AI Assistant Configuration

The Higson AI Assistant is a beta feature that integrates an LLM provider directly into Higson Studio. Before the AI Assistant can be used, two steps are required: enabling the feature flag and configuring an LLM provider with a model.


Enabling the AI Assistant#

Property Description Default value
higson.studio.ai.features.enabled Enables the AI Assistant — the Documentation Assistant chat panel and AI actions (Explain / Generate) in Functions and Decision Tables. When false, all AI controllers and adapters are not loaded. false
higson.studio.ai.features.thinking-budget Global override for the reasoning budget of models that support reasoning mode. 0 disables reasoning, a positive value sets the token limit, empty value uses the default settings of each feature.

Configuring an LLM Provider#

The provider is configured under higson.studio.ai.provider.

Provider Properties#

Property Required Description Default value
higson.studio.ai.provider.code yes Unique identifier for the provider (e.g., vertex-ai, openai).
higson.studio.ai.provider.name yes Display name of the provider, used in logs and the interface.
higson.studio.ai.provider.type yes Adapter type: VERTEX_AI or OPENAI_COMPATIBLE.
higson.studio.ai.provider.base-url yes Base URL of the provider API.
higson.studio.ai.provider.api-key for OPENAI_COMPATIBLE API key / access token.
higson.studio.ai.provider.auth-header-name no Name of the authentication header. Authorization
higson.studio.ai.provider.auth-header-prefix no Prefix for the header value (with a trailing space). Bearer
higson.studio.ai.provider.additional-headers no Map of additional headers appended to every request.
higson.studio.ai.provider.connection-timeout no Connection timeout in seconds. 10
higson.studio.ai.provider.response-timeout no Timeout for waiting for a response in seconds. 300
higson.studio.ai.provider.read-timeout no Data read timeout in seconds. 120
higson.studio.ai.provider.vertex-project-id for VERTEX_AI Google Cloud project ID.
higson.studio.ai.provider.vertex-credentials-path for VERTEX_AI Path to the service account key file (JSON). If not set, Application Default Credentials (ADC) are used.

vertex-project-id and vertex-credentials-path apply only to the VERTEX_AI adapter. api-key applies only to OPENAI_COMPATIBLE. All other provider and model properties are shared between both adapter types.

Model Properties#

Property Required Description Default value
higson.studio.ai.model.name yes Display name of the model.
higson.studio.ai.model.model-id yes Model identifier used in the provider API (e.g., gemini-2.5-flash, gpt-4o-mini).
higson.studio.ai.model.max-output-tokens no Maximum number of tokens the model can generate in a single response. 4096
higson.studio.ai.model.context-window no Context window size of the model. Informational value.
higson.studio.ai.model.supports-streaming no Whether the model supports streaming responses. true
higson.studio.ai.model.supports-json-mode no Whether the model supports structured JSON output mode. false
higson.studio.ai.model.supports-tools no Whether the model supports tool-calling. Required by the Documentation Assistant. true

The Documentation Assistant requires a model with supports-tools: true. Models that do not support tool-calling can still be used for simpler actions such as Explain and Generate.


See Also#

If you want to connect an external AI client (such as Claude Desktop, Cursor, or VS Code) to Higson instead of using the built-in AI Assistant, see Enabling the MCP Server.