datapaw-cli

Standalone command-line interface for DataPaw.

This package provides the datapaw command and connects local host workflows through datapaw-host-core.

The command automatically loads DATAPAW_ENV_FILE when set. Otherwise, an editable checkout is discovered from the CLI package location and the repository root .env is loaded. Existing process environment variables take precedence over dotenv values. Outside an editable checkout, the current working directory is used as the fallback configuration root.

Available command groups:

• plan
• execute
• run
• chat
• datasource
• semantic
• doctor

Run datapaw doctor before the first task to check supported Python/Node versions, uv, model and authentication configuration, DataPaw Home, DataBridge MCP, Docker, Neo4j, and the DataBridge API. datapaw doctor --json emits a machine-readable report that reports whether credentials are set without printing their values.

datasource manages the datasources configured in DataBridge:

datapaw datasource list
datapaw datasource get <datasource_id> [--show-config]
datapaw datasource create --name <n> --type <t> \
  (--config-file <f.json> | --config '<json>') [--test]
datapaw datasource update <datasource_id> [--name <n>] [--type <t>] [--config-file <f>]
datapaw datasource delete <datasource_id> [--yes]
datapaw datasource test (<datasource_id> | --type <t> --config-file <f>)

Set DATAPAW_CM_BASE_URL to the DataBridge origin. It defaults to http://127.0.0.1:8765. Every command emits JSON and masks saved passwords, AccessKeys, and STS tokens before writing anything to stdout; there is no option that prints stored credentials. list uses the credential-free discovery endpoint, while the management subcommands call the /api/semantic-config routes and need an API key with the credentials:manage scope when scoped keys are configured (DATAPAW_CLIENT_API_TOKEN or DATAPAW_API_TOKEN).

semantic manages the DataBridge semantic configuration layer. The seven resources domain, dataset, column, dimension, binding, metric, and formula share the same verbs:

datapaw semantic <resource> list [filters] [--page N --size N | --all]
datapaw semantic <resource> get <id>
datapaw semantic <resource> create (field flags | --file payload.json)
datapaw semantic <resource> update <id> (field flags | --file payload.json)
datapaw semantic <resource> delete <id> [--yes]

binding and formula also support dataset-level batch deletion with delete --dataset-id <id>. Reads require the query scope and writes the manage scope. Two more subcommands cover import and publishing:

# Import a semantic configuration workbook into the draft store
datapaw semantic import --file demo_semantic_config.xlsx

# Publish (weave) the configuration into the graph store
datapaw semantic weave submit --datasource-id <id> [--mode FULL] [--wait]
datapaw semantic weave list [--datasource-name <n>] [--task-name <n>]
datapaw semantic weave kill <task_id>

With --wait the CLI polls until the task reaches a terminal state (success, failed, killed), printing progress to stderr and the final task record as JSON to stdout; the exit code is 0 only for success.

The default CLI model uses provider openai together with LLM_MODEL, OPENAI_API_KEY, and OPENAI_BASE_URL. Set the corresponding DATAPAW_MODEL_PROVIDER, DATAPAW_MODEL_NAME, DATAPAW_MODEL_API_KEY, and DATAPAW_MODEL_BASE_URL variables to override individual values for the CLI.

Output and logs

The CLI writes streamed agent events, final answers, and execution summaries to stdout. Interactive prompts and concise command errors remain explicit CLI output on the terminal rather than raw log records. Python logging is written to ${DATAPAW_HOME}/host/datapaw.log at INFO level and above; DATAPAW_HOME defaults to ~/.datapaw.

The log rotates at 50 MB and keeps two backups as datapaw.log.1 and datapaw.log.2. Rotation assumes one active CLI process per DATAPAW_HOME. Direct users of datapaw-host-core continue to own their process logging configuration.

This package intentionally omits serve and local initialization helpers from the public CLI help.