Struct AgentConfig 

pub struct AgentConfig {

    pub compact_context: bool,
    pub max_tool_iterations: usize,
    pub max_history_messages: usize,
    pub max_context_tokens: usize,
    pub parallel_tools: bool,
    pub tool_dispatcher: String,
    pub tool_call_dedup_exempt: Vec<String>,
    pub tool_filter_groups: Vec<ToolFilterGroup>,
    pub max_system_prompt_chars: usize,
    pub thinking: ThinkingConfig,
    pub history_pruning: HistoryPrunerConfig,
    pub context_aware_tools: bool,
    pub eval: EvalConfig,
    pub auto_classify: Option<AutoClassifyConfig>,
    pub context_compression: ContextCompressionConfig,
    pub max_tool_result_chars: usize,
    pub keep_tool_context_turns: usize,
}

Agent orchestration configuration ([agent] section).

Fields

compact_context: bool

When true: bootstrap_max_chars=6000, rag_chunk_limit=2. Use for 13B or smaller models.

max_tool_iterations: usize

Maximum tool-call loop turns per user message. Default: 10. Setting to 0 falls back to the safe default of 10.

max_history_messages: usize

Maximum conversation history messages retained per session. Default: 50.

max_context_tokens: usize

Maximum estimated tokens for conversation history before compaction triggers. Uses ~4 chars/token heuristic. When this threshold is exceeded, older messages are summarized to preserve context while staying within budget. Default: 32000.

parallel_tools: bool

Enable parallel tool execution within a single iteration. Default: false.

tool_dispatcher: String

Tool dispatch strategy (e.g. "auto"). Default: "auto".

tool_call_dedup_exempt: Vec<String>

Tools exempt from the within-turn duplicate-call dedup check. Default: [].

tool_filter_groups: Vec<ToolFilterGroup>

Per-turn MCP tool schema filtering groups.

When non-empty, only MCP tools matched by an active group are included in the tool schema sent to the LLM for that turn. Built-in tools always pass through. Default: [] (no filtering — all tools included).

max_system_prompt_chars: usize

Maximum characters for the assembled system prompt. When > 0, the prompt is truncated to this limit after assembly (keeping the top portion which contains identity and safety instructions). 0 means unlimited. Useful for small-context models (e.g. glm-4.5-air ~8K tokens → set to 8000).

thinking: ThinkingConfig

Thinking/reasoning level control. Configures how deeply the model reasons per message. Users can override per-message with /think:<level> directives.

history_pruning: HistoryPrunerConfig

History pruning configuration for token efficiency.

context_aware_tools: bool

Enable context-aware tool filtering (only surface relevant tools per iteration).

eval: EvalConfig

Post-response quality evaluator configuration.

auto_classify: Option<AutoClassifyConfig>

Automatic complexity-based classification fallback.

context_compression: ContextCompressionConfig

Context compression configuration for automatic conversation compaction.

max_tool_result_chars: usize

Maximum characters for a single tool result before truncation. Head (2/3) and tail (1/3) are preserved with a truncation marker in the middle. Set to 0 to disable truncation. Default: 50000.

keep_tool_context_turns: usize

Number of most recent conversation turns whose full tool-call/result messages are preserved in channel conversation history. Older turns keep only the final assistant text. Set to 0 to disable (previous behavior). Default: 2.

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.

impl<'de> Deserialize<'de> for AgentConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl JsonSchema for AgentConfig

fn schema_name() -> Cow<'static, str>

The name of the generated JSON Schema.

fn schema_id() -> Cow<'static, str>

Returns a string that uniquely identifies the schema produced by this type.

fn json_schema(generator: &mut SchemaGenerator) -> Schema

Generates a JSON Schema for this type.

fn inline_schema() -> bool

Whether JSON Schemas generated for this type should be included directly in parent schemas, rather than being re-used where possible using the $ref keyword.

impl Serialize for AgentConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.