Mini-App With Custom AI Chat Interface

Field	Value
Complexity	Medium to High
Example	Script editor or EyePad-style mini-app with an embedded AI chat.
End Goal	Add a chat panel that can answer questions, inspect editor context, and call approved tools.
Key Technologies	AI runtime, AI chat UI, AI tools/plugins, Blazor

Use When

• You are building a mini-app or editor and want an AI chat inside it.
• The chat should know what script, tab, error, or selection the user is looking at.
• The chat may call tools such as docs search, diagnostics, or safe editor actions.

Shape

• Put a chat panel next to the editor or in a tool window.
• Build the AI session from the editor context the user can actually see.
• Add tools one by one: read selected text, read diagnostics, search docs, then propose or apply edits if needed.
• Ask for approval before tools change files, scripts, or app state.

Choices

• Is this chat tied to one editor tab, one workspace, or the whole mini-app?
• Which context is included automatically?
• Which tools are read-only?
• Which tools need explicit approval?
• Does this assistant need docs search, or only local editor context?

APIs To Look Up

• AiEngine
• IAiChatSession
• IAiChatSessionHandle
• AiSessionOptions
• AiChatViewModel
• AiChatComponent
• AiKernelPlugin
• IAiPluginHost
• AiFunctionInvocationPolicy
• AiDocsKnowledgeBasePlugin
• IBm25KnowledgeBase

Avoid

• Making documentation search mandatory for every AI chat.
• Tools that change files or app data without clear names and approval behavior.
• Letting the assistant guess hidden editor data.
• Assuming script-only globals exist in normal helper/plugin classes.
• Persisting active session ids or transient request data.

Related Maps

• ml/ai-runtime.md
• ml/ai-kernel-plugins.md
• scripting/runtime.md
• windows-subsystems/blazor-windows.md