 can run a config as a task. `LdpRunOperator` takes the data interval of the DAG run
499
+ as the window, `ti.try_number` as the attempt, and a run id made of the DAG id, the run id of the DAG run and the task id. So
500
+ a retry by Airflow is the next attempt of the same run, a backfill over a day that is in the table finds the key and writes
501
+ nothing, and a day is loaded once however Airflow got to it. The [guide](docs/airflow.md) installs Airflow, runs the daily
502
+ drops of the restaurant example from `examples/restaurant/airflow_dag.py`, and retries a late drop. Its commands, in short, are these.
503
+
504
+ ```bash
505
+ export LDP_AIRFLOW_DEMO="$PWD/ldp_demo/airflow/data"
506
+ python examples/restaurant/airflow_dag.py --prepare "$LDP_AIRFLOW_DEMO" # the seven daily drops and a windowed orders config
507
+ airflow standalone # terminal 1, with the DAG file in its dags folder
508
+ airflow dags unpause restaurant_orders # terminal 2, catches up seven days one after the other
509
+ ldp commits "$LDP_AIRFLOW_DEMO/config/orders.json" # terminal 2, one publish for each day
510
+ ```
511
+
512
+ - The window is `(data_interval_start, data_interval_end)`, so the key of an interval is the one
513
+ `ldp plan CONFIG --window START/END` prints for it.
514
+ - A data gate that holds a run back, and a config that cannot run, fail the task with `AirflowFailException`, so Airflow does
515
+ not retry them. A skipped duplicate succeeds, and the `PipelineResult` as a dict is the XCom of the task. Any other error is
516
+ left to the `retries` of the task.
517
+ - Give the DAG a `CronDataIntervalTimetable("@daily", timezone="UTC")` as its schedule and not the string `"@daily"`. Airflow 3
518
+ makes a trigger timetable for a cron string, whose runs have an empty data interval, and the operator fails them and says why.
519
+ - The mapping is plain Python and Airflow is imported when the operator is used, so `local_data_platform.integrations.airflow`
520
+ imports without it. There is no extra for it, because Airflow has to be installed with its own constraints file. Airflow 3.1.7
521
+ and later install next to LDP, and a team that cannot do that runs `ldp run CONFIG --window ... --run-id ... --attempt ...`
522
+ from a BashOperator, which the guide shows. The [contract](docs/design/v0_1_1.md#airflow-013) has the rest.
523
+
524
+ ### A DuckDB file as a source
525
+
526
+ A `DUCKDB` source reads one table, or the result of one query, from a DuckDB database file and
527
+ publishes it to Iceberg like any other source, with the same quality checks, run events and
528
+ snapshots. It is how tables built with SQL (or with dbt, through `ldp dbt`) reach the lakehouse:
529
+
530
+ ```json
531
+ {
532
+ "identifier": "churn_by_month",
533
+ "metadata": {
534
+ "source": {"name": "churn_by_month", "format": "DUCKDB", "path": "target/build.duckdb",
535
+ "relation": "gold.churn_by_month"},
536
+ "target": {"name": "churn_by_month", "format": "ICEBERG", "write_mode": "overwrite",
537
+ "catalog": {"identifier": "gold", "warehouse_path": "warehouse"}}
538
+ }
539
+ }
540
+ ```
541
+
542
+ Use `relation` (`table`, `schema.table` or `catalog.schema.table`) or `sql`, not both. The file is
543
+ opened read-only and closed before the Iceberg write starts, and a `sql` source must be exactly one
544
+ `SELECT` or `WITH` query. It needs the `duckdb` extra. `ldp plan` reports a missing file or a
545
+ source with both or neither of `relation` and `sql`.
546
+
547
+ ### An Excel sheet as a source
548
+
549
+ An `XLSX` source reads one sheet of an `.xlsx` workbook and writes it to a CSV file (`ExcelToCSV`) or an
550
+ Iceberg table (`ExcelToIceberg`), so a spreadsheet that someone downloads from a website stops being converted
551
+ by hand:
552
+
553
+ ```json
554
+ {
555
+ "identifier": "sales",
556
+ "metadata": {
557
+ "source": {"name": "sales", "format": "XLSX", "path": "downloads/sales.xlsx", "sheet": "Sales",
558
+ "header_row": 1},
559
+ "target": {"name": "sales", "format": "CSV", "path": "drive/sales.csv"}
560
+ }
561
+ }
562
+ ```
563
+
564
+ `sheet` is the sheet's name or its position from 0, and the first sheet by default. `header_row` is the row
565
+ that holds the column names, counted from 1, and the rows above it are skipped. Dates and datetimes become
566
+ timestamps, whole numbers `int64`, other numbers `float64`, TRUE and FALSE `bool`, and text `string`. A column
567
+ that mixes kinds becomes `string`, and a warning names it. A formula cell gives the value Excel saved with it,
568
+ and empty rows at the end of the sheet and empty columns at its right edge are dropped. Formatting, charts and
569
+ macros are ignored, and a merged range gives its value to its first cell and nulls to the rest. A `.xls` file, a
570
+ sheet that does not exist, and an empty or repeated header cell fail with a message that names the file and the
571
+ sheet. It needs the `excel` extra, and `ldp plan` reports a missing file or a `.xls` path. The CSV is replaced
572
+ atomically, so a folder that Google Drive syncs never holds a half-written file; see the recipe
573
+ [Excel to CSV in a Google Drive folder](docs/recipes.md#excel-to-csv-in-a-google-drive-folder).
574
+
575
+ ## Catalogs and object storage
576
+
577
+ The catalog block's `type` picks the catalog. It defaults to `local`, so a catalog block without
578
+ `type` keeps working unchanged.
579
+
580
+ | `type` | Keys | What you get |
581
+ |---|---|---|
582
+ | `local` (alias `LocalIceberg`) | `identifier`, `warehouse_path` | A SQLite catalog file and a warehouse folder |
583
+ | `sql` (alias `sqlite`) | `uri` (a SQLAlchemy URI such as `sqlite:///…` or `postgresql+psycopg://…`), `warehouse`, `name`, `password_env` | pyiceberg's `SqlCatalog`. Its tables are the ones Iceberg's Java `JdbcCatalog` uses, so Spark can share it |
584
+ | `rest` | `uri`, `warehouse`, `name`, `token_env`, `credential_env`, and optional `properties` (such as `s3.endpoint`) | pyiceberg's `RestCatalog`, for any Iceberg REST catalog, such as Polaris, Nessie or Lakekeeper |
585
+ | `glue` | `name`, `warehouse`, and optional `properties` | pyiceberg's `GlueCatalog` (the `glue` extra). AWS credentials come from boto3's default chain |
586
+
587
+ What has been tested: `local` and `sql` on SQLite; `rest` against Apache Iceberg's REST catalog
588
+ test server (`tools/rest_fixture`, opt-in); `glue` against a local moto server. Postgres and
589
+ hosted REST catalogs (Polaris, S3 Tables, Unity and so on) are not tested yet.
590
+
591
+ The namespace tables live in comes from `identifier` for `local` catalogs, and from `namespace`,
592
+ falling back to `identifier`, for every other type.
593
+ `properties` are passed through to pyiceberg, and `properties_env` maps a property to the
594
+ environment variable holding its value. A REST target looks like this:
595
+
596
+ ```json
597
+ "catalog": {
598
+ "type": "rest",
599
+ "name": "lake",
600
+ "uri": "https://catalog.example.com/api/catalog",
601
+ "warehouse": "analytics",
602
+ "namespace": "rides",
603
+ "token_env": "LAKE_TOKEN",
604
+ "properties": {"s3.endpoint": "http://127.0.0.1:9000"}
605
+ }
606
+ ```
607
+
608
+ **Secrets never go in configs.** A config names the environment variable that holds a token,
609
+ credential or password (`token_env`, `credential_env`, `password_env`, `properties_env`). A key or
610
+ property whose name contains `token`, `secret`, `password` or `credential` and holds a literal value
611
+ is rejected, and so is a password inside a `uri`; log lines mask them. A `postgresql+psycopg://`
612
+ URI also needs a SQLAlchemy driver, such as `pip install "psycopg[binary]"`.
613
+
614
+ `create_catalog` builds the same catalog a config would. For a read-only check before running
615
+ anything, use `ldp catalog test SPEC`; calling `create_catalog` on an `sql` spec creates the SQLite
616
+ file if it does not exist yet:
617
+
618
+ ```python
619
+ from local_data_platform.catalog.provider import catalog_namespace, create_catalog
620
+
621
+ spec = {"type": "sql", "name": "lake", "uri": "sqlite:////tmp/lake/catalog.db", "warehouse": "file:///tmp/lake",
622
+ "namespace": "rides"}
623
+ catalog = create_catalog(spec)
624
+ print(catalog_namespace(spec), catalog.list_namespaces())
625
+ ```
626
+
627
+ `register_catalog_type(name)` is a decorator that adds your own type. `Iceberg(...,
628
+ catalog_obj=catalog)` writes through a catalog object you already have.
629
+
630
+ **Object storage.** CSV, Parquet and JSON read and write through `local_data_platform.fs`, so their
631
+ paths can be local, `file://`, `s3://` or `gs://`, all through `pyarrow.fs`. Settings come from the
632
+ environment, never from a config: S3 credentials from the AWS default chain, `AWS_ENDPOINT_URL_S3`
633
+ or `AWS_ENDPOINT_URL` for MinIO, LocalStack or moto, and `AWS_REGION`; `STORAGE_EMULATOR_HOST` for
634
+ a GCS emulator. A URI with credentials in it (`s3://key:secret@bucket/...`) is rejected. Writes are
635
+ atomic: locally a temp file is renamed into place, and on an object store the data is uploaded only
636
+ once it is complete, so readers see the old object or the new one. The S3 tests run against a local
637
+ moto server; `gs://` reads and writes have no test against a GCS emulator yet.
638
+
639
+ ## Exactly-once writes
640
+
641
+ A plain `Iceberg.put(df, mode)` is **direct mode**. It keeps the semantics of the
642
+ [hardening contract](docs/design/v0_1_1.md) (appending twice gives duplicates, overwriting twice
643
+ doesn't) with three fixes:
644
+
645
+ - **One commit per write.** Adding new columns and writing the data happen in one transaction.
646
+ - **Counts from metadata.** `rows_before` and `rows_after` come from the snapshots' `total-records`,
647
+ not from a scan.
648
+ - **A lock on local catalogs.** `overwrite` and `upsert` take an exclusive file lock,
649
+ `<warehouse>/.ldp/locks/<namespace>.<table>.lock`, so two overwrites or upserts on one laptop
650
+ can't interleave. Appends take no lock. An append that loses a commit race reloads the table
651
+ and runs again with a backoff (`CommitPolicy` sets the limits), so several processes can append
652
+ to one table at once.
653
+
654
+ Direct `overwrite` and `upsert` on a shared remote catalog are single-writer. For several
655
+ writers, pass a `CommitContext` and `put` switches to the **staged publish protocol**: it writes
656
+ to a private branch, checks the staged snapshot, then fast-forwards `main` in one catalog
657
+ compare-and-swap that asserts both `main` and the branch. The idempotency key is stored on the
658
+ snapshot as `ldp.idempotency-key` (with `ldp.run-id` and `ldp.attempt`), so re-running a key finds
659
+ the published snapshot and writes nothing:
660
+
661
+ ```python
662
+ import pyarrow as pa
663
+
664
+ from local_data_platform.format.iceberg import Iceberg
665
+ from local_data_platform.format.iceberg.commit import CommitContext
666
+
667
+ table = Iceberg("fares", {"identifier": "nyc", "warehouse_path": "warehouse"}, write_mode="upsert",
668
+ join_cols=["ride_id"])
669
+ batch = pa.table({"ride_id": [1, 2, 3], "fare": [12.5, 8.0, 30.25]})
670
+ context = CommitContext.create("fares/2024-01-01") # a new run id; looks for the key 7 days back
671
+
672
+ first = table.put(batch, commit=context)
673
+ again = table.put(batch, commit=context) # the same key: finds the published snapshot
674
+ print(first.attempts, again.skipped_duplicate) # 1 True
675
+ ```
676
+
677
+ If another writer moves `main` first, the attempt rebases and retries within its `CommitPolicy`; an
678
+ upsert is recomputed against the new rows, never blindly rebased. A superseded attempt is fenced by
679
+ removing its branch, so it can't publish afterwards. `WriteResult` gains `branch`,
680
+ `idempotency_key`, `attempts` and `skipped_duplicate`. What this guarantees is at most one effect
681
+ on `main` per key, which with retries means effectively once. It doesn't remove duplicates a
682
+ producer sent twice upstream. `ldp commits
683
+ CONFIG` lists the publishes on a table with their keys.
684
+ [Exactly-once writes](docs/exactly_once.md) has the protocol and its tests, including the
685
+ multi-process upsert race.
686
+
687
+ ## Run events and the `_ldp` namespace
688
+
689
+ `Pipeline.run(mode=None, *, commit=None, sink=None)` emits a `RunEvent` at each step:
690
+ `run.started`, `run.extracted`, `quality.evaluated`, then one of `run.published`,
691
+ `run.skipped_duplicate`, `run.blocked_quality` or `run.failed`, and finally `run.finished`.
692
+ `PipelineResult` gains `run_id` (a uuid7), `idempotency_key` and `published_snapshot_id`.
693
+
694
+ | Sink | Writes |
695
+ |---|---|
696
+ | `NullSink` | Nothing (the default, so a run emits no events unless you choose a sink) |
697
+ | `JsonlSink(path)` | One JSON event per line |
698
+ | `OpenLineageSink(path_or_url)` | OpenLineage 1.x `RunEvent`s with schema, data-quality and output-statistics facets |
699
+ | `IcebergSink(catalog_spec, base_dir)` | Batched appends to `_ldp.runs` and `_ldp.quality_results`, both day-partitioned, in your own catalog (the MCP server also uses it for `_ldp.audit`) |
700
+ | `MultiSink` | Several of the above |
701
+
702
+ ```python
703
+ from local_data_platform import Config
704
+ from local_data_platform.events import JsonlSink
705
+ from local_data_platform.pipeline.registry import create_pipeline
706
+
707
+ result = create_pipeline(Config.from_json("rides.json")).run(sink=JsonlSink("runs.jsonl"))
708
+ print(result.run_id, result.published_snapshot_id)
709
+ ```
710
+
711
+ Or set the default sinks in the config with `metadata.observability`, for example
712
+ `{"sinks": ["iceberg", "jsonl"]}`; by default the `jsonl` sink writes `.ldp/events.jsonl` in the
713
+ config's folder, and `ldp run` uses these sinks too. Because `_ldp.runs` and
714
+ `_ldp.quality_results` are ordinary Iceberg tables in your catalog, you can query your run history
715
+ with any engine; `ldp runs CONFIG` lists it, and `events.read_iceberg_events` and `summarize_runs`
716
+ read it from Python. `local_data_platform.spec` gives each config a stable `spec_hash` and an
717
+ `idempotency_key(config, window)`, and validates configs against the `ldp/v1` JSON Schema
718
+ (`ldp plan CONFIG` shows all three). See
719
+ [Run events and the `_ldp` namespace](docs/observability.md).
720
+
721
+ ## SQL with DuckDB
722
+
723
+ `DuckDBEngine.register_iceberg(table, alias, snapshot_id=None, row_filter=None, native=None)`
724
+ uses DuckDB's `iceberg` extension when it can: the table becomes a view over `iceberg_scan` at the
725
+ chosen snapshot, which streams and pushes filters down instead of loading the table into memory.
726
+ If the extension can't load, it falls back to the in-memory scan with a warning. It never
727
+ downloads the extension on its own, so it works offline; `DuckDBEngine(install_extensions=True)` allows
728
+ that. `native=True` or `native=False` forces one path.
729
+
730
+ ```python
731
+ from local_data_platform.engine.duckdb import DuckDBEngine
732
+ from local_data_platform.format.iceberg import Iceberg
733
+
734
+ table = Iceberg("rides", {"identifier": "nyc", "warehouse_path": "warehouse"})
735
+ with DuckDBEngine() as duck:
736
+ duck.register_iceberg(table, "rides") # native iceberg_scan when available
737
+ duck.register_iceberg(table, "rides_in_memory", native=False)
738
+ print(duck.query("SELECT count(*) FROM rides").to_pylist())
739
+ print(duck.snapshots("rides")) # from iceberg_snapshots
740
+ ```
741
+
742
+ `open_lake(path)` registers every Iceberg table that some dataset configs name, so SQL can join them. `path` is a
743
+ config file, a folder of configs, or a list of them, and `ldp query` uses it when it is given a folder or more than
744
+ one config. Each table is a view named `namespace.table` and, when no other table has the same name, a view with the
745
+ bare name. A namespace of several parts, such as `a.b`, is one schema named `a.b`, so SQL quotes it: `"a.b".events`.
746
+ `snapshot_id=` or `as_of=` read one table as it was, and refuse a lake of several. Nothing is written and
747
+ nothing is downloaded, and the MCP SDK is not needed.
748
+
749
+ ```python
750
+ from local_data_platform.lake import open_lake
751
+
752
+ with open_lake("config") as lake:
753
+ lake.query("SELECT o.id, c.label FROM raw.orders o JOIN raw.customers c ON c.id = o.id")
754
+ with open_lake("config/raw.orders.json", as_of="2026-07-31 23:59:59") as lake: # UTC unless it has an offset
755
+ lake.query("SELECT count(*) FROM orders")
756
+ ```
757
+
758
+ `attach_rest(name, uri, warehouse, token_env=None)` attaches a whole REST catalog to DuckDB.
759
+ `engine.router` estimates how many bytes a scan reads from the manifests, after partition pruning,
760
+ and `choose_engine` picks DuckDB below 50 GiB by default, and PySpark above that when it's
761
+ installed. No pipeline calls the router yet.
762
+
763
+ ## Transform with dbt
764
+
765
+ `ldp dbt` runs a [dbt](https://www.getdbt.com/) project over your Iceberg tables and writes the models you mark back
766
+ to Iceberg, so what dbt builds is versioned, can be read as it was, and is there for `ldp query`, `ldp runs` and
767
+ `ldp mcp` like any other table. It needs the `dbt` extra, and the `excel` extra when a source is an Excel sheet.
768
+
769
+ ```bash
770
+ pip install "local-data-platform[dbt,excel]"
771
+ ldp dbt init my_project --sources config --warehouse warehouse # a project that builds as soon as its sources exist
772
+ ldp dbt build my_project # register the sources, run dbt, then publish the models marked ldp_publish
773
+ ldp dbt test my_project # register the sources and run the data tests, and publish nothing
774
+ ldp dbt docs my_project --serve # generate the dbt docs and serve them on http://127.0.0.1:8080
775
+ ldp dbt exec my_project compile # any other dbt command, with the sources registered
776
+ ```
777
+
778
+ Before dbt runs, every Iceberg table that the configs in `--sources` describe becomes a source in a DuckDB file inside
779
+ the project, so a model reads it with `{{ source('raw', 'orders') }}`. After a build in which every model and every
780
+ data test passed, each model whose config has `meta={'ldp_publish': true}` is written to Iceberg through the
781
+ [`DUCKDB` source](#a-duckdb-file-as-a-source), with the same quality checks, run events and snapshots as any other load.
782
+ A table or view model overwrites its table, and an incremental model with a `unique_key` upserts on it. Nothing is
783
+ published when dbt fails. Everything after the project folder goes to dbt unchanged, and the options of `ldp dbt`
784
+ itself go before it.
785
+
786
+ - The [warehousing guide](docs/warehousing.md) walks through it on a restaurant chain's Excel and CSV exports, from
787
+ the raw tables to tested models, churn, an incremental load and time travel. `make demo-warehouse` runs the same
788
+ example offline in about 25 seconds.
789
+ - [How `ldp dbt` works](docs/design/warehousing.md) has the contract, including the meta keys and the schema rules.
790
+
791
+ ## Orchestration
792
+
793
+ A workflow file declares the pipelines, dbt models and maintenance of a night once. LDP runs it for the right window, backfills a range of windows, retries what failed, and runs it on a schedule from cron or launchd, with no service. Every run goes by the idempotency key of its window, so a window is never loaded twice, whoever starts it.
794
+
795
+ ```bash
796
+ make demo-orchestration # the restaurant's nightly drops, a bad row, a backfill, a fix and a scheduler tick
797
+ ```
798
+
799
+ - The [orchestration guide](docs/orchestration.md) walks through it on the restaurant example, command by command, from a window and its key to a workflow file, a backfill that stops at a bad row, a replaced day, a retried task, launchd and cron, and Airflow, Dagster and Prefect.
800
+ - `make demo-orchestration` runs `examples/restaurant/run.py --layer orchestration`. It drops three nights of orders into a folder, with a bad row in the third. `ldp backfill` publishes two days and a data gate holds the third back, with exit status 3. After the fix, the same command runs only the third, and a third run skips every day. `ldp schedule tick` then loads the fourth night, and `ldp runs`, `ldp commits`, `ldp schedule list` and `ldp schedule install` show what happened. It needs no extra, runs the `models` task when the `dbt` extra is installed, and prints a skip line when it is not.
801
+ - The pieces are described above, under [Windows](#windows), [Workflow files](#workflow-files), [Backfill a range of windows](#backfill-a-range-of-windows), [Run on a schedule](#run-on-a-schedule) and [Run from Airflow](#run-from-airflow), and the exit statuses are in the [table](#use-it-from-the-command-line) of the command line section.
802
+
803
+ ## MCP server for agents
804
+
805
+ `local_data_platform.mcp_server` serves your tables to an AI agent over the Model Context Protocol,
806
+ on stdio and read-only. Install the `mcp` extra, then point your MCP client at `ldp mcp`
807
+ (`python -m local_data_platform.mcp_server` takes the same options):
808
+
809
+ ```json
810
+ {
811
+ "mcpServers": {
812
+ "ldp": {
813
+ "command": "/abs/path/to/.venv/bin/ldp",
814
+ "args": ["mcp", "--config", "/abs/path/to/rides.json", "--max-rows", "500"]
815
+ }
816
+ }
817
+ }
818
+ ```
819
+
820
+ | Tool | Returns |
821
+ |---|---|
822
+ | `list_tables()` | Every table |
823
+ | `describe_table(table)` | Schema, partition spec, row count, snapshots, the latest `_ldp` run and quality status, and freshness |
824
+ | `query(sql, max_rows)` | The result of one read-only SQL statement |
825
+ | `sample_rows(table, n)` | A sample of rows |
826
+ | `table_history(table)` | The table's snapshot history |
827
+ | `get_dataset(name)` | A pinned dataset version |
828
+
829
+ The guardrails are on by default. A query must be exactly one `SELECT` or `WITH` statement, as
830
+ DuckDB's parser sees it, so `COPY`, `ATTACH`, `INSTALL`, `SET`, `read_csv('/etc/passwd')` and
831
+ multi-statement SQL are rejected. DuckDB runs with external access off and file access limited to
832
+ the served tables' folders, and its configuration is locked once the tables are registered.
833
+ Results are capped in rows and time, `--allow` limits the tables, and every call is audited: the
834
+ tool, the SQL, row counts, duration and any error, never the result data. Records go to a JSONL file
835
+ as each call ends, and to the `_ldp.audit` Iceberg table in the first served catalog when the
836
+ server stops (`--no-iceberg-audit` turns that off).
837
+ `make demo-agent` runs a scripted client against the demo table. See
838
+ [MCP server for agents](docs/agents.md).
839
+
840
+ ## Reproducible datasets and robotics
841
+
842
+ A dataset version pins a table to one snapshot, with an optional row filter and column list, and
843
+ is saved as a JSON manifest under `<warehouse>/.ldp/datasets/`. Loading it later returns exactly
844
+ the same rows, however the table has changed since:
845
+
846
+ ```python
847
+ from local_data_platform import datasets
848
+ from local_data_platform.format.iceberg import Iceberg
849
+
850
+ episodes = Iceberg("gold_episode_stats", {"identifier": "robots", "warehouse_path": "warehouse"})
851
+ train = datasets.pin(episodes, "humanoid_train", row_filter="success = true",
852
+ selected_fields=["episode_id", "robot_id", "task", "frame_count"])
853
+ rows = datasets.load(train) # the same rows at any later time
854
+ datasets.export(train, "exports/humanoid_train_v1.parquet") # or format="jsonl"
855
+ ```
856
+
857
+ Each pin also tags the snapshot (`ldp_ds_<name>_v<N>`), so snapshot expiry keeps it.
858
+ `datasets.list_versions(name, warehouse="warehouse")` and
859
+ `datasets.get_version("name@vN", warehouse="warehouse")` find versions again. From the command
860
+ line, `ldp datasets pin CONFIG NAME`, `ldp datasets list` and `ldp datasets export` do the same.
861
+
862
+ The temporal quality checks suit sensor and event data:
863
+
864
+ | `check` | Keys | Passes when |
865
+ |---|---|---|
866
+ | `monotonic` | `column`, `group_by`, `strict`, `order_by` | The column never decreases (or, with `strict`, always increases) within each group |
867
+ | `max_skew` | `column_a`, `column_b`, `max_ms` | Two timestamps on each row are at most `max_ms` apart |
868
+ | `rate_below` | `predicate_column`, `max_rate`, `group_by` | The share of rows where the boolean column is true stays below `max_rate` |
869
+ | `max_gap` | `column`, `max_ms`, `group_by` | Consecutive timestamps within each group are at most `max_ms` apart |
870
+
871
+ [`examples/robot_episodes/`](examples/robot_episodes/) puts it together: deterministic
872
+ humanoid-robot episodes and sensor frames, bronze, silver and gold tables (episodes partitioned by
873
+ `day(start_ts)` and `bucket(8, robot_id)`), quality configs for the frame-drop rate, RGB/depth sync
874
+ skew, monotonic frame timestamps and accepted tasks, a pinned training split, DuckDB research
875
+ queries and an optional Scala Spark per-robot aggregate. It runs offline in about a second with
876
+ `make demo-robotics`. See [Robot-episode datasets](docs/robotics.md).
877
+
878
+ ## Table maintenance
879
+
880
+ ```python
881
+ import datetime as dt
882
+
883
+ from local_data_platform import maintenance
884
+ from local_data_platform.format.iceberg import Iceberg
885
+
886
+ table = Iceberg("rides", {"identifier": "nyc", "warehouse_path": "warehouse"})
887
+ plan = maintenance.expire_snapshots(table, older_than=dt.timedelta(days=7), dry_run=True)
888
+ print(plan["expired_snapshot_ids"], plan["protected_snapshot_ids"]["idempotency_key"])
889
+ print(maintenance.find_orphans(table)) # files no metadata references; deletes nothing
890
+ report = maintenance.remove_orphans(table) # a dry run unless dry_run=False
891
+ ```
892
+
893
+ Expiry always keeps the current snapshot, branch and tag heads and `main`'s last 20 snapshots
894
+ (`retain_last`), and it never expires a snapshot that carries an idempotency key newer than the
895
+ table's idempotency horizon (7 days by default), because re-runs rely on finding it. Orphans are
896
+ files under the table's location that no metadata references and that are older than 72 hours
897
+ (`older_than_hours`), so the files of a write still in progress are left alone. `ldp maintain
898
+ CONFIG` runs both from the command line, as a dry run unless `--apply` is given. There is no
899
+ compaction yet.
900
+
901
+ ## Spark (experimental)
902
+
903
+ `engine.spark` reads and writes the same tables from Spark, in two ways: `ScalaSparkJob`, which runs
904
+ the Scala job in [`spark/`](spark/README.md) with scala-cli (it fetches its own JDK, Spark and
905
+ Iceberg), and `SparkEngine`, a PySpark engine (the `spark` extra plus a JDK 17+). Both open the same
906
+ SQLite catalog as pyiceberg, and pyiceberg reads back what Spark writes. `spark_catalog_conf_for`
907
+ turns a catalog block into Spark settings for the `local`, `sql` and `rest` types:
908
+
909
+ ```python
910
+ from local_data_platform.engine.spark import spark_catalog_conf_for
911
+
912
+ conf = spark_catalog_conf_for({"identifier": "nyc", "warehouse_path": "warehouse"})
913
+ print(conf["spark.sql.catalog.nyc.catalog-impl"]) # org.apache.iceberg.jdbc.JdbcCatalog
914
+ ```
915
+
916
+ `make demo-spark` runs the Scala job on the demo table, and `ldp spark CONFIG` on any config's
917
+ table. The integration tests are opt-in
918
+ (`make spark-test`, which sets `LDP_RUN_SPARK=1`). No pipeline uses Spark, and DuckDB is the
919
+ supported SQL engine. [Spark](docs/spark.md) covers versions, the shared catalog and the limits.
920
+
921
+ ## Architecture
922
+
923
+ ```text
924
+ rides.json ──► Config.from_json ──► registry.create_pipeline ──► Pipeline.run(mode, commit=, sink=)
925
+ │
926
+ ┌──────────────────────────────────────────────────────────┘
927
+ ▼
928
+ source.get() ──► transforms ──► quality checks ──► target.put(mode, commit=)
929
+ CSV · Parquet · your pyarrow fail: raise direct: one transaction, file lock
930
+ JSON (local, s3, callables DataQualityError, staged: branch ─► checks ─► publish (CAS)
931
+ gs via fs.py) · write nothing CSV · Parquet targets
932
+ Iceberg · BigQuery │
933
+ ▼
934
+ catalog provider: local (SQLite + folder) · sql · rest · glue ── warehouse: file · s3 · gs
935
+ │
936
+ RunEvents ──► JSONL · OpenLineage · Iceberg (_ldp.runs, _ldp.quality_results)
937
+ │
938
+ read by: DuckDB (iceberg_scan) · Spark (experimental) · MCP server (read-only, audited to _ldp.audit) · datasets.pin
939
+ ```
940
+
941
+ Data moves between steps as `pyarrow.Table`s. The package lives in `src/local_data_platform/`:
942
+
943
+ | Module | Purpose |
944
+ |---|---|
945
+ | `config`, `paths` | `Config` and path resolution relative to the config file |
946
+ | `spec` | The `ldp/v1` JSON Schema, config validation, `spec_hash` and `idempotency_key` |
947
+ | `fs` | Local and object-storage IO (`file://`, `s3://`, `gs://`) with atomic writes |
948
+ | `format` | `CSV`, `Parquet` and `Iceberg` tables (`get` / `put`), and `WriteResult` |
949
+ | `format.iceberg.commit` | The staged publish protocol: `CommitContext`, `write_once`, `stage`, `publish`, `fence` |
950
+ | `catalog` | `create_catalog` and the `local`, `sql`, `rest` and `glue` types; `LocalIcebergCatalog` |
951
+ | `store` | Sources: BigQuery (`[bigquery]` extra), JSON files, and older import paths kept for compatibility |
952
+ | `quality` | Checks, including the temporal ones in `quality.temporal`, `run_checks` and `QualityReport` |
953
+ | `events` | `RunEvent` and the event sinks |
954
+ | `engine.duckdb` | `DuckDBEngine` for SQL over Iceberg and Arrow tables (`[duckdb]` extra) |
955
+ | `engine.router` | Scan-size estimates and engine choice |
956
+ | `engine.spark` | Experimental: `SparkEngine`, `ScalaSparkJob` and `spark_catalog_conf_for`; see [Spark](docs/spark.md) |
957
+ | `mcp_server` | The read-only MCP server (`[mcp]` extra) |
958
+ | `datasets` | Snapshot-pinned dataset versions |
959
+ | `maintenance` | Snapshot expiry with the idempotency floor, and orphan-file detection |
960
+ | `pipeline` | `Pipeline`, the built-in pipelines and the registry (`create_pipeline`) |
961
+ | `cli`, `demo` | The `ldp` command and the demo |
962
+ | `github`, `issue` | GitHub REST helpers used by the project's docs tooling |
963
+
964
+ [Why a registry instead of if/else](docs/design/factory_registry.md) explains how pipelines are
965
+ chosen. The [SaaS architecture](docs/design/saas_architecture.md) is a proposal for where this
966
+ could go; none of its control plane exists.
967
+
968
+ ## Development
969
+
970
+ ```bash
971
+ make install # create .venv and install the package with the dev and docs extras
972
+ make lint # flake8
973
+ make test # pytest, offline: the Spark and REST integration tests skip
974
+ make demo # run the demo into ./ldp_demo
975
+ make demo-all # every demo; demo-warehouse needs the dbt and excel extras, demo-spark and demo-rest need scala-cli
976
+ make demo-orchestration # the restaurant's nightly workflow, a backfill and a scheduler tick
977
+ make spark-test # opt-in Spark integration tests (LDP_RUN_SPARK=1, scala-cli)
978
+ make rest-test # opt-in REST catalog integration tests (LDP_RUN_REST=1, scala-cli)
979
+ make dbt-test # the tests that need the dbt extra (pip install -e ".[dev,dbt]" first)
980
+ make docs # mkdocs build --strict
981
+ make build # build the sdist and wheel into dist/
982
+ make smoke # install the wheel in a fresh venv and run ldp --version and ldp demo
983
+ ```
984
+
985
+ The JVM demos and tests use [scala-cli](https://scala-cli.virtuslab.org), which provisions its own
986
+ Temurin 17 JDK; the first run downloads it with Spark, Iceberg and their jars (about 500 MB).
987
+ `tests/test_rest_catalog.py` and `make demo-rest` start the Iceberg REST fixture in
988
+ `tools/rest_fixture` (`make demo-rest` on port 8181; `REST_PORT=` changes it) and stop it
989
+ afterwards, also when a step fails.
990
+
991
+ CI runs lint, tests on Python 3.12 and 3.13 against both pyiceberg 0.11.0 and the newest supported
992
+ release, the wheel smoke test and a strict docs build. The Spark and REST jobs are in a separate,
993
+ opt-in workflow (`.github/workflows/jvm.yml`): they run when started by hand, weekly, or on a pull
994
+ request labelled `jvm`, and never block a merge. See the [changelog](CHANGELOG.md) for what changed
995
+ in each release, and the [contributing guide](docs/contributing.md) to get involved.
996
+
997
+ ## Known limitations
998
+
999
+ - **The engine router is unused.** Nothing calls `engine.router` yet.
1000
+ - **Remote catalogs are lightly tested.** `rest` is tested against Iceberg's REST test server and
1001
+ `glue` against moto. Postgres, hosted REST catalogs and GCS have no integration tests yet.
1002
+ - **Direct overwrite and upsert on a shared catalog are single-writer.** Several writers need the
1003
+ staged protocol (`commit=`). The local file lock only covers processes on one machine.
1004
+ - **Exactly once means at most one effect per key.** Duplicates a producer sent upstream are not
1005
+ removed.
1006
+ - **An upsert batch needs every column of the table.** An upsert replaces whole rows, so a batch
1007
+ that lacks one of the table's columns is rejected with a `ConfigError` that names them (appends
1008
+ accept it).
1009
+ - **Mostly in memory.** Reads return a whole `pyarrow.Table`. DuckDB's native `iceberg_scan`
1010
+ streams, but pipelines don't; there are no streaming or incremental reads.
1011
+ - **Spark is experimental.** It needs a JDK (or scala-cli) and downloads jars on first use, no
1012
+ pipeline uses it, and its tests are opt-in.
1013
+ - **The MCP server is read-only and local.** It speaks MCP over stdio only. Its `_ldp.audit` rows
1014
+ are appended when it stops, so a server that is killed leaves them only in the JSONL audit file.
1015
+ - **No compaction.** Maintenance expires snapshots and finds orphans; it doesn't rewrite data files.
1016
+ - **BigQuery is read-only.** It can be a source, not a target.
1017
+ - **`ldp dbt` publishes one model at a time.** Nothing is published when dbt fails, but the writes of a
1018
+ successful build are separate commits, so a write that fails, such as a blocked quality check or a model
1019
+ with no rows, leaves the models before it published. A model published with an upsert that gains or loses
1020
+ a column needs one `ldp run CONFIG --mode overwrite`.
1021
+ - **The restaurant example is in the repo, not in the package.** `pip install` does not include
1022
+ `examples/`, so the warehousing and orchestration guides and `make demo-warehouse` and `make demo-orchestration`
1023
+ need a clone.
1024
+ - **Tasks run one at a time.** A workflow runs its tasks in order and a backfill its windows in order, which also
1025
+ respects DuckDB's single writer. There are no parallel tasks.
1026
+ - **The scheduler is a tick, not a daemon.** `ldp schedule tick` has to be called by cron or launchd, and there is no
1027
+ foreground loop and no systemd or Windows Task Scheduler recipe. A window that failed is run again by every tick, every
1028
+ 5 minutes under the recipes, with no wait between ticks. `retries` are the retries inside one tick.
1029
+ - **A dbt task builds its whole selection for every window.** dbt gets `window_start` and `window_end`, but only a
1030
+ model that reads them is limited to the window, and the models of the restaurant example are not.
1031
+ - **Airflow runs one config for each task.** DAGs built from workflow files, an `LdpTaskOperator` and managed Airflow
1032
+ recipes are planned for 0.2.0.
1033
+
1034
+ ## Roadmap
1035
+
1036
+ From 0.1.1 on, the package version equals the milestone number. Each milestone adds the next layer
1037
+ of a data platform, and every layer runs on a laptop first. 0.1.3 Orchestration is released, and these
1038
+ are the milestones after it:
1039
+
1040
+ | Version | Milestone | What it adds |
1041
+ |---|---|---|
1042
+ | 0.1.4 | Self Serving Gold Layer | Gold tables with owners and docs, releases, metrics, a catalog and exports |
1043
+ | 0.1.5 | Monitoring | SLAs, freshness, volume, schema and metric monitors, alerts and `ldp status` |
1044
+ | 0.1.6 | Business Intelligence Reporting Dashboarding | Offline HTML reports, Slack delivery, a live dashboard and BI tool recipes |
1045
+ | 0.1.7 | Data Science Insights | Churn, retention, funnels, sessions, training sets and a notebook |
1046
+ | 0.1.8 | LLM | Catalog search and `ldp ask`, text to SQL that corrects itself |
1047
+ | 0.1.9 | Launch Documentation | The Learn path, generated reference pages, onboarding and contributor setup |
1048
+ | 0.2.0 | Cloud Integration | Profiles, cloud catalogs and storage, Snowflake and one container image |
1049
+
1050
+ The [roadmap](docs/roadmap.md) has the goal, issues and demo of each milestone.
1051
+
1052
+ ## References
1053
+
1054
+ - [PyIceberg](https://py.iceberg.apache.org)
1055
+ - [DuckDB Iceberg extension](https://duckdb.org/docs/extensions/iceberg.html)
1056
+ - [Model Context Protocol](https://modelcontextprotocol.io)
1057
+ - [OpenLineage](https://openlineage.io)
1058
+ - [NEAR Lake Framework](https://docs.near.org/concepts/advanced/near-lake-framework)
1059
+ - [Reliable change data capture using Iceberg](https://medium.com/@tushar.choudhary.de/reliable-cdc-apache-spark-ingestion-pipeline-using-iceberg-5d8f0fee6fd6)
1060
+ - [Internals of Apache PyIceberg](https://medium.com/@tushar.choudhary.de/internals-of-apache-pyiceberg-10c2302a5c8b)
1061
+
1062
+ ## License
1063
+
1064
+ MIT. See [LICENSE](LICENSE).
1065
+