intersystems-community/ai-hub-dev-template/ai-hub-config-store

AI Hub Config Store

Use this skill when AI Hub components should read model or MCP settings from governed IRIS configuration instead of raw environment variables.

When to Use

• Creating LLM configurations for ObjectScript, LangChain, or other AI Hub consumers
• Creating MCP server configurations for local or remote tools
• Storing secrets in IRIS Wallet and referencing them from configuration JSON
• Explaining the boundary between Wallet secret storage and Config Store metadata

Core Workflow

1. Create IRIS security resources for use and edit permissions.
2. Create or choose a Wallet collection.
3. Store secrets with %Wallet.KeyValue.
4. Create config entries with %ConfigStore.Configuration.
5. Reference Wallet values through secret://... URIs.
6. Retrieve configs by area, type, subtype, and name.

Critical Rules

• Store secrets in the Wallet, not directly in general config JSON when governed storage is expected.
• Use Config Store for structured metadata and references to those secrets.
• Keep naming consistent with the <area>.<type>.<subtype>.<name> pattern implied by the docs.
• Pair this skill with ai-hub-langchain when the consumer is Python LangChain code.

Common Mistakes

• Treating Wallet and Config Store as interchangeable
• Forgetting to define UseResource and EditResource
• Using raw API keys in config JSON when the Wallet should hold them
• Mixing up logical config names with full dotted identifiers

References

• Config Store reference