pub struct TimeoutConfig {
pub llm_seconds: u64,
pub llm_request_timeout_secs: u64,
pub embedding_seconds: u64,
pub a2a_seconds: u64,
pub max_parallel_tools: usize,
pub context_prep_timeout_secs: u64,
pub no_providers_backoff_secs: u64,
}Expand description
Timeout configuration for external operations, nested under [timeouts] in TOML.
All timeouts are in seconds. Exceeding a timeout returns an error to the agent loop rather than blocking indefinitely.
§Example (TOML)
[timeouts]
llm_seconds = 60
embedding_seconds = 15
max_parallel_tools = 4Fields§
§llm_seconds: u64Timeout for streaming LLM first-token responses, in seconds. Default: 120.
llm_request_timeout_secs: u64Total wall-clock timeout for a complete LLM request (all tokens), in seconds.
Default: 600.
embedding_seconds: u64Timeout for embedding API calls, in seconds. Default: 30.
a2a_seconds: u64Timeout for A2A agent-to-agent calls, in seconds. Default: 30.
max_parallel_tools: usizeMaximum number of tool calls that may execute concurrently in a single turn.
Default: 8.
context_prep_timeout_secs: u64Maximum wall-clock time (seconds) allowed for advance_context_lifecycle (memory recall,
graph retrieval, proactive compression, context assembly) before it is aborted and the
agent proceeds with a degraded (cached) context.
Setting this too low may skip useful memory recall; setting it too high blocks the agent
when embed providers are rate-limited or unavailable. Default: 30.
no_providers_backoff_secs: u64How long to wait (seconds) before retrying a turn after the previous turn ended with
no providers available. Prevents a busy-wait loop when all LLM backends are down.
Default: 2.
Trait Implementations§
Source§impl Clone for TimeoutConfig
impl Clone for TimeoutConfig
Source§fn clone(&self) -> TimeoutConfig
fn clone(&self) -> TimeoutConfig
1.0.0 · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
source. Read more