cupel 1.2.0

Context window management pipeline for LLM applications
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50

use crate::model::ContextItem;
use crate::scorer::Scorer;

/// Rank-based scorer: more recent items score higher.
///
/// Items without a timestamp receive 0.0. A single timestamped item scores 1.0.
///
/// # Examples
///
/// ```
/// use cupel::{ContextItemBuilder, RecencyScorer, Scorer};
/// use chrono::Utc;
///
/// let item = ContextItemBuilder::new("recent message", 5)
///     .timestamp(Utc::now())
///     .build()?;
///
/// let score = RecencyScorer.score(&item, &[item.clone()]);
/// assert_eq!(score, 1.0); // single timestamped item scores 1.0
/// # Ok::<(), cupel::CupelError>(())
/// ```
#[derive(Debug, Clone, Copy)]
pub struct RecencyScorer;

impl Scorer for RecencyScorer {
    fn score(&self, item: &ContextItem, all_items: &[ContextItem]) -> f64 {
        let item_ts = match item.timestamp() {
            Some(ts) => ts,
            None => return 0.0,
        };

        let mut count_with_timestamp: usize = 0;
        let mut rank: usize = 0;

        for other in all_items {
            if let Some(ts) = other.timestamp() {
                count_with_timestamp += 1;
                if ts < item_ts {
                    rank += 1;
                }
            }
        }

        if count_with_timestamp <= 1 {
            return 1.0;
        }

        rank as f64 / (count_with_timestamp - 1) as f64
    }
}