Documentation
¶
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type ElicitResult ¶
type ElicitResult struct {
// Action is either "accept" or "decline".
Action ElicitResultAction `json:"action"`
// Payload is the user supplied map that conforms to the supplied schema
// when Action == "accept". It is nil if the action is "decline".
Payload map[string]any `json:"payload,omitempty"`
// Reason optionally carries a human readable explanation when the user
// declines an elicitation. It is empty for accept actions.
Reason string `json:"reason,omitempty"`
}
ElicitResult represents the outcome of the elicitation prompt.
type ElicitResultAction ¶
type ElicitResultAction string
ElicitResultAction defines the action selected by the user after the elicitation process finished.
const ( // ElicitResultActionAccept indicates that the user supplied a payload that // satisfies the schema and accepted the request. ElicitResultActionAccept ElicitResultAction = "accept" // ElicitResultActionDecline signals that the user declined to provide a // payload. ElicitResultActionDecline ElicitResultAction = "decline" )
type Elicitation ¶
type Elicitation struct {
mcpproto.ElicitRequestParams `json:",inline"`
// CallbackURL is a server-relative URL that the UI can POST to
// with a body {action, payload} to resolve this elicitation.
// It is optional and, when present, preferred over generic
// form submission so both LLM- and tool-initiated prompts share
// a unified posting contract.
CallbackURL string `json:"callbackURL,omitempty"`
}
Elicitation describes a request to obtain a user-supplied payload that conforms to the supplied JSON Schema document (Schema field). The struct embeds mcp-protocol’s ElicitRequestParams so that callers that already work with the full protocol can freely convert between the two without manual copying.
Only a subset of the original fields is actually used by Agently today. The additional Schema string is a convenience copy of RequestedSchema encoded as a single JSON document so that generic front-ends (CLI/stdio) do not have to reconstruct it from the sub-fields.
func (*Elicitation) IsEmpty ¶
func (e *Elicitation) IsEmpty() bool
IsEmpty reports whether the elicitation is effectively empty (i.e. there is nothing to ask the user). The heuristic is intentionally simple so that it does not require full JSON-Schema parsing.
type Plan ¶
type Plan struct {
ID string `yaml:"id,omitempty" json:"id,omitempty"` // Unique identifier for the plan
Intention string `yaml:"intention,omitempty" json:"intention,omitempty"` // Optional summary of the user’s goal
Steps Steps `yaml:"steps" json:"steps"` // Ordered list of steps to execute
Elicitation *Elicitation `yaml:"elicitation,omitempty" json:"elicitation,omitempty"` // Optional elicitation details if user input is needed
}
Plan represents an ordered strategy composed of one or more steps.
type Step ¶
type Step struct {
ID string `json:"id,omitempty" yaml:"id,omitempty"`
Type string `yaml:"type" json:"type"` // "tool", "elicitation", "abort" etc.
Name string `yaml:"name,omitempty" json:"name,omitempty"` // Tool/function name (if applicable)
Args map[string]interface{} `yaml:"args,omitempty" json:"args,omitempty"` // Tool arguments matching tool schema
Reason string `yaml:"reason,omitempty" json:"reason,omitempty"` // Explanation of the step
Content string `yaml:"content,omitempty" json:"content,omitempty"`
// Structured elicitation payload (when Type=="elicitation").
Elicitation *Elicitation `json:"elicitation,omitempty"`
// Retries specifies how many times to retry this tool on error or empty result
Retries int `yaml:"retries,omitempty" json:"retries,omitempty"`
// ResponseID carries the provider response.id of the assistant message
// that requested this tool call (continuation anchor).
ResponseID string `json:"responseId,omitempty" yaml:"responseId,omitempty"`
}
Step represents a single atomic action in a Plan.
Source Files