From oracle-ai-data-platform-workbench-engineer-agent
Ingests CSV/JSON/Parquet files into managed AIDP Delta tables via the `aidp` CLI (one-step or three-step upload→infer→create). Use when the user says "load this file into a table" or "create a table from a file."
How this skill is triggered — by the user, by Claude, or both
Slash command
/oracle-ai-data-platform-workbench-engineer-agent:aidp-ingest-file-to-tableThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Land a file into a managed AIDP table, either in one call or via the staged 3-step flow when you need to
aidp-ingest-file-to-table — file → managed Delta tableLand a file into a managed AIDP table, either in one call or via the staged 3-step flow when you need to
review/adjust the inferred schema. This is a control-plane flow on the DataLake schema/tables
resource. Primary engine: the official Oracle aidp CLI (same REST API + auth); oci raw-request is
the fallback when the CLI isn't installed.
Per references/aidp-cli-map.md: schema generate-temp-file-upload-target
→ schema infer / infer-with-preview → schema create-data-table / create-table (also
schema retrieve-par). All commands take --instance-id <DATALAKE_OCID> --auth api_key --profile DEFAULT --region <r>.
# 3-step (control): stage → infer → create
aidp schema generate-temp-file-upload-target --instance-id <DATALAKE_OCID> --auth api_key --profile DEFAULT --region us-ashburn-1 # returns upload target / PAR (also: retrieve-par)
aidp schema infer-with-preview --instance-id <DATALAKE_OCID> --auth api_key --profile DEFAULT --region us-ashburn-1 # review columns/types/preview (or: infer)
aidp schema create-data-table --body-file .aidp/payloads/create-data-table-<name>.json \
--instance-id <DATALAKE_OCID> --auth api_key --profile DEFAULT --region us-ashburn-1 # or: create-table
Mutating ops (create-data-table/create-table, upload): persist the body to .aidp/payloads/
and confirm with the user before running (see references/payloads.md).
Fallback (no CLI) — same REST + auth via oci raw-request against
…/20240831/dataLakes/<OCID>/… (auth ladder in references/oci-raw-request.md):
POST /tables/actions/uploadDataFile (multipart/binary may need PAR upload — see aidp-volumes),
POST /tables/actions/inferSchema, POST /tables/actions/createTable (with catalogKey, schemaKey,
table name, finalized columns, source format, load options), verify GET /tables?catalogKey=<cat>&schemaKey=<cat.schema>.
Verify-first (no-fabrication): the upload/infer/create action shapes are UNVERIFIED in this env (not yet in
references/rest-endpoint-map.md). Confirm with a live probe (start with aGET /tables?catalogKey=…&schemaKey=…200 against the target schema) before any write; record results.
Live-verified 2026-06-10 on de-agent (CSV →
de_ingest_test, 3 rows) — correction: theuploadDataFile/inferSchema/createTableaction names above are WRONG. The working flow is theschema-resource 3-step: (1)generate-temp-file-upload-targetreturns a PAR +ociFilePath; (2) PUT the file bytes to the PAR (HTTP 200); (3)infer-with-preview— itslocationMUST be theociFilePathOCI URI, not theuploadKey(passinguploadKey→ 400); (4)create-data-tablereturns 202 + adatalake-async-operation-key(poll toSUCCEEDED).create-data-tableis HEADERLESS/POSITIONAL:header=trueis ignored at create, sotableFieldsmust use the reader column names_c0/_c1/_c2…— naming themid/name/amtfails the async op withUNRESOLVED_COLUMN. Rename afterward viaALTER TABLE … RENAME COLUMN.
catalog.schema.table
(create the schema first if needed).aidp schema create-table referencing the source file, format, and options —
fastest when the schema infers cleanly.generate-temp-file-upload-target → infer-with-preview (review columns/types
with the user; fix types/headers/delimiters) → create-data-table with the finalized columns.202 with an async-operation key — poll until terminal (async
convention in references/oci-raw-request.md; track via aidp-observability).aidp schema list-tables / GET /tables?…; report the fully-qualified table name and row/column summary.aidp-federate,
not this skill (this is file→table).aidp-workspace-files, aidp-volumes, aidp-profiling-tablesnpx claudepluginhub anthropics/claude-plugins-official --plugin oracle-ai-data-platform-workbench-engineer-agent2plugins reuse this skill
First indexed Jun 12, 2026
Imports data into AWS data lake (S3 Tables or Iceberg) from S3 files, local uploads, JDBC (Oracle, PostgreSQL, MySQL, SQL Server, RDS), Redshift, Snowflake, BigQuery, DynamoDB, or Glue tables. For one-time loads, pipelines, migrations.
Runs write-side Spark SQL on an AIDP lakehouse: DDL (CREATE/ALTER/DROP), DML (INSERT/UPDATE/DELETE/MERGE upsert), and Delta maintenance (OPTIMIZE, VACUUM, time travel, RESTORE, schema evolution, liquid clustering). Use for any non-SELECT SQL.
Registers Parquet, CSV, JSON, Arrow IPC, or Avro files as persistent external tables in DataFusion sessions. Auto-detects format, explores schema, and persists state for reuse across skills.