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
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775//! LLM error classification and sanitization.
//!
//! Classifies raw LLM API errors into 8 categories using pattern matching
//! against error messages and HTTP status codes. Handles error formats from
//! all 19+ providers OpenFang supports: Anthropic, OpenAI, Gemini, Groq,
//! DeepSeek, Mistral, Together, Fireworks, Ollama, vLLM, LM Studio,
//! Perplexity, Cohere, AI21, Cerebras, SambaNova, HuggingFace, XAI, Replicate.
//!
//! Pattern matching is done via case-insensitive substring checks with no
//! external regex dependency, keeping the crate dependency graph lean.
use serde::Serialize;
// ---------------------------------------------------------------------------
// Types
// ---------------------------------------------------------------------------
/// Classified LLM error category.
#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize)]
pub enum LlmErrorCategory {
/// 429, quota exceeded, too many requests.
RateLimit,
/// 503, overloaded, service unavailable, high demand.
Overloaded,
/// Request timeout, deadline exceeded, ETIMEDOUT, ECONNRESET.
Timeout,
/// 402, payment required, insufficient credits/balance.
Billing,
/// 401/403, invalid API key, unauthorized, forbidden.
Auth,
/// Context length exceeded, max tokens, context window.
ContextOverflow,
/// Invalid request format, malformed tool_use, schema violation.
Format,
/// Model not found, unknown model, NOT_FOUND.
ModelNotFound,
}
/// Classified error with metadata.
#[derive(Debug, Clone, Serialize)]
pub struct ClassifiedError {
/// The classified category.
pub category: LlmErrorCategory,
/// `true` for RateLimit, Overloaded, Timeout.
pub is_retryable: bool,
/// `true` only for Billing.
pub is_billing: bool,
/// Retry delay parsed from the error message, if available.
pub suggested_delay_ms: Option<u64>,
/// User-safe message (no raw API details).
pub sanitized_message: String,
/// Original error message for logging.
pub raw_message: String,
}
// ---------------------------------------------------------------------------
// Pattern tables (case-insensitive substring checks)
// ---------------------------------------------------------------------------
/// Context overflow patterns -- checked first because they are highly specific.
const CONTEXT_OVERFLOW_PATTERNS: &[&str] = &[
"context_length_exceeded",
"context length",
"context_length",
"maximum context",
"context window",
"token limit",
"too many tokens",
"max_tokens_exceeded",
"max tokens exceeded",
"prompt is too long",
"input too long",
"context.length",
];
/// Billing patterns.
const BILLING_PATTERNS: &[&str] = &[
"payment required",
"insufficient credits",
"credit balance",
"billing",
"insufficient balance",
"usage limit",
];
/// Auth patterns.
const AUTH_PATTERNS: &[&str] = &[
"invalid api key",
"invalid api_key",
"invalid apikey",
"incorrect api key",
"invalid token",
"unauthorized",
"forbidden",
"authentication",
"permission denied",
];
/// Rate-limit patterns.
const RATE_LIMIT_PATTERNS: &[&str] = &[
"rate limit",
"rate_limit",
"too many requests",
"exceeded quota",
"exceeded your quota",
"resource exhausted",
"resource_exhausted",
"quota exceeded",
"tokens per minute",
"requests per minute",
"tpm limit",
"rpm limit",
];
/// Model-not-found patterns.
const MODEL_NOT_FOUND_PATTERNS: &[&str] = &[
"model not found",
"model_not_found",
"unknown model",
"does not exist",
"not_found",
"model unavailable",
"model_unavailable",
"no such model",
"invalid model",
"is not found",
];
/// Format / bad-request patterns (catch-all for 400-class issues).
const FORMAT_PATTERNS: &[&str] = &[
"invalid request",
"invalid_request",
"malformed",
"tool_use",
"schema",
"validation error",
"validation_error",
"invalid parameter",
"invalid_parameter",
"missing required",
"bad request",
"bad_request",
];
/// Overloaded patterns.
const OVERLOADED_PATTERNS: &[&str] = &[
"overloaded",
"overloaded_error",
"service unavailable",
"service_unavailable",
"high demand",
"capacity",
"server_error",
"internal server error",
"internal_server_error",
];
/// Timeout / network patterns.
const TIMEOUT_PATTERNS: &[&str] = &[
"timeout",
"timed out",
"deadline exceeded",
"etimedout",
"econnreset",
"econnrefused",
"econnaborted",
"epipe",
"ehostunreach",
"enetunreach",
"connection reset",
"connection refused",
"network error",
"fetch failed",
];
// ---------------------------------------------------------------------------
// Classification
// ---------------------------------------------------------------------------
/// Check if `haystack` (lowercased) contains any pattern from `patterns`.
fn matches_any(haystack: &str, patterns: &[&str]) -> bool {
patterns.iter().any(|p| haystack.contains(p))
}
/// Classify a raw error message + optional HTTP status into a category.
///
/// Priority order (most specific first):
/// 1. ContextOverflow 2. Billing (402) 3. Auth (401/403)
/// 4. RateLimit (429) 5. ModelNotFound 6. Format (400)
/// 7. Overloaded (503/500) 8. Timeout (network)
///
/// If nothing matches, falls back to `Format` for structured errors or
/// `Timeout` for network-sounding messages.
pub fn classify_error(message: &str, status: Option<u16>) -> ClassifiedError {
let lower = message.to_lowercase();
let delay = extract_retry_delay(message);
// Helper to build ClassifiedError
let build = |category: LlmErrorCategory| ClassifiedError {
category,
is_retryable: matches!(
category,
LlmErrorCategory::RateLimit | LlmErrorCategory::Overloaded | LlmErrorCategory::Timeout
),
is_billing: category == LlmErrorCategory::Billing,
suggested_delay_ms: delay,
sanitized_message: sanitize_for_user(category, message),
raw_message: message.to_string(),
};
// --- Status-code fast paths (some statuses are unambiguous) ---
if let Some(code) = status {
match code {
429 => return build(LlmErrorCategory::RateLimit),
402 => return build(LlmErrorCategory::Billing),
401 => return build(LlmErrorCategory::Auth),
403 => {
// 403 could be auth OR rate-limit on some providers; check message
if matches_any(&lower, RATE_LIMIT_PATTERNS) {
return build(LlmErrorCategory::RateLimit);
}
return build(LlmErrorCategory::Auth);
}
_ => {}
}
}
// --- Pattern matching in priority order ---
// 1. Context overflow (very specific patterns)
if matches_any(&lower, CONTEXT_OVERFLOW_PATTERNS) {
return build(LlmErrorCategory::ContextOverflow);
}
// 2. Billing
if matches_any(&lower, BILLING_PATTERNS) {
return build(LlmErrorCategory::Billing);
}
if status == Some(402) {
return build(LlmErrorCategory::Billing);
}
// 3. Auth
if matches_any(&lower, AUTH_PATTERNS) {
return build(LlmErrorCategory::Auth);
}
if matches!(status, Some(401) | Some(403)) {
return build(LlmErrorCategory::Auth);
}
// 4. Rate limit
if matches_any(&lower, RATE_LIMIT_PATTERNS) {
return build(LlmErrorCategory::RateLimit);
}
if status == Some(429) {
return build(LlmErrorCategory::RateLimit);
}
// 5. Model not found
if matches_any(&lower, MODEL_NOT_FOUND_PATTERNS) {
return build(LlmErrorCategory::ModelNotFound);
}
// Composite check: "model" + "not found" anywhere in the message
if lower.contains("model") && lower.contains("not found") {
return build(LlmErrorCategory::ModelNotFound);
}
// 6. Format / bad request (before overloaded, since 400 is more specific)
if matches_any(&lower, FORMAT_PATTERNS) {
return build(LlmErrorCategory::Format);
}
if status == Some(400) {
return build(LlmErrorCategory::Format);
}
// 7. Overloaded
if matches_any(&lower, OVERLOADED_PATTERNS) {
return build(LlmErrorCategory::Overloaded);
}
if matches!(status, Some(500) | Some(503)) {
return build(LlmErrorCategory::Overloaded);
}
// 8. Timeout / network
if matches_any(&lower, TIMEOUT_PATTERNS) {
return build(LlmErrorCategory::Timeout);
}
// --- HTML error page detection (Cloudflare etc.) ---
if is_html_error_page(message) {
return build(LlmErrorCategory::Overloaded);
}
// --- Fallback ---
// If there's a status code in the 5xx range, treat as overloaded.
if let Some(code) = status {
if (500..600).contains(&code) {
return build(LlmErrorCategory::Overloaded);
}
if (400..500).contains(&code) {
return build(LlmErrorCategory::Format);
}
}
// Last resort: if the message mentions network-like terms, call it timeout;
// otherwise default to format (unknown structured error).
if lower.contains("connect") || lower.contains("network") || lower.contains("dns") {
build(LlmErrorCategory::Timeout)
} else {
build(LlmErrorCategory::Format)
}
}
// ---------------------------------------------------------------------------
// Sanitization
// ---------------------------------------------------------------------------
/// Produce a user-friendly error message.
///
/// Maps each category to a human-readable description, capped at 200 chars.
pub fn sanitize_for_user(category: LlmErrorCategory, _raw: &str) -> String {
let msg = match category {
LlmErrorCategory::RateLimit => {
"The AI provider is rate-limiting requests. Retrying shortly..."
}
LlmErrorCategory::Overloaded => "The AI provider is temporarily overloaded. Retrying...",
LlmErrorCategory::Timeout => "The request timed out. Check your network connection.",
LlmErrorCategory::Billing => "Billing issue with the AI provider. Check your API plan.",
LlmErrorCategory::Auth => "Authentication failed. Check your API key configuration.",
LlmErrorCategory::ContextOverflow => {
"The conversation is too long for the model's context window."
}
LlmErrorCategory::Format => {
"LLM request failed. Check your API key and model configuration in Settings."
}
LlmErrorCategory::ModelNotFound => {
"The requested model was not found. Check the model name."
}
};
// Cap at 200 chars (all built-in messages are under 200, but defensive).
if msg.chars().count() > 200 {
let end = msg
.char_indices()
.nth(197)
.map(|(i, _)| i)
.unwrap_or(msg.len());
format!("{}...", &msg[..end])
} else {
msg.to_string()
}
}
// ---------------------------------------------------------------------------
// Retry-After extraction
// ---------------------------------------------------------------------------
/// Try to extract a retry delay (in milliseconds) from the error message.
///
/// Recognizes patterns like:
/// - `retry after 30` (seconds)
/// - `retry-after: 30` (seconds)
/// - `try again in 30` (seconds)
/// - `retry after 500ms` (milliseconds)
///
/// Returns `None` if no recognizable delay is found.
pub fn extract_retry_delay(message: &str) -> Option<u64> {
let lower = message.to_lowercase();
// Patterns to search for, each followed by a number.
const PREFIXES: &[&str] = &["retry after ", "retry-after: ", "try again in "];
for prefix in PREFIXES {
if let Some(start) = lower.find(prefix) {
let after = &lower[start + prefix.len()..];
// Parse the leading digits.
let num_str: String = after.chars().take_while(|c| c.is_ascii_digit()).collect();
if let Ok(value) = num_str.parse::<u64>() {
if value == 0 {
continue;
}
// Check for "ms" suffix (milliseconds).
let rest = &after[num_str.len()..];
if rest.starts_with("ms") {
return Some(value);
}
// Default: treat as seconds, convert to ms.
return Some(value.saturating_mul(1000));
}
}
}
None
}
// ---------------------------------------------------------------------------
// Transient error detection
// ---------------------------------------------------------------------------
/// Check if an error is likely transient (network hiccup, temporary overload).
///
/// This is a quick heuristic that does not require full classification.
pub fn is_transient(message: &str) -> bool {
let lower = message.to_lowercase();
matches_any(&lower, TIMEOUT_PATTERNS)
|| matches_any(&lower, OVERLOADED_PATTERNS)
|| matches_any(&lower, RATE_LIMIT_PATTERNS)
}
// ---------------------------------------------------------------------------
// HTML / Cloudflare error detection
// ---------------------------------------------------------------------------
/// Detect if the response body is a Cloudflare error page or raw HTML
/// instead of expected JSON.
///
/// Checks for: `<!DOCTYPE`, `<html`, Cloudflare error codes (521-530),
/// `cf-error-code`.
pub fn is_html_error_page(body: &str) -> bool {
let lower = body.to_lowercase();
// HTML markers
if lower.contains("<!doctype") || lower.contains("<html") {
return true;
}
// Cloudflare error code header/attribute
if lower.contains("cf-error-code") || lower.contains("cf-error-type") {
return true;
}
// Cloudflare error status codes in text (e.g., "Error 522" or "522:")
for code in 521..=530 {
let code_str = code.to_string();
if lower.contains(&code_str) && lower.contains("cloudflare") {
return true;
}
}
false
}
// ===========================================================================
// Tests
// ===========================================================================
#[cfg(test)]
mod tests {
use super::*;
// -----------------------------------------------------------------------
// Classification tests
// -----------------------------------------------------------------------
#[test]
fn test_classify_rate_limit() {
// Standard 429
let e = classify_error("Too Many Requests", Some(429));
assert_eq!(e.category, LlmErrorCategory::RateLimit);
assert!(e.is_retryable);
// Pattern: "rate limit"
let e = classify_error("You have hit the rate limit for this API", None);
assert_eq!(e.category, LlmErrorCategory::RateLimit);
// Pattern: "quota exceeded"
let e = classify_error("Resource exhausted: quota exceeded", None);
assert_eq!(e.category, LlmErrorCategory::RateLimit);
// Pattern: "tokens per minute"
let e = classify_error("You exceeded your tokens per minute limit", None);
assert_eq!(e.category, LlmErrorCategory::RateLimit);
// Pattern: "RPM"
let e = classify_error("RPM limit reached, slow down", None);
assert_eq!(e.category, LlmErrorCategory::RateLimit);
}
#[test]
fn test_classify_overloaded() {
let e = classify_error("The server is currently overloaded", None);
assert_eq!(e.category, LlmErrorCategory::Overloaded);
assert!(e.is_retryable);
let e = classify_error("Service unavailable due to high demand", None);
assert_eq!(e.category, LlmErrorCategory::Overloaded);
// Status 503
let e = classify_error("Please try again later", Some(503));
assert_eq!(e.category, LlmErrorCategory::Overloaded);
// Status 500
let e = classify_error("Something went wrong", Some(500));
assert_eq!(e.category, LlmErrorCategory::Overloaded);
}
#[test]
fn test_classify_timeout() {
let e = classify_error("ETIMEDOUT: request timed out", None);
assert_eq!(e.category, LlmErrorCategory::Timeout);
assert!(e.is_retryable);
let e = classify_error("ECONNRESET", None);
assert_eq!(e.category, LlmErrorCategory::Timeout);
let e = classify_error("ECONNREFUSED: connection refused", None);
assert_eq!(e.category, LlmErrorCategory::Timeout);
let e = classify_error("fetch failed: network error", None);
assert_eq!(e.category, LlmErrorCategory::Timeout);
let e = classify_error("deadline exceeded while waiting for response", None);
assert_eq!(e.category, LlmErrorCategory::Timeout);
}
#[test]
fn test_classify_billing() {
let e = classify_error("Payment required", Some(402));
assert_eq!(e.category, LlmErrorCategory::Billing);
assert!(e.is_billing);
assert!(!e.is_retryable);
let e = classify_error("Insufficient credits in your account", None);
assert_eq!(e.category, LlmErrorCategory::Billing);
let e = classify_error("Your credit balance is too low", None);
assert_eq!(e.category, LlmErrorCategory::Billing);
}
#[test]
fn test_classify_auth() {
let e = classify_error("Invalid API key provided", Some(401));
assert_eq!(e.category, LlmErrorCategory::Auth);
assert!(!e.is_retryable);
let e = classify_error("Forbidden: you do not have access", Some(403));
assert_eq!(e.category, LlmErrorCategory::Auth);
let e = classify_error("Incorrect API key format", None);
assert_eq!(e.category, LlmErrorCategory::Auth);
let e = classify_error("Authentication failed for this endpoint", None);
assert_eq!(e.category, LlmErrorCategory::Auth);
}
#[test]
fn test_classify_context_overflow() {
let e = classify_error("This model's maximum context length is 128000 tokens", None);
assert_eq!(e.category, LlmErrorCategory::ContextOverflow);
let e = classify_error("context_length_exceeded", Some(400));
assert_eq!(e.category, LlmErrorCategory::ContextOverflow);
let e = classify_error("prompt is too long for the context window", None);
assert_eq!(e.category, LlmErrorCategory::ContextOverflow);
let e = classify_error("input too long: exceeds maximum context", None);
assert_eq!(e.category, LlmErrorCategory::ContextOverflow);
}
#[test]
fn test_classify_format() {
let e = classify_error("Invalid request: missing 'messages' field", None);
assert_eq!(e.category, LlmErrorCategory::Format);
let e = classify_error("Malformed JSON in request body", None);
assert_eq!(e.category, LlmErrorCategory::Format);
let e = classify_error("Validation error: tool_use block missing id", None);
assert_eq!(e.category, LlmErrorCategory::Format);
// Status 400 without more specific patterns
let e = classify_error("Something is wrong with your request", Some(400));
assert_eq!(e.category, LlmErrorCategory::Format);
}
#[test]
fn test_classify_model_not_found() {
let e = classify_error("Model 'gpt-5-ultra' not found", None);
assert_eq!(e.category, LlmErrorCategory::ModelNotFound);
let e = classify_error("The model does not exist or you lack access", None);
assert_eq!(e.category, LlmErrorCategory::ModelNotFound);
let e = classify_error("Unknown model: claude-99", None);
assert_eq!(e.category, LlmErrorCategory::ModelNotFound);
}
#[test]
fn test_status_code_override() {
// Even though message says "overloaded", status 429 wins
let e = classify_error("server overloaded", Some(429));
assert_eq!(e.category, LlmErrorCategory::RateLimit);
// Status 402 overrides message
let e = classify_error("something generic happened", Some(402));
assert_eq!(e.category, LlmErrorCategory::Billing);
// Status 401 overrides message
let e = classify_error("generic error text", Some(401));
assert_eq!(e.category, LlmErrorCategory::Auth);
}
#[test]
fn test_retryable_categories() {
// Retryable
assert!(classify_error("rate limit", None).is_retryable);
assert!(classify_error("overloaded", None).is_retryable);
assert!(classify_error("timeout", None).is_retryable);
// Not retryable
assert!(!classify_error("", Some(402)).is_retryable); // Billing
assert!(!classify_error("", Some(401)).is_retryable); // Auth
assert!(!classify_error("context_length_exceeded", None).is_retryable); // ContextOverflow
assert!(!classify_error("model not found", None).is_retryable); // ModelNotFound
}
#[test]
fn test_billing_flag() {
let e = classify_error("payment required", Some(402));
assert!(e.is_billing);
let e = classify_error("rate limit exceeded", None);
assert!(!e.is_billing);
let e = classify_error("insufficient credits", None);
assert!(e.is_billing);
}
#[test]
fn test_sanitize_messages() {
let msg = sanitize_for_user(LlmErrorCategory::RateLimit, "raw error details here");
assert!(msg.contains("rate-limiting"));
assert!(!msg.contains("raw error"));
let msg = sanitize_for_user(LlmErrorCategory::Auth, "sk-xxxx invalid");
assert!(msg.contains("Authentication"));
assert!(!msg.contains("sk-xxxx"));
let msg = sanitize_for_user(LlmErrorCategory::ContextOverflow, "");
assert!(msg.contains("context window"));
let msg = sanitize_for_user(LlmErrorCategory::ModelNotFound, "");
assert!(msg.contains("model"));
// All messages should be under 200 chars
for cat in [
LlmErrorCategory::RateLimit,
LlmErrorCategory::Overloaded,
LlmErrorCategory::Timeout,
LlmErrorCategory::Billing,
LlmErrorCategory::Auth,
LlmErrorCategory::ContextOverflow,
LlmErrorCategory::Format,
LlmErrorCategory::ModelNotFound,
] {
let m = sanitize_for_user(cat, "test");
assert!(
m.len() <= 200,
"Message for {:?} too long: {}",
cat,
m.len()
);
}
}
#[test]
fn test_extract_retry_delay() {
assert_eq!(
extract_retry_delay("Rate limited. Retry after 30 seconds"),
Some(30_000)
);
assert_eq!(extract_retry_delay("retry-after: 5"), Some(5_000));
assert_eq!(
extract_retry_delay("Please try again in 10 seconds"),
Some(10_000)
);
assert_eq!(extract_retry_delay("Retry after 500ms"), Some(500));
}
#[test]
fn test_extract_retry_delay_none() {
assert_eq!(extract_retry_delay("Something went wrong"), None);
assert_eq!(extract_retry_delay(""), None);
assert_eq!(extract_retry_delay("rate limit exceeded"), None);
}
#[test]
fn test_is_transient() {
assert!(is_transient("Connection reset by peer"));
assert!(is_transient("ECONNRESET"));
assert!(is_transient("Request timed out after 30s"));
assert!(is_transient("Service unavailable"));
assert!(is_transient("rate limit exceeded"));
// Non-transient
assert!(!is_transient("invalid api key"));
assert!(!is_transient("model not found"));
assert!(!is_transient("context_length_exceeded"));
}
#[test]
fn test_is_html_error_page() {
assert!(is_html_error_page(
"<!DOCTYPE html><html><body>Error</body></html>"
));
assert!(is_html_error_page("<html lang='en'>502 Bad Gateway</html>"));
assert!(!is_html_error_page(r#"{"error": "rate limit"}"#));
assert!(!is_html_error_page("plain text error message"));
}
#[test]
fn test_cloudflare_detection() {
assert!(is_html_error_page(
"<!DOCTYPE html><html><body>cloudflare 522 connection timed out</body></html>"
));
assert!(is_html_error_page(
"<html><head><meta cf-error-code='1015'></head></html>"
));
}
#[test]
fn test_unknown_error_defaults() {
// An error with no recognizable pattern and no status code
let e = classify_error("??? something unknown ???", None);
// Should default to Format (unknown structured error)
assert_eq!(e.category, LlmErrorCategory::Format);
// Network-sounding message without explicit pattern
let e = classify_error("failed to connect to host", None);
assert_eq!(e.category, LlmErrorCategory::Timeout);
}
#[test]
fn test_gemini_specific_errors() {
// Gemini model not found format
let e = classify_error(
"models/gemini-ultra is not found for API version v1beta",
None,
);
assert_eq!(e.category, LlmErrorCategory::ModelNotFound);
// Gemini overloaded
let e = classify_error("The model is overloaded. Please try again later.", None);
assert_eq!(e.category, LlmErrorCategory::Overloaded);
// Gemini resource exhausted (rate limit)
let e = classify_error("Resource exhausted: request rate limit exceeded", None);
assert_eq!(e.category, LlmErrorCategory::RateLimit);
}
#[test]
fn test_anthropic_specific_errors() {
// Anthropic overloaded_error
let e = classify_error(
r#"{"type":"error","error":{"type":"overloaded_error","message":"Overloaded"}}"#,
Some(529),
);
assert_eq!(e.category, LlmErrorCategory::Overloaded);
// Anthropic rate limit
let e = classify_error(
"rate_limit_error: Number of request tokens has exceeded your per-minute rate limit",
Some(429),
);
assert_eq!(e.category, LlmErrorCategory::RateLimit);
// Anthropic invalid API key
let e = classify_error(
r#"{"type":"error","error":{"type":"authentication_error","message":"invalid x-api-key"}}"#,
Some(401),
);
assert_eq!(e.category, LlmErrorCategory::Auth);
}
}