Skip to Content

Console Video Tutorials

English learning series

Watch the new isA Console learning series with English on-screen captions, chapter navigation, transcripts, and practical review points.

Console is an agent orchestration and management platform. Start with usable capabilities, then progress through model training and testing, Agent Builder, and Maestro automation.

Learning path

TopicTask and result evidence
MCP ServerSave the connection and discover capabilities actually provided by the server.
Tools, Resources, Prompts, and SkillsConfigure contracts, content, and dependencies; test inputs and outputs; bind them to a real consumer.
Model testing and evaluationFix samples and parameters, inspect actual responses or case-level evaluation results, and retain repeatable comparisons.
Model trainingUse approved data, submit training, inspect artifacts, and compare actual tests before and after training. Parameter tuning does not replace training.
Agent BuilderCombine a role, model, and MCP capabilities; save a version, run it, and inspect output.
MaestroDesign triggers, nodes, mappings, and branches; validate and execute; inspect results, recovery, or human intervention.

Related guides: MCP Platform, Models and Routing, Agent Studio, and Workflows.

Before watching

Use the account, workspace, and existing permissions for the recorded environment. Draft creation, version saving, model execution, and agent or workflow release are separate actions; every film identifies the step actually completed. Demonstration inputs do not establish a real business conclusion.

Manual model experiments, structured output, Prompt configuration and recovery, and private Agent draft maintenance are task excerpts. Trace, usage, historical workflow, and capability-contract checks are supporting lessons. A complete orchestration result requires actual execution and consumer output.

Watch by environment

Only independently reviewed, encoded films appear below; missing topics have no placeholder player. These lessons have Chinese narration and captions, with chapters for individual actions.

把一段运营数据变成可校对摘要与代码

Core tutorial excerpt

Goal: 把一段示例运营数据整理成摘要,再准备调用代码

Result: 模型已生成摘要并打开调用代码;仍待人工校对与应用接入验证

Live execution (48s)

把一段运营数据变成可校对摘要与代码

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 完成单次演示数据摘要与代码查看;不宣称已完成应用发布。

让抽检记录成为可追溯的交接清单

Core tutorial excerpt

Goal: 换班时,需要把抽检记录的事实、待确认资料和责任动作分清

Result: 同一份质量交接Prompt草稿已读回并封存未激活版本;另在试验场手动代入虚构20件包装、1件外袋破损记录,得到三项清单供人工复核。未建立原生绑定执行,也未发布、激活或放行。

Configuration and verification (68s)

让抽检记录成为可追溯的交接清单

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 保存质检Prompt草稿并手工验证;不生成真实质检结论或业务审批。

把客服反馈提取成可核对的JSON

Core tutorial excerpt

Goal: 反馈缺少产品检查和绝对日期,需要提取结构化事实并交接配置

Result: 合成DEMO107经真实Qwen3.8 27B Fast请求返回JSON,未知产品状态与日期保留null;本次Schema及九项检查通过,导出语法检查通过的当前会话JS示例,未实际接入执行。

Live execution (56s)

把客服反馈提取成可核对的JSON

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 实际执行两组合成客服样例并导出JavaScript;未创建工单或联系客户。

保留实验基线,试新样例后恢复对照

Core tutorial excerpt

Goal: 试验信息完整的新反馈,同时保留缺失信息的原样例供对照

Result: 同一浏览器保存原会话与预设,新会话手动加载后DEMO108真实Qwen3.8 27B Fast调用成功并九项检查通过;另存独立会话。硬刷新后手选原DEMO107,模型、参数、System Prompt、Schema及原响应全部一致,未重新推理。

Local experiment session (59s)

保留实验基线,试新样例后恢复对照

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 本机浏览器实验会话的基线恢复;不是团队共享或跨设备重放。

恢复质检交接草稿的验收规则

Core tutorial excerpt

Goal: 修订抽检交接规范后,需要找回原验收规则并确认变量完整

Result: 保存演示修订3,再从既有封存草稿恢复为修订4;硬刷新后正文与{audience}、{inspection}两个变量等于原修订2,演示标记消失,仍为draft/inactive。

Draft configuration (46s)

恢复质检交接草稿的验收规则

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 恢复既有质检交接草稿并刷新读回;草稿未激活。

配置并导出客服交接 Skill

Core tutorial excerpt

Goal: 要把客户事实保留与未知信息处理规则变成团队可检查、可交接的 Skill 规范

Result: 选择 L4,配置 Skill 与边界,写好 SKILL.md;保存并通过服务端校验,刷新读回同一草稿,再导出实际 ZIP。信任等级仍待评审,未发布未激活。

Configuration and verification (85s)

配置并导出客服交接 Skill

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 已配置并保存Skill草稿、通过服务端校验、刷新读回及导出ZIP;信任等级待评审,未发布、未激活、未被Agent执行。

配置并核对客服上下文 Resource 草稿

Core tutorial excerpt

Goal: 希望把事实保留和人工核实清单组织成可管理的静态资源

Result: 填写 Markdown 与正确 URI,保存草稿,核对草稿预览和持久化读回;保持未激活。

Draft configuration (90s)

配置并核对客服上下文 Resource 草稿

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 仅完成静态 Resource 配置、保存读回与前端草稿预览;5/5为前端检查,未封存、审核、发布、激活或共享,未进行实际 resources/read 或 Agent 消费。

客服专家交付可复核交接规则包

Core tutorial excerpt

Goal: 把客服外包装破损与产品状态判定的交接经验交付为可校验规则包

Result: 保存 Skill 草稿并通过服务端格式校验

Draft configuration (145s)

客服专家交付可复核交接规则包

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 完成草稿保存与校验;未激活或下发至一线客服座席。

把客服反馈整理成可核对字段与对照实验

Core tutorial excerpt

Goal: 将非结构化客户反馈解析为规范字段并保留实验对照

Result: 在模型 Playground 中完成字段提取与比对

Local experiment session (147s)

把客服反馈整理成可核对字段与对照实验

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 仅进行合成样本比对;未自动创建工单或联系客户。

使用 Maestro 工作流生成统一客服交接包

Core tutorial excerpt

Goal: 把已有结构化客服信息交给固定工作流,生成可复核交接包

Result: 完成工作流确定性字段映射并输出统一交接记录

Live execution (112s)

使用 Maestro 工作流生成统一客服交接包

Open workbench
Jump to Key Action

This lesson uses demonstration or sample data.

  • 运行确定性映射工作流;未调用外部大模型做主观判定。

接手 Agent 前,先核对配置与边界

Supporting review and troubleshooting

Goal: 接手已有业务 Agent,准备后续维护与验证

Result: 已查看职责、模型、提示词和工具状态,停在接手检查阶段

Read-only review (47s)

接手 Agent 前,先核对配置与边界

Open workbench
Jump to Key Action
  • 只读核对既有Agent配置与边界;未新建或运行Agent。

旧运行完成后,核对版本与当前绑定

Supporting review and troubleshooting

Goal: 已有历史完成记录的大促编排当前提示需要处理,需要确定旧运行基线并定位当前依赖缺口

Result: 核对两份历史发布版及同次completed运行的固定release/revision,回看持久化事件;当前服务端校验仅确认指定工具能力版本不可用,形成节点与引用的排查交接。未修复、发布或重新运行。

Read-only review (42s)

旧运行完成后,核对版本与当前绑定

Open workbench
Jump to Key Action
  • 回看既有运行和当前缺失能力引用;未重新绑定、发布或运行。

从失败客服运行找到步骤与证据缺口

Supporting review and troubleshooting

Goal: 历史保修咨询Trace为错误,需要定位具体失败步骤并明确下一步取证

Result: 在SN真实历史Trace中定位回复生成步骤error,核对保修问题输入及仅含statuserror/duration的输出;确认后端未附错误详情,检索success也无命中文档内容,形成继续补齐上游记录的交接。未修复或重新运行。

Read-only review (41s)

从失败客服运行找到步骤与证据缺口

Open workbench
Jump to Key Action
  • 定位历史失败步骤和证据缺口;未证明根因、修复或重跑成功。

All product videos · AEP governance · DXP data application

Console overview