uip ixp

Commands for managing IXP (Intelligent Document Processing) projects, taxonomy, documents, labelling, and deployments using the `uip ixp` tool.

uip ixp manages IXP projects — UiPath's intelligent document-processing product for training and deploying document-extraction models. A project holds a taxonomy (the schema of what to extract — data types, field groups, and fields), a set of training documents, and human-reviewed labellings (confirmed/corrected predictions) that improve the trained model. Once trained, a model version is deployed to an Orchestrator folder, where activities and Maestro Flow can call it at runtime. ixp is the command prefix; the underlying package is @uipath/ixp-tool.

资源

Each resource has its own reference page. The verb set matches the commands registered by uip ixp — run uip ixp --help to confirm the full surface on your installation.

资源	用途
项目	Create, configure, and manage IXP projects — including model publishing, metrics, and taxonomy import/export.
taxonomy — fields, groups, data-types	Edit a project's extraction schema: data types (reusable value definitions), field groups (document-type containers), and the fields inside them.
documents & labellings	Upload/download/delete training documents, and confirm, correct, or mark-missing the model's predictions on them.
deployments	Deploy a trained model version to an Orchestrator folder, move a deployment to another version, and inspect a deployed version's taxonomy.

概念

• Taxonomy — a project's extraction schema, made of three related concepts: data types (entity_defs — reusable type definitions like "Email Address" or a "Citizenship" choice list, each field references one by name), field groups (label_defs — document-type containers, e.g. "Invoice" or a repeatable "Line Items" group), and fields — the individual named values inside a group (e.g. "Invoice Number" inside the "Invoice" group), each backed by a data type and per-field extraction instructions. uip ixp projects get-taxonomy/import-taxonomy move the whole taxonomy as one artifact (its own internal format — data types and field groups together); uip ixp taxonomy's fields/groups/data-types commands edit one piece at a time.
• Training loop — create a project (optionally seeded with documents and an auto-suggested taxonomy) → refine the taxonomy (data-types/groups/fields) → upload more documents → review the model's predictions (labellings get-predictions) and confirm or correct them (labellings confirm/unconfirm/mark-missing) → check projects get-metrics → projects publish a model version → deployments create it to an Orchestrator folder.
• Model versions — a project accumulates trained model versions as labelling and retraining progress. projects publish/unpublish/untag/list-models manage which versions are published and tagged (live/staging); deployments create/upgrade control which published version an Orchestrator folder actually serves at runtime — publishing a version and deploying it are separate steps.
• Output key casing — most uip ixp responses forward the underlying IXP API payload verbatim, which means camelCase or snake_case keys, not this CLI's usual PascalCase — each page calls this out per verb where it applies. This is deliberate passthrough behavior, not a bug.

大纲

uip ixp <resource> <verb> [options]

另请参阅

• Tools (plugins) — the install model for uip ixp and other tool packages.
• uip tools — installs and manages tool packages.
• 全局选项
• 退出代码