> ## Documentation Index
> Fetch the complete documentation index at: https://docs.xpertai.cn/llms.txt
> Use this file to discover all available pages before exploring further.

# Workbench and Resource Chat

> Use ChatKit, Agent Workbench, and visualized results in resource context.

The UOSE system provides not only APIs but also an interactive user-facing workbench. In resource details, the ontology workspace, or Agent Workbench, users can let the ontology assistant carry current context for conversations, analysis, and operations.

## Resource Chat

Resource chat on the resource details page carries the current resource context:

* resourceId.
* resourceType.
* snapshotId.
* graphVersion.
* currently selected entity.
* current graph or search context.

Users can directly ask about the current resource, such as:

* Which entity types were synced from this resource.
* Why a metric cannot be queried.
* Which actions a SAP Entity Set supports.
* Which fields, primary keys, and foreign keys exist in the current database table.

The ontology assistant should use UOSE Agent Tools to obtain facts instead of guessing from page text.

## Ontology Workspace Chat

The assistant in the ontology workspace is for cross-resource semantic operations. It is suitable for:

* Comparing snapshot status across resources.
* Finding entities and explaining match methods.
* Analyzing missing graph relationships.
* Guiding users into resource details for troubleshooting.

When ChatKit target is not configured or Ontology Assistant is not initialized, the page prompts administrators to complete runtime initialization.

## Agent Workbench

Agent Workbench hosts Assistant execution processes and result display. After UOSE actions execute, they can generate workbench visualization events, such as:

* Semantic model query results rendered as trends, tables, or metric summaries.
* OData read results rendered as collection tables or charts.
* Database query results rendered as tables and column descriptions.

Visualization does not change action semantics. It only presents the effect in a user-understandable form.

## Context Passing

In ChatKit or MCP scenarios, identity and context must be passed correctly:

* Platform token is used to access UOSE or the MCP server.
* Business user token is used to access resources that require user identity at the source.
* assistantId or assistantCode is used to restrict the resources accessible by that Assistant.
* tenantId and organizationId are used for data isolation.

When these contexts are missing, tools should reject access instead of falling back to global default resources.

## Recommendations

* In resource details, inspect the graph and attributes before asking the assistant to run analysis.
* Let the assistant follow the standard UOSE process and do not ask it to directly compose backend APIs.
* For high-risk operations, review the simulation result and approval reason first.
* After execution, use audit references or workbench artifacts to review results.
* For cross-resource questions, first locate resources in the ontology workspace, then enter single-resource execution.