Datasource

datasource

Reference data discovery skill (read-only).

Purpose

Query and browse reference data as standalone read-only operations. Use when user wants to explore available options without submitting a request. This skill owns standalone business-group scope discovery. Dedicated resource-pool and resource skills still handle standalone resource-pool and resource browsing.

Terminology Mapping

Treat SmartCMP business group as a generic organizational scope. Users may describe the same concept as:

• tenant
• 租户
• 部门
• BU
• Department
• 项目
• Project

Resolve these terms against SmartCMP business-group data unless the user is clearly referring to some other system-level tenant concept. Mirror the user's wording in replies when it helps readability, but keep the SmartCMP field names businessGroupName and bgId when calling scripts or building request data.

Trigger Conditions

Activate this skill when user intent matches:

Intent	Keywords
View catalogs	"show catalogs", "list services", "available services"
View business-group scopes	"show business groups", "show tenants", "查看租户", "查看部门", "查看项目"
List applications	"list applications", "show apps"
List OS templates	"list OS templates", "available OS"
List images	"list images", "available images"
Show resource details	"resource details", "show resource", "analyze resource data"

NOT for: Resource provisioning -> use request skill instead. NOT for: Direct "查询资源池", "查看所有资源", "查看所有云主机", or "查看某个云主机详情" requests -> use resource-pool or resource.

Scripts

Most scripts are located in scripts/.

Script	Description	Arguments
scripts/list_all_business_groups.py	List all business-group scopes (tenant / 租户 / 部门 / BU / 项目) from the standalone UI directory endpoint	[QUERY_VALUE]
scripts/list_services.py	List published service catalogs; the request skill also uses its internal exact-detail mode	[KEYWORD] or --catalog-id <CATALOG_ID>
scripts/list_resource.py	List resource evidence by resource ID from /nodes/{resourceId}/view first, with legacy resource fallback	<RESOURCE_ID> [RESOURCE_ID ...]
../shared/scripts/list_applications.py	List applications for a selected business group	<BUSINESS_GROUP_ID>
../shared/scripts/list_components.py	List component metadata for a catalog sourceKey	<SOURCE_KEY>
../shared/scripts/list_images.py	List image options for a selected resource pool and logic template	<RESOURCE_BUNDLE_ID> <LOGIC_TEMPLATE_ID> <CLOUD_ENTRY_TYPE>

scripts/list_resource.py emits resource.data as the canonical SmartCMP resource evidence pack and also emits a normalized type + properties view per resource. It currently calls PATCH /nodes/{resourceId}/view because SmartCMP exposes this read-only view through PATCH; switch to GET after the CMP API bug is fixed. If /view is unavailable or returns no data, fallback to GET /nodes/{resourceId} and GET /nodes/{resourceId}/details; preserve the primary /view error in errors. If fallback provides resource data, keep fetchStatus=ok and set fallbackUsed=true.

Environment Setup

Option 1: Direct Cookie

# PowerShell - CMP_URL auto-normalizes (adds /platform-api if missing)
$env:CMP_URL = "<your-cmp-host>"           # e.g., "cmp.example.com" or "https://cmp.example.com/platform-api"
$env:CMP_COOKIE = '<full cookie string>'

# Bash
export CMP_URL="<your-cmp-host>"
export CMP_COOKIE="<full cookie string>"

Option 2: Auto-Login (Recommended)

Automatically obtains and caches cookies (30-minute TTL). Auth URL is auto-inferred.

# PowerShell
$env:CMP_URL = "<your-cmp-host>"
$env:CMP_USERNAME = "<username>"
$env:CMP_PASSWORD = "<password>"

# Bash
export CMP_URL="<your-cmp-host>"
export CMP_USERNAME="<username>"
export CMP_PASSWORD="<password>"

Workflow Examples

Example 1: List Business-Group Scopes

User: "Show available tenants or projects"

python scripts/list_all_business_groups.py

Output: Numbered list + ##BUSINESS_GROUP_DIRECTORY_META_START## ... ##BUSINESS_GROUP_DIRECTORY_META_END##

Example 2: List Available Catalogs

User: "Show available catalogs"

python scripts/list_services.py

Output: Numbered list + ##CATALOG_META_START## ... ##CATALOG_META_END##

Example 3: Show Resource Details

User: "Show resource details for ID X"

python scripts/list_resource.py <resource_id>

This calls PATCH /nodes/{resourceId}/view first. resource.data is the evidence pack passed to standalone IT Ops skills. If the primary view fails, the script falls back to the older resource/detail APIs. If fallback provides resource data, the result remains fetchStatus=ok with fallbackUsed=true.

Data Flow

scripts/list_all_business_groups.py [query_value]
  -> standalone business-group scope discovery

scripts/list_services.py [keyword]
  -> (catalogId, sourceKey)

scripts/list_services.py --catalog-id <catalog_id>
  -> exactly one normalized catalog detail for internal request-skill use

Output Meta Blocks

Script	Meta Block
scripts/list_all_business_groups.py	##BUSINESS_GROUP_DIRECTORY_META_START## ... ##BUSINESS_GROUP_DIRECTORY_META_END##
scripts/list_services.py	##CATALOG_META_START## ... ##CATALOG_META_END##

Critical Rules

All operations are read-only - no data is created or modified.

Scripts are shared with the request skill.

Standalone directory queries for all business-group scopes belong to datasource. Standalone resource-pool and resource browsing belong to resource-pool and resource.

On error ([ERROR]), report to user immediately; do NOT self-debug.

Error Handling

Error	Resolution
401 / Token expired	Ask user to refresh cookie
Missing arguments	Check script usage in docstring

References

• WORKFLOW.md - Detailed script usage and query flows