cognis 0.3.0

Cognis umbrella crate: agent builder, multi-agent orchestration, memory, middleware (rate limit, retry, PII, prompt caching), built-in tools, and re-exports of cognis-core, cognis-graph, cognis-llm, and cognis-rag.
Documentation 
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
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147

//! Retry the LLM call with exponential backoff on retryable errors.

use std::sync::Arc;
use std::time::Duration;

use async_trait::async_trait;

use cognis_core::Result;
use cognis_llm::chat::ChatResponse;

use super::{Middleware, MiddlewareCtx, Next};

/// Retry the underlying LLM call when it fails with a retryable
/// `CognisError` (rate-limited, network, timeout, transient provider error).
pub struct ModelRetry {
    max_attempts: u32,
    initial_delay: Duration,
    multiplier: f64,
    max_delay: Duration,
}

impl ModelRetry {
    /// Build with default backoff: 100ms initial, 2x multiplier, 30s cap.
    pub fn new(max_attempts: u32) -> Self {
        Self {
            max_attempts,
            initial_delay: Duration::from_millis(100),
            multiplier: 2.0,
            max_delay: Duration::from_secs(30),
        }
    }

    /// Override the initial delay.
    pub fn with_initial_delay(mut self, d: Duration) -> Self {
        self.initial_delay = d;
        self
    }

    /// Override the backoff multiplier.
    pub fn with_multiplier(mut self, m: f64) -> Self {
        self.multiplier = m;
        self
    }

    /// Override the cap on per-attempt delay.
    pub fn with_max_delay(mut self, d: Duration) -> Self {
        self.max_delay = d;
        self
    }
}

#[async_trait]
impl Middleware for ModelRetry {
    async fn call(&self, ctx: MiddlewareCtx, next: Arc<dyn Next>) -> Result<ChatResponse> {
        let mut delay = self.initial_delay;
        let mut last_err = None;
        for attempt in 0..self.max_attempts {
            match next.invoke(ctx.clone()).await {
                Ok(r) => return Ok(r),
                Err(e) if !e.is_retryable() => return Err(e),
                Err(e) => {
                    let suggested = e.retry_delay().unwrap_or(delay);
                    last_err = Some(e);
                    if attempt + 1 >= self.max_attempts {
                        break;
                    }
                    let sleep_for = suggested.min(self.max_delay);
                    tokio::time::sleep(sleep_for).await;
                    delay = Duration::from_secs_f64(
                        (delay.as_secs_f64() * self.multiplier).min(self.max_delay.as_secs_f64()),
                    );
                }
            }
        }
        Err(last_err.unwrap_or_else(|| {
            cognis_core::CognisError::Internal("retry exhausted with no error".into())
        }))
    }

    fn name(&self) -> &str {
        "ModelRetry"
    }
}

#[cfg(test)]
mod tests {
    use super::super::tests_util::*;
    use super::*;
    use crate::middleware::MiddlewarePipeline;

    use std::sync::atomic::{AtomicUsize, Ordering};
    use std::sync::Arc;

    use cognis_core::{CognisError, Message};
    use cognis_llm::chat::ChatOptions;
    use cognis_llm::Client;

    #[tokio::test]
    async fn retries_until_success() {
        let attempts = Arc::new(AtomicUsize::new(0));
        let attempts_for_provider = attempts.clone();
        let provider = make_flaky_provider(move |i| {
            attempts_for_provider.store(i + 1, Ordering::SeqCst);
            if i < 2 {
                Err(CognisError::Network {
                    status_code: Some(503),
                    message: "boom".into(),
                })
            } else {
                Ok("ok".into())
            }
        });
        let client = Client::new(provider);
        let pipe = MiddlewarePipeline::new()
            .push(ModelRetry::new(5).with_initial_delay(Duration::from_millis(1)))
            .build(client);
        let r = pipe
            .invoke(
                vec![Message::human("hi")],
                Vec::new(),
                ChatOptions::default(),
            )
            .await
            .unwrap();
        assert_eq!(r.message.content(), "ok");
        assert_eq!(attempts.load(Ordering::SeqCst), 3);
    }

    #[tokio::test]
    async fn non_retryable_propagates() {
        let provider =
            make_flaky_provider(|_| Err(CognisError::AuthenticationFailed("nope".into())));
        let client = Client::new(provider);
        let pipe = MiddlewarePipeline::new()
            .push(ModelRetry::new(5).with_initial_delay(Duration::from_millis(1)))
            .build(client);
        let err = pipe
            .invoke(
                vec![Message::human("hi")],
                Vec::new(),
                ChatOptions::default(),
            )
            .await
            .unwrap_err();
        assert!(matches!(err, CognisError::AuthenticationFailed(_)));
    }
}