Documentation
¶
Overview ¶
Package trait defines the Trait entity — who an agent is.
Traits are personality dimensions that influence HOW an agent works. They are declarative, not imperative — you don't tell the agent what to do, you tell it who it is.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Dimension ¶
type Dimension struct {
Name string `json:"name"`
LowLabel string `json:"low_label"`
HighLabel string `json:"high_label"`
Value float64 `json:"value"`
}
Dimension represents a bipolar personality axis. Value ranges from 0.0 to 1.0 where each extreme has a label.
type Influence ¶
type Influence struct {
Target InfluenceTarget `json:"target"`
Value any `json:"value"`
Condition string `json:"condition,omitempty"`
Weight float64 `json:"weight,omitempty"`
}
Influence describes a runtime modification applied by a trait.
type InfluenceTarget ¶
type InfluenceTarget string
InfluenceTarget is what a trait influence modifies at runtime.
const ( TargetPromptInjection InfluenceTarget = "prompt_injection" TargetTemperature InfluenceTarget = "temperature" TargetMaxSteps InfluenceTarget = "max_steps" TargetToolSelection InfluenceTarget = "tool_selection" TargetResponseStyle InfluenceTarget = "response_style" )
type ListFilter ¶
ListFilter controls pagination and matching for trait listing. Scope arrives on the context; Exact narrows to rows stored at precisely that depth instead of everything beneath it.
type Store ¶
type Store interface {
CreateTrait(ctx context.Context, trait *Trait) error
GetTrait(ctx context.Context, traitID id.TraitID) (*Trait, error)
GetTraitByName(ctx context.Context, name string) (*Trait, error)
UpdateTrait(ctx context.Context, trait *Trait) error
DeleteTrait(ctx context.Context, traitID id.TraitID) error
ListTraits(ctx context.Context, filter *ListFilter) ([]*Trait, error)
CountTraits(ctx context.Context, filter *ListFilter) (int64, error)
}
Store defines persistence for traits.
type Trait ¶
type Trait struct {
cortex.Entity
ID id.TraitID `json:"id"`
Name string `json:"name"`
Description string `json:"description,omitempty"`
Scope cortex.Scope `json:"scope"`
Dimensions []Dimension `json:"dimensions,omitempty"`
Influences []Influence `json:"influences,omitempty"`
Category Category `json:"category,omitempty"`
Metadata map[string]any `json:"metadata,omitempty"`
}
Trait represents a personality dimension that influences agent behavior.
func (*Trait) Sections ¶
Sections returns the trait's contribution to an agent's system prompt: one section per prompt-injection influence, in the order the influences are declared. A trait with no prompt injections returns nothing, which is the common case for traits that only move temperature or tool selection.
order is where the trait's first section sits in the assembled prompt, and further sections take consecutive orders from there. A caller walking an agent's trait list starts at prompt.OrderTrait and advances by the number of sections each trait returned, since assembly falls back to sorting by ID when two sections share an order and that would reorder the traits alphabetically.
The heading stays inside Body rather than moving to Title so the assembled text is byte-identical to the prompt this injection has always produced. Title is for sections a host authors.
The first injection is addressed as "trait:<name>". A trait with more than one gets "trait:<name>:2" onwards, so the ordinary single injection keeps the ID an overlay author would guess.
Source Files