edgequake-llm 0.6.1

Multi-provider LLM abstraction library with caching, rate limiting, and cost tracking
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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
//! Retry executor for LLM operations with exponential backoff.
//!
//! This module provides a retry executor that applies the appropriate
//! retry strategy based on the error type.
//!
//! # Usage
//!
//! ```ignore
//! use edgequake_llm::retry::RetryExecutor;
//! use edgequake_llm::error::RetryStrategy;
//!
//! let executor = RetryExecutor::new();
//! let result = executor.execute(
//!     &RetryStrategy::network_backoff(),
//!     || async { provider.complete(messages, options).await },
//! ).await;
//! ```
//!
//! @implements specs/improve-tools/006-error-handling.md
//! @iteration OODA-11

use crate::error::{LlmError, RetryStrategy};
use std::future::Future;
use std::time::Duration;
use tokio::time::sleep;
use tracing::{debug, info, warn};

/// Executor for retry logic with configurable backoff strategies.
///
/// The executor wraps async operations and automatically retries them
/// according to the specified retry strategy.
#[derive(Debug, Default)]
pub struct RetryExecutor {
    /// Optional callback for logging retry attempts.
    log_retries: bool,
}

impl RetryExecutor {
    /// Create a new retry executor.
    pub fn new() -> Self {
        Self { log_retries: true }
    }

    /// Create a retry executor without logging.
    pub fn silent() -> Self {
        Self { log_retries: false }
    }

    /// Execute an async operation with automatic retry based on strategy.
    ///
    /// # Arguments
    ///
    /// * `strategy` - The retry strategy to use
    /// * `operation` - Async closure that performs the operation
    ///
    /// # Returns
    ///
    /// The result of the operation, or the last error if all retries fail.
    ///
    /// # Example
    ///
    /// ```ignore
    /// let executor = RetryExecutor::new();
    /// let result = executor.execute(
    ///     &RetryStrategy::network_backoff(),
    ///     || async { make_api_call().await },
    /// ).await;
    /// ```
    pub async fn execute<F, Fut, T>(
        &self,
        strategy: &RetryStrategy,
        mut operation: F,
    ) -> Result<T, LlmError>
    where
        F: FnMut() -> Fut,
        Fut: Future<Output = Result<T, LlmError>>,
    {
        match strategy {
            RetryStrategy::NoRetry => operation().await,

            RetryStrategy::WaitAndRetry { wait } => {
                self.execute_wait_and_retry(*wait, operation).await
            }

            RetryStrategy::ExponentialBackoff {
                base_delay,
                max_delay,
                max_attempts,
            } => {
                self.execute_exponential_backoff(*base_delay, *max_delay, *max_attempts, operation)
                    .await
            }

            RetryStrategy::ReduceContext => {
                // We can't automatically reduce context here.
                // The caller should catch ReduceContext strategy and handle it.
                // Just attempt once.
                operation().await
            }
        }
    }

    /// Execute with wait-and-retry strategy.
    async fn execute_wait_and_retry<F, Fut, T>(
        &self,
        wait: Duration,
        mut operation: F,
    ) -> Result<T, LlmError>
    where
        F: FnMut() -> Fut,
        Fut: Future<Output = Result<T, LlmError>>,
    {
        match operation().await {
            Ok(v) => Ok(v),
            Err(e) => {
                if self.log_retries {
                    warn!("Operation failed, waiting {:?} before retry: {}", wait, e);
                }
                sleep(wait).await;
                operation().await
            }
        }
    }

    /// Execute with exponential backoff strategy.
    async fn execute_exponential_backoff<F, Fut, T>(
        &self,
        base_delay: Duration,
        max_delay: Duration,
        max_attempts: u32,
        mut operation: F,
    ) -> Result<T, LlmError>
    where
        F: FnMut() -> Fut,
        Fut: Future<Output = Result<T, LlmError>>,
    {
        let mut delay = base_delay;
        let mut attempts = 0;

        loop {
            attempts += 1;

            match operation().await {
                Ok(v) => {
                    if attempts > 1 && self.log_retries {
                        info!("Operation succeeded after {} attempts", attempts);
                    }
                    return Ok(v);
                }
                Err(e) => {
                    if attempts >= max_attempts {
                        if self.log_retries {
                            warn!(
                                "Operation failed after {} attempts, giving up: {}",
                                attempts, e
                            );
                        }
                        return Err(e);
                    }

                    // Check if the error itself suggests we shouldn't retry
                    let error_strategy = e.retry_strategy();
                    if matches!(error_strategy, RetryStrategy::NoRetry) {
                        if self.log_retries {
                            debug!("Error is non-retryable, stopping: {}", e);
                        }
                        return Err(e);
                    }

                    if self.log_retries {
                        warn!(
                            "Attempt {}/{} failed, retrying in {:?}: {}",
                            attempts, max_attempts, delay, e
                        );
                    }

                    sleep(delay).await;
                    delay = (delay * 2).min(max_delay);
                }
            }
        }
    }

    /// Execute an operation with automatic strategy detection.
    ///
    /// This variant automatically determines the retry strategy from
    /// the first error that occurs.
    ///
    /// # Example
    ///
    /// ```ignore
    /// let executor = RetryExecutor::new();
    /// let result = executor.execute_auto(|| async {
    ///     make_api_call().await
    /// }).await;
    /// ```
    pub async fn execute_auto<F, Fut, T>(&self, mut operation: F) -> Result<T, LlmError>
    where
        F: FnMut() -> Fut,
        Fut: Future<Output = Result<T, LlmError>>,
    {
        match operation().await {
            Ok(v) => Ok(v),
            Err(e) => {
                let strategy = e.retry_strategy();

                if !strategy.should_retry() {
                    return Err(e);
                }

                if self.log_retries {
                    debug!("First attempt failed, using strategy {:?}: {}", strategy, e);
                }

                // Sleep before next attempt based on strategy
                match &strategy {
                    RetryStrategy::WaitAndRetry { wait } => {
                        sleep(*wait).await;
                        operation().await
                    }
                    RetryStrategy::ExponentialBackoff {
                        base_delay,
                        max_delay,
                        max_attempts,
                    } => {
                        sleep(*base_delay).await;
                        // Continue with remaining attempts
                        self.execute_exponential_backoff(
                            *base_delay * 2, // Already waited base_delay
                            *max_delay,
                            max_attempts - 1, // Already used one attempt
                            operation,
                        )
                        .await
                    }
                    RetryStrategy::ReduceContext | RetryStrategy::NoRetry => {
                        // Shouldn't reach here due to should_retry check
                        Err(e)
                    }
                }
            }
        }
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use std::sync::atomic::{AtomicU32, Ordering};
    use std::sync::Arc;

    #[tokio::test]
    async fn test_no_retry_succeeds() {
        let executor = RetryExecutor::silent();
        let result = executor
            .execute(&RetryStrategy::NoRetry, || async { Ok::<_, LlmError>(42) })
            .await;
        assert_eq!(result.unwrap(), 42);
    }

    #[tokio::test]
    async fn test_no_retry_fails_immediately() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute(&RetryStrategy::NoRetry, || {
                let count = call_count_clone.clone();
                async move {
                    count.fetch_add(1, Ordering::SeqCst);
                    Err::<i32, _>(LlmError::AuthError("bad key".to_string()))
                }
            })
            .await;

        assert!(result.is_err());
        assert_eq!(call_count.load(Ordering::SeqCst), 1);
    }

    #[tokio::test]
    async fn test_exponential_backoff_retries() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute(
                &RetryStrategy::ExponentialBackoff {
                    base_delay: Duration::from_millis(1),
                    max_delay: Duration::from_millis(10),
                    max_attempts: 3,
                },
                || {
                    let count = call_count_clone.clone();
                    async move {
                        let attempts = count.fetch_add(1, Ordering::SeqCst) + 1;
                        if attempts < 3 {
                            Err(LlmError::NetworkError("failed".to_string()))
                        } else {
                            Ok(42)
                        }
                    }
                },
            )
            .await;

        assert_eq!(result.unwrap(), 42);
        assert_eq!(call_count.load(Ordering::SeqCst), 3);
    }

    #[tokio::test]
    async fn test_exponential_backoff_gives_up() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute(
                &RetryStrategy::ExponentialBackoff {
                    base_delay: Duration::from_millis(1),
                    max_delay: Duration::from_millis(10),
                    max_attempts: 3,
                },
                || {
                    let count = call_count_clone.clone();
                    async move {
                        count.fetch_add(1, Ordering::SeqCst);
                        Err::<i32, _>(LlmError::NetworkError("always fails".to_string()))
                    }
                },
            )
            .await;

        assert!(result.is_err());
        assert_eq!(call_count.load(Ordering::SeqCst), 3);
    }

    #[tokio::test]
    async fn test_wait_and_retry() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute(
                &RetryStrategy::WaitAndRetry {
                    wait: Duration::from_millis(1),
                },
                || {
                    let count = call_count_clone.clone();
                    async move {
                        let attempts = count.fetch_add(1, Ordering::SeqCst) + 1;
                        if attempts < 2 {
                            Err(LlmError::RateLimited("wait".to_string()))
                        } else {
                            Ok(42)
                        }
                    }
                },
            )
            .await;

        assert_eq!(result.unwrap(), 42);
        assert_eq!(call_count.load(Ordering::SeqCst), 2);
    }

    #[tokio::test]
    async fn test_stops_on_permanent_error() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute(
                &RetryStrategy::ExponentialBackoff {
                    base_delay: Duration::from_millis(1),
                    max_delay: Duration::from_millis(10),
                    max_attempts: 5,
                },
                || {
                    let count = call_count_clone.clone();
                    async move {
                        count.fetch_add(1, Ordering::SeqCst);
                        // Return a non-retryable error
                        Err::<i32, _>(LlmError::AuthError("invalid".to_string()))
                    }
                },
            )
            .await;

        assert!(result.is_err());
        // Should stop after first attempt since AuthError is non-retryable
        assert_eq!(call_count.load(Ordering::SeqCst), 1);
    }

    #[tokio::test]
    async fn test_reduce_context_executes_once() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute(&RetryStrategy::ReduceContext, || {
                let count = call_count_clone.clone();
                async move {
                    count.fetch_add(1, Ordering::SeqCst);
                    Err::<i32, _>(LlmError::TokenLimitExceeded {
                        max: 4096,
                        got: 5000,
                    })
                }
            })
            .await;

        assert!(result.is_err());
        assert_eq!(call_count.load(Ordering::SeqCst), 1);
    }

    #[tokio::test]
    async fn test_reduce_context_success() {
        let executor = RetryExecutor::silent();

        let result = executor
            .execute(&RetryStrategy::ReduceContext, || async {
                Ok::<_, LlmError>(42)
            })
            .await;

        assert_eq!(result.unwrap(), 42);
    }

    #[tokio::test]
    async fn test_execute_auto_success() {
        let executor = RetryExecutor::new();

        let result = executor
            .execute_auto(|| async { Ok::<_, LlmError>(99) })
            .await;

        assert_eq!(result.unwrap(), 99);
    }

    #[tokio::test]
    async fn test_execute_auto_non_retryable_error() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute_auto(|| {
                let count = call_count_clone.clone();
                async move {
                    count.fetch_add(1, Ordering::SeqCst);
                    Err::<i32, _>(LlmError::AuthError("invalid".to_string()))
                }
            })
            .await;

        assert!(result.is_err());
        assert_eq!(call_count.load(Ordering::SeqCst), 1);
    }

    #[tokio::test]
    async fn test_execute_auto_retryable_network_error() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute_auto(|| {
                let count = call_count_clone.clone();
                async move {
                    let attempts = count.fetch_add(1, Ordering::SeqCst) + 1;
                    if attempts < 2 {
                        Err(LlmError::NetworkError("failed".to_string()))
                    } else {
                        Ok(42)
                    }
                }
            })
            .await;

        assert_eq!(result.unwrap(), 42);
        // First call fails, then auto-retry succeeds
        assert!(call_count.load(Ordering::SeqCst) >= 2);
    }

    #[tokio::test]
    async fn test_wait_and_retry_both_fail() {
        let executor = RetryExecutor::silent();
        let call_count = Arc::new(AtomicU32::new(0));
        let call_count_clone = call_count.clone();

        let result = executor
            .execute(
                &RetryStrategy::WaitAndRetry {
                    wait: Duration::from_millis(1),
                },
                || {
                    let count = call_count_clone.clone();
                    async move {
                        count.fetch_add(1, Ordering::SeqCst);
                        Err::<i32, _>(LlmError::RateLimited("wait".to_string()))
                    }
                },
            )
            .await;

        assert!(result.is_err());
        assert_eq!(call_count.load(Ordering::SeqCst), 2);
    }

    #[tokio::test]
    async fn test_new_constructor_logging() {
        let executor = RetryExecutor::new();
        assert!(executor.log_retries);
    }

    #[tokio::test]
    async fn test_silent_constructor() {
        let executor = RetryExecutor::silent();
        assert!(!executor.log_retries);
    }

    #[tokio::test]
    async fn test_default_constructor() {
        let executor = RetryExecutor::default();
        assert!(!executor.log_retries);
    }
}