Struct AgentConfig 

pub struct AgentConfig {

    pub api_base: String,
    pub api_key: String,
    pub model: String,
    pub max_iterations: usize,
    pub max_tool_calls_per_task: usize,
    pub workspace: String,
    pub system_prompt: Option<String>,
    pub temperature: Option<f64>,
    pub skill_dirs: Vec<String>,
    pub enable_task_planning: bool,
    pub enable_memory: bool,
    pub enable_memory_vector: bool,
    pub verbose: bool,
    pub soul_path: Option<String>,
    pub context_append: Option<String>,
    pub max_consecutive_failures: Option<usize>,
    pub goal_boundaries: Option<GoalBoundaries>,
    pub skip_history_for_planning: bool,
    pub read_only_tools: bool,
}

Agent configuration.

Fields

api_base: String

OpenAI-compatible API base URL (e.g. “https://api.openai.com/v1”)

api_key: String

API key

model: String

Model name (e.g. “gpt-4o”, “claude-3-5-sonnet-20241022”)

max_iterations: usize

Maximum iterations for the agent loop

max_tool_calls_per_task: usize

Maximum tool calls per task

workspace: String

Workspace root path

system_prompt: Option<String>

System prompt override (optional)

temperature: Option<f64>

Temperature (0.0 - 2.0)

skill_dirs: Vec<String>

Skills directories to load (reserved for multi-dir support)

enable_task_planning: bool

Enable task planning

enable_memory: bool

Enable memory tools

enable_memory_vector: bool

Enable memory vector search (requires memory_vector feature + embedding API)

verbose: bool

Verbose output

soul_path: Option<String>

Path to SOUL.md identity document (optional). Resolution: explicit path > .skilllite/SOUL.md > ~/.skilllite/SOUL.md

context_append: Option<String>

Optional extra context to append to system prompt (e.g. from RPC params.context.append). Generic extension point for callers to inject domain-specific rules without modifying SkillLite.

max_consecutive_failures: Option<usize>

[Run mode] Max consecutive tool failures before stopping (prevents infinite retry loops). None = no limit (chat mode). Some(N) = stop after N consecutive failures (run mode).

goal_boundaries: Option<GoalBoundaries>

[Run mode] Extracted goal boundaries (scope, exclusions, completion conditions). Injected into planning when set.

skip_history_for_planning: bool

When true, exclude conversation history from the planning prompt. Use when each task is self-contained and history from previous turns would corrupt planning (e.g. multi-turn agent orchestration, batch task dispatch). Default: false.

read_only_tools: bool

Restrict the tool registry to read-only operations. Used by replay/eval flows that must not mutate the workspace.

Implementations

impl AgentConfig

pub fn from_env() -> Self

Load from environment variables with sensible defaults. Also reads .env file from current directory if present. Uses unified config layer: SKILLLITE_* with fallback to OPENAI_* / BASE_URL / API_KEY / MODEL.

Trait Implementations

impl Clone for AgentConfig

fn clone(&self) -> AgentConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AgentConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for AgentConfig

fn default() -> Self

Returns the “default value” for a type.