Struct RlmCompletion 

pub struct RlmCompletion {
    pub trace_id: Uuid,
    pub outcome: RlmOutcome,
    pub iterations: usize,
    pub subcalls: usize,
    pub input_tokens: usize,
    pub output_tokens: usize,
    pub elapsed_ms: u64,
    pub reason: Option<String>,
    pub root_model: String,
    pub subcall_model_used: Option<String>,
}

Terminal record for a single RLM invocation.

Persisted durably so cost post-mortems and trace-driven tuning jobs can reconstruct how each compaction or big-tool-output run behaved.

Examples

use codetether_agent::session::{RlmCompletion, RlmOutcome};
use uuid::Uuid;

let c = RlmCompletion {
    trace_id: Uuid::nil(),
    outcome: RlmOutcome::Converged,
    iterations: 4,
    subcalls: 1,
    input_tokens: 42_000,
    output_tokens: 2_100,
    elapsed_ms: 3_412,
    reason: None,
    root_model: "zai/glm-5".into(),
    subcall_model_used: None,
};
assert!(c.compression_ratio() > 0.0);
assert!(c.compression_ratio() < 1.0);

Fields

trace_id: Uuid

Correlates this completion with its progress events.

outcome: RlmOutcome

Why the loop finished.

iterations: usize

Iterations actually executed.

subcalls: usize

Sub-LLM calls issued during the loop.

input_tokens: usize

Input tokens (the raw blob handed to the router).

output_tokens: usize

Output tokens (the produced summary / final answer).

elapsed_ms: u64

Total wall-clock duration in milliseconds.

reason: Option<String>

Human-readable explanation for non-RlmOutcome::Converged outcomes.

root_model: String

Model used for iteration 1 (the root model).

subcall_model_used: Option<String>

Model used for iterations ≥ 2, if RlmConfig::subcall_model was configured and resolved. None means all iterations used Self::root_model.

Implementations

impl RlmCompletion

pub fn compression_ratio(&self) -> f64

output_tokens / input_tokens, clamped so zero-input loops return 0.0 rather than NaN.

Values below 1.0 indicate compression; above 1.0 indicate expansion (rare, but possible when the model rewrites a terse blob into a verbose explanation).

Examples

use codetether_agent::session::{RlmCompletion, RlmOutcome};
use uuid::Uuid;

let c = RlmCompletion {
    trace_id: Uuid::nil(),
    outcome: RlmOutcome::Converged,
    iterations: 1, subcalls: 0,
    input_tokens: 1000, output_tokens: 100,
    elapsed_ms: 0, reason: None,
    root_model: "m".into(), subcall_model_used: None,
};
assert!((c.compression_ratio() - 0.1).abs() < 1e-9);

Trait Implementations

impl Clone for RlmCompletion

fn clone(&self) -> RlmCompletion

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RlmCompletion

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for RlmCompletion

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for RlmCompletion

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

where __S: Serializer,

Serialize this value into the given Serde serializer.