api_endpoints_test.rs 71.3 KB
Newer Older
1
2
mod common;

3
4
5
6
7
use axum::{
    body::Body,
    extract::Request,
    http::{header::CONTENT_TYPE, StatusCode},
};
8
9
10
use common::mock_worker::{HealthStatus, MockWorker, MockWorkerConfig, WorkerType};
use reqwest::Client;
use serde_json::json;
11
use sglang_router_rs::config::{
12
    CircuitBreakerConfig, ConnectionMode, PolicyConfig, RetryConfig, RouterConfig, RoutingMode,
13
};
14
use sglang_router_rs::core::WorkerManager;
15
use sglang_router_rs::routers::{RouterFactory, RouterTrait};
16
use sglang_router_rs::server::AppContext;
17
18
use std::sync::Arc;
use tower::ServiceExt;
19
20
21
22

/// Test context that manages mock workers
struct TestContext {
    workers: Vec<MockWorker>,
23
    router: Arc<dyn RouterTrait>,
24
25
26
    _client: Client,
    _config: RouterConfig,
    app_context: Arc<AppContext>,
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
}

impl TestContext {
    async fn new(worker_configs: Vec<MockWorkerConfig>) -> Self {
        // Create default router config
        let config = RouterConfig {
            mode: RoutingMode::Regular {
                worker_urls: vec![],
            },
            policy: PolicyConfig::Random,
            host: "127.0.0.1".to_string(),
            port: 3002,
            max_payload_size: 256 * 1024 * 1024,
            request_timeout_secs: 600,
            worker_startup_timeout_secs: 1,
            worker_startup_check_interval_secs: 1,
43
            discovery: None,
44
45
            dp_aware: false,
            api_key: None,
46
47
48
            metrics: None,
            log_dir: None,
            log_level: None,
49
            request_id_headers: None,
50
            max_concurrent_requests: 64,
51
52
53
            queue_size: 0,
            queue_timeout_secs: 60,
            rate_limit_tokens_per_second: None,
54
            cors_allowed_origins: vec![],
55
            retry: RetryConfig::default(),
56
            circuit_breaker: CircuitBreakerConfig::default(),
57
58
            disable_retries: false,
            disable_circuit_breaker: false,
59
            health_check: sglang_router_rs::config::HealthCheckConfig::default(),
60
            enable_igw: false,
61
62
63
            connection_mode: ConnectionMode::Http,
            model_path: None,
            tokenizer_path: None,
64
            history_backend: sglang_router_rs::config::HistoryBackend::Memory,
65
66
67
68
69
        };

        Self::new_with_config(config, worker_configs).await
    }

70
71
72
73
    async fn new_with_config(
        mut config: RouterConfig,
        worker_configs: Vec<MockWorkerConfig>,
    ) -> Self {
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
        let mut workers = Vec::new();
        let mut worker_urls = Vec::new();

        // Start mock workers if any
        for worker_config in worker_configs {
            let mut worker = MockWorker::new(worker_config);
            let url = worker.start().await.unwrap();
            worker_urls.push(url);
            workers.push(worker);
        }

        if !workers.is_empty() {
            tokio::time::sleep(tokio::time::Duration::from_millis(200)).await;
        }

89
90
91
92
93
94
95
96
97
98
        // Update config with worker URLs if not already set
        if let RoutingMode::Regular {
            worker_urls: ref mut urls,
        } = config.mode
        {
            if urls.is_empty() {
                *urls = worker_urls.clone();
            }
        }

99
100
101
102
103
        let client = Client::builder()
            .timeout(std::time::Duration::from_secs(config.request_timeout_secs))
            .build()
            .unwrap();

104
105
        // Create app context
        let app_context = common::create_test_context(config.clone());
106

107
108
        // Initialize workers in the registry before creating router
        if !worker_urls.is_empty() {
109
            WorkerManager::initialize_workers(&config, &app_context.worker_registry, None)
110
111
112
113
                .await
                .expect("Failed to initialize workers");
        }

114
115
        // Create router
        let router = RouterFactory::create_router(&app_context).await.unwrap();
116
        let router = Arc::from(router);
117

118
119
        // Wait for router to discover workers
        if !workers.is_empty() {
120
121
122
            tokio::time::sleep(tokio::time::Duration::from_millis(500)).await;
        }

123
124
125
        Self {
            workers,
            router,
126
127
128
            _client: client,
            _config: config,
            app_context,
129
        }
130
131
    }

132
    async fn create_app(&self) -> axum::Router {
133
        common::test_app::create_test_app_with_context(
134
            Arc::clone(&self.router),
135
            Arc::clone(&self.app_context),
136
137
138
139
140
141
142
143
144
145
146
147
148
149
        )
    }

    async fn shutdown(mut self) {
        for worker in &mut self.workers {
            worker.stop().await;
        }
    }
}

#[cfg(test)]
mod health_tests {
    use super::*;

150
151
152
153
    #[tokio::test]
    async fn test_liveness_endpoint() {
        let ctx = TestContext::new(vec![]).await;
        let app = ctx.create_app().await;
154

155
156
157
158
159
        let req = Request::builder()
            .method("GET")
            .uri("/liveness")
            .body(Body::empty())
            .unwrap();
160

161
162
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
163

164
        ctx.shutdown().await;
165
166
    }

167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
    #[tokio::test]
    async fn test_readiness_with_healthy_workers() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18001,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let req = Request::builder()
            .method("GET")
            .uri("/readiness")
            .body(Body::empty())
            .unwrap();
185

186
187
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
188

189
        ctx.shutdown().await;
190
191
    }

192
193
194
    #[tokio::test]
    async fn test_readiness_with_unhealthy_workers() {
        let ctx = TestContext::new(vec![]).await;
195

196
        let app = ctx.create_app().await;
197

198
199
200
201
202
        let req = Request::builder()
            .method("GET")
            .uri("/readiness")
            .body(Body::empty())
            .unwrap();
203

204
205
206
        let resp = app.oneshot(req).await.unwrap();
        // With no workers, readiness should return SERVICE_UNAVAILABLE
        assert_eq!(resp.status(), StatusCode::SERVICE_UNAVAILABLE);
207

208
        ctx.shutdown().await;
209
210
    }

211
212
213
214
215
216
217
218
219
220
221
222
    #[tokio::test]
    async fn test_health_endpoint_details() {
        let ctx = TestContext::new(vec![
            MockWorkerConfig {
                port: 18003,
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
            },
            MockWorkerConfig {
                port: 18004,
223
224
225
226
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
227
228
229
            },
        ])
        .await;
230

231
        let app = ctx.create_app().await;
232

233
234
235
236
237
        let req = Request::builder()
            .method("GET")
            .uri("/health")
            .body(Body::empty())
            .unwrap();
238

239
240
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
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
        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_health_generate_endpoint() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18005,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let req = Request::builder()
            .method("GET")
            .uri("/health_generate")
            .body(Body::empty())
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert!(body_json.is_object());

        ctx.shutdown().await;
274
275
276
277
278
279
280
    }
}

#[cfg(test)]
mod generation_tests {
    use super::*;

281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
    #[tokio::test]
    async fn test_generate_success() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18101,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "text": "Hello, world!",
            "stream": false
        });
298

299
300
301
302
303
304
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
305

306
307
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
308

309
310
311
312
313
314
315
316
317
318
319
320
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert!(body_json.get("text").is_some());
        assert!(body_json.get("meta_info").is_some());
        let meta_info = &body_json["meta_info"];
        assert!(meta_info.get("finish_reason").is_some());
        assert_eq!(meta_info["finish_reason"]["type"], "stop");

        ctx.shutdown().await;
    }
321

322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
    #[tokio::test]
    async fn test_generate_streaming() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18102,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "text": "Stream test",
            "stream": true
338
339
        });

340
341
342
343
344
345
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
346

347
348
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
349

350
351
352
353
        // For streaming responses, the router might use chunked encoding or other streaming mechanisms
        // The exact content-type can vary based on the router implementation
        // Just verify we got a successful response
        // Note: In a real implementation, we'd check for text/event-stream or appropriate streaming headers
354

355
356
        ctx.shutdown().await;
    }
357

358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
    #[tokio::test]
    async fn test_generate_with_worker_failure() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18103,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 1.0, // Always fail
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "text": "This should fail",
            "stream": false
        });
375

376
377
378
379
380
381
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
382

383
384
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::INTERNAL_SERVER_ERROR);
385

386
387
        ctx.shutdown().await;
    }
388

389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
    #[tokio::test]
    async fn test_v1_chat_completions_success() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18104,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "model": "test-model",
            "messages": [
                {"role": "user", "content": "Hello!"}
            ],
            "stream": false
        });
409

410
411
412
413
414
415
        let req = Request::builder()
            .method("POST")
            .uri("/v1/chat/completions")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
416

417
418
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
419

420
421
422
423
424
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert!(body_json.get("choices").is_some());
425

426
        ctx.shutdown().await;
427
    }
428
}
429

430
431
432
#[cfg(test)]
mod model_info_tests {
    use super::*;
433

434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
    #[tokio::test]
    async fn test_get_server_info() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18201,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let req = Request::builder()
            .method("GET")
            .uri("/get_server_info")
            .body(Body::empty())
            .unwrap();
452

453
454
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
455

456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert!(body_json.is_object());
        // Check for actual sglang server fields
        assert!(body_json.get("version").is_some());
        assert!(body_json.get("model_path").is_some());
        assert!(body_json.get("tokenizer_path").is_some());
        assert!(body_json.get("port").is_some());
        assert!(body_json.get("max_num_batched_tokens").is_some());
        assert!(body_json.get("schedule_policy").is_some());

        ctx.shutdown().await;
    }
471

472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
    #[tokio::test]
    async fn test_get_model_info() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18202,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let req = Request::builder()
            .method("GET")
            .uri("/get_model_info")
            .body(Body::empty())
            .unwrap();
490

491
492
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
493

494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert!(body_json.is_object());
        // Check for actual sglang model info fields
        assert_eq!(
            body_json.get("model_path").and_then(|v| v.as_str()),
            Some("mock-model-path")
        );
        assert_eq!(
            body_json.get("tokenizer_path").and_then(|v| v.as_str()),
            Some("mock-tokenizer-path")
        );
        assert_eq!(
            body_json.get("is_generation").and_then(|v| v.as_bool()),
            Some(true)
        );
        assert!(body_json.get("preferred_sampling_params").is_some());

        ctx.shutdown().await;
515
516
    }

517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
    #[tokio::test]
    async fn test_v1_models() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18203,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let req = Request::builder()
            .method("GET")
            .uri("/v1/models")
            .body(Body::empty())
            .unwrap();
535

536
537
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
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
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert!(body_json.get("object").is_some());
        assert_eq!(
            body_json.get("object").and_then(|v| v.as_str()),
            Some("list")
        );

        let data = body_json.get("data").and_then(|v| v.as_array());
        assert!(data.is_some());

        let models = data.unwrap();
        assert!(!models.is_empty());

        let first_model = &models[0];
        assert_eq!(
            first_model.get("id").and_then(|v| v.as_str()),
            Some("mock-model")
        );
        assert_eq!(
            first_model.get("object").and_then(|v| v.as_str()),
            Some("model")
        );
        assert!(first_model.get("created").is_some());
        assert_eq!(
            first_model.get("owned_by").and_then(|v| v.as_str()),
            Some("organization-owner")
        );

        ctx.shutdown().await;
    }
572

573
574
575
576
    #[tokio::test]
    async fn test_model_info_with_no_workers() {
        let ctx = TestContext::new(vec![]).await;
        let app = ctx.create_app().await;
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
        // Test server info with no workers
        let req = Request::builder()
            .method("GET")
            .uri("/get_server_info")
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        // Router may return various error codes when no workers
        assert!(
            resp.status() == StatusCode::OK
                || resp.status() == StatusCode::SERVICE_UNAVAILABLE
                || resp.status() == StatusCode::NOT_FOUND
                || resp.status() == StatusCode::INTERNAL_SERVER_ERROR,
            "Unexpected status code: {:?}",
            resp.status()
        );

        // Test model info with no workers
        let req = Request::builder()
            .method("GET")
            .uri("/get_model_info")
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        // Router may return various error codes when no workers
        assert!(
            resp.status() == StatusCode::OK
                || resp.status() == StatusCode::SERVICE_UNAVAILABLE
                || resp.status() == StatusCode::NOT_FOUND
                || resp.status() == StatusCode::INTERNAL_SERVER_ERROR,
            "Unexpected status code: {:?}",
            resp.status()
        );

        // Test v1/models with no workers
        let req = Request::builder()
            .method("GET")
            .uri("/v1/models")
            .body(Body::empty())
            .unwrap();
        let resp = app.oneshot(req).await.unwrap();
        // Router may return various error codes when no workers
        assert!(
            resp.status() == StatusCode::OK
                || resp.status() == StatusCode::SERVICE_UNAVAILABLE
                || resp.status() == StatusCode::NOT_FOUND
                || resp.status() == StatusCode::INTERNAL_SERVER_ERROR,
            "Unexpected status code: {:?}",
            resp.status()
        );

        ctx.shutdown().await;
630
631
    }

632
633
634
635
636
637
638
639
640
641
642
643
    #[tokio::test]
    async fn test_model_info_with_multiple_workers() {
        let ctx = TestContext::new(vec![
            MockWorkerConfig {
                port: 18204,
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
            },
            MockWorkerConfig {
                port: 18205,
644
645
646
647
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
648
649
650
            },
        ])
        .await;
651

652
        let app = ctx.create_app().await;
653

654
655
656
657
        // Test that model info is consistent across workers
        for _ in 0..5 {
            let req = Request::builder()
                .method("GET")
658
                .uri("/get_model_info")
659
660
                .body(Body::empty())
                .unwrap();
661

662
            let resp = app.clone().oneshot(req).await.unwrap();
663
664
            assert_eq!(resp.status(), StatusCode::OK);

665
666
667
668
            let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
                .await
                .unwrap();
            let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
669
            assert_eq!(
670
                body_json.get("model_path").and_then(|v| v.as_str()),
671
672
                Some("mock-model-path")
            );
673
        }
674

675
        ctx.shutdown().await;
676
677
    }

678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
    #[tokio::test]
    async fn test_model_info_with_unhealthy_worker() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18206,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 1.0, // Always fail
        }])
        .await;

        let app = ctx.create_app().await;

        let req = Request::builder()
            .method("GET")
            .uri("/get_model_info")
            .body(Body::empty())
            .unwrap();
696

697
698
699
700
701
702
703
704
        let resp = app.oneshot(req).await.unwrap();
        // Worker with fail_rate: 1.0 should always return an error status
        assert!(
            resp.status() == StatusCode::INTERNAL_SERVER_ERROR
                || resp.status() == StatusCode::SERVICE_UNAVAILABLE,
            "Expected error status for always-failing worker, got: {:?}",
            resp.status()
        );
705

706
707
708
        ctx.shutdown().await;
    }
}
709

710
711
712
#[cfg(test)]
mod worker_management_tests {
    use super::*;
713

714
715
716
717
718
719
720
721
722
723
724
725
    #[tokio::test]
    async fn test_add_new_worker() {
        let ctx = TestContext::new(vec![]).await;
        let app = ctx.create_app().await;

        // Start a mock worker
        let mut worker = MockWorker::new(MockWorkerConfig {
            port: 18301,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
726
        });
727
        let url = worker.start().await.unwrap();
728

729
730
731
        // Add the worker
        let req = Request::builder()
            .method("POST")
732
            .uri(format!("/add_worker?url={}", url))
733
734
            .body(Body::empty())
            .unwrap();
735

736
737
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
738

739
740
741
742
743
744
        // List workers to verify
        let req = Request::builder()
            .method("GET")
            .uri("/list_workers")
            .body(Body::empty())
            .unwrap();
745

746
747
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
748

749
750
751
752
753
754
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        let workers = body_json["urls"].as_array().unwrap();
        assert!(workers.iter().any(|w| w.as_str().unwrap() == url));
755

756
757
        worker.stop().await;
        ctx.shutdown().await;
758
759
    }

760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
    #[tokio::test]
    async fn test_remove_existing_worker() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18302,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // Get the worker URL
        let req = Request::builder()
            .method("GET")
            .uri("/list_workers")
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        let workers = body_json["urls"].as_array().unwrap();
        let worker_url = workers[0].as_str().unwrap();

        // Remove the worker
        let req = Request::builder()
            .method("POST")
790
            .uri(format!("/remove_worker?url={}", worker_url))
791
792
            .body(Body::empty())
            .unwrap();
793

794
795
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
796

797
798
799
800
801
802
803
804
805
806
807
808
809
        // Verify it's removed
        let req = Request::builder()
            .method("GET")
            .uri("/list_workers")
            .body(Body::empty())
            .unwrap();
        let resp = app.oneshot(req).await.unwrap();
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        let workers = body_json["urls"].as_array().unwrap();
        assert!(workers.is_empty());
810

811
        ctx.shutdown().await;
812
813
    }

814
815
816
817
    #[tokio::test]
    async fn test_add_worker_invalid_url() {
        let ctx = TestContext::new(vec![]).await;
        let app = ctx.create_app().await;
818

819
820
821
822
823
824
        // Invalid URL format
        let req = Request::builder()
            .method("POST")
            .uri("/add_worker?url=not-a-valid-url")
            .body(Body::empty())
            .unwrap();
825

826
827
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);
828

829
830
831
832
833
834
        // Missing URL parameter
        let req = Request::builder()
            .method("POST")
            .uri("/add_worker")
            .body(Body::empty())
            .unwrap();
835

836
837
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);
838

839
840
841
842
843
844
        // Empty URL
        let req = Request::builder()
            .method("POST")
            .uri("/add_worker?url=")
            .body(Body::empty())
            .unwrap();
845

846
847
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);
848

849
        ctx.shutdown().await;
850
851
    }

852
853
854
855
856
857
858
859
860
861
862
    #[tokio::test]
    async fn test_add_duplicate_worker() {
        // Start a mock worker
        let mut worker = MockWorker::new(MockWorkerConfig {
            port: 18303,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        });
        let url = worker.start().await.unwrap();
863

864
865
        let ctx = TestContext::new(vec![]).await;
        let app = ctx.create_app().await;
866

867
868
869
        // Add worker first time
        let req = Request::builder()
            .method("POST")
870
            .uri(format!("/add_worker?url={}", url))
871
872
873
874
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
875

876
        tokio::time::sleep(tokio::time::Duration::from_millis(500)).await;
877

878
879
880
        // Try to add same worker again
        let req = Request::builder()
            .method("POST")
881
            .uri(format!("/add_worker?url={}", url))
882
883
884
885
886
            .body(Body::empty())
            .unwrap();
        let resp = app.oneshot(req).await.unwrap();
        // Should return error for duplicate
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);
887

888
889
890
        worker.stop().await;
        ctx.shutdown().await;
    }
891

892
893
894
895
896
897
898
899
900
901
902
    #[tokio::test]
    async fn test_add_unhealthy_worker() {
        // Start unhealthy worker
        let mut worker = MockWorker::new(MockWorkerConfig {
            port: 18304,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Unhealthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        });
        let url = worker.start().await.unwrap();
903

904
905
        let ctx = TestContext::new(vec![]).await;
        let app = ctx.create_app().await;
906

907
908
909
        // Try to add unhealthy worker
        let req = Request::builder()
            .method("POST")
910
            .uri(format!("/add_worker?url={}", url))
911
912
913
            .body(Body::empty())
            .unwrap();
        let resp = app.oneshot(req).await.unwrap();
914

915
916
917
918
919
        // Router should reject unhealthy workers
        assert!(
            resp.status() == StatusCode::BAD_REQUEST
                || resp.status() == StatusCode::SERVICE_UNAVAILABLE
        );
920

921
922
        worker.stop().await;
        ctx.shutdown().await;
923
    }
924
925
926
927
928
}

#[cfg(test)]
mod router_policy_tests {
    use super::*;
929

930
931
932
933
934
    #[tokio::test]
    async fn test_random_policy() {
        let ctx = TestContext::new(vec![
            MockWorkerConfig {
                port: 18801,
935
936
937
938
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
939
940
941
            },
            MockWorkerConfig {
                port: 18802,
942
                worker_type: WorkerType::Regular,
943
                health_status: HealthStatus::Healthy,
944
945
                response_delay_ms: 0,
                fail_rate: 0.0,
946
947
948
949
950
951
952
953
954
955
956
            },
        ])
        .await;

        // Send multiple requests and verify they succeed
        let app = ctx.create_app().await;

        for i in 0..10 {
            let payload = json!({
                "text": format!("Request {}", i),
                "stream": false
957
958
            });

959
960
961
962
963
964
            let req = Request::builder()
                .method("POST")
                .uri("/generate")
                .header(CONTENT_TYPE, "application/json")
                .body(Body::from(serde_json::to_string(&payload).unwrap()))
                .unwrap();
965

966
967
968
            let resp = app.clone().oneshot(req).await.unwrap();
            assert_eq!(resp.status(), StatusCode::OK);
        }
969

970
971
        ctx.shutdown().await;
    }
972

973
974
975
976
977
978
979
980
981
982
983
984
985
986
    #[tokio::test]
    async fn test_worker_selection() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18203,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let _payload = json!({
            "text": "Test selection",
            "stream": false
987
        });
988
989

        // Check that router has the worker
990
991
        // TODO: Update test after worker management refactoring
        // For now, skip this check
992
993

        ctx.shutdown().await;
994
995
996
    }
}

997
998
999
#[cfg(test)]
mod responses_endpoint_tests {
    use super::*;
1000
    use reqwest::Client as HttpClient;
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080

    #[tokio::test]
    async fn test_v1_responses_non_streaming() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18950,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "input": "Hello Responses API",
            "model": "mock-model",
            "stream": false
        });

        let req = Request::builder()
            .method("POST")
            .uri("/v1/responses")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert_eq!(body_json["object"], "response");
        assert_eq!(body_json["status"], "completed");

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_v1_responses_streaming() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18951,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "input": "Hello Responses API",
            "model": "mock-model",
            "stream": true
        });

        let req = Request::builder()
            .method("POST")
            .uri("/v1/responses")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        // Check that content-type indicates SSE
        let headers = resp.headers().clone();
        let ct = headers
            .get("content-type")
            .and_then(|v| v.to_str().ok())
            .unwrap_or("");
        assert!(ct.contains("text/event-stream"));

        // We don't fully consume the stream in this test harness.
        ctx.shutdown().await;
    }
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095

    #[tokio::test]
    async fn test_v1_responses_get() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18952,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // First create a response to obtain an id
1096
        let resp_id = "test-get-resp-id-123";
1097
1098
1099
        let payload = json!({
            "input": "Hello Responses API",
            "model": "mock-model",
1100
1101
1102
1103
            "stream": false,
            "store": true,
            "background": true,
            "request_id": resp_id
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
        });
        let req = Request::builder()
            .method("POST")
            .uri("/v1/responses")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        // Retrieve the response
        let req = Request::builder()
            .method("GET")
            .uri(format!("/v1/responses/{}", resp_id))
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let get_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert_eq!(get_json["object"], "response");

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_v1_responses_cancel() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18953,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // First create a response to obtain an id
1145
        let resp_id = "test-cancel-resp-id-456";
1146
1147
1148
        let payload = json!({
            "input": "Hello Responses API",
            "model": "mock-model",
1149
1150
1151
1152
            "stream": false,
            "store": true,
            "background": true,
            "request_id": resp_id
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
        });
        let req = Request::builder()
            .method("POST")
            .uri("/v1/responses")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        // Cancel the response
        let req = Request::builder()
            .method("POST")
            .uri(format!("/v1/responses/{}/cancel", resp_id))
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let cancel_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
        assert_eq!(cancel_json["status"], "cancelled");

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_v1_responses_delete_and_list_not_implemented() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18954,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // Use an arbitrary id for delete/list
        let resp_id = "resp-test-123";

        let req = Request::builder()
            .method("DELETE")
            .uri(format!("/v1/responses/{}", resp_id))
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::NOT_IMPLEMENTED);

        let req = Request::builder()
            .method("GET")
            .uri(format!("/v1/responses/{}/input", resp_id))
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::NOT_IMPLEMENTED);

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_v1_responses_get_multi_worker_fanout() {
        // Start two mock workers
        let ctx = TestContext::new(vec![
            MockWorkerConfig {
                port: 18960,
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
            },
            MockWorkerConfig {
                port: 18961,
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
            },
        ])
        .await;

        let app = ctx.create_app().await;

        // Create a background response with a known id
        let rid = format!("resp_{}", 18960); // arbitrary unique id
        let payload = json!({
            "input": "Hello Responses API",
            "model": "mock-model",
            "background": true,
            "store": true,
            "request_id": rid,
        });

        let req = Request::builder()
            .method("POST")
            .uri("/v1/responses")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        // Using the router, GET should succeed by fanning out across workers
        let req = Request::builder()
            .method("GET")
            .uri(format!("/v1/responses/{}", rid))
            .body(Body::empty())
            .unwrap();
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        // Validate only one worker holds the metadata: direct calls
        let client = HttpClient::new();
        let mut ok_count = 0usize;
1269
1270
1271
1272
1273
1274
        // Get the actual worker URLs from the context
        let worker_urls: Vec<String> = vec![
            "http://127.0.0.1:18960".to_string(),
            "http://127.0.0.1:18961".to_string(),
        ];
        for url in worker_urls {
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
            let get_url = format!("{}/v1/responses/{}", url, rid);
            let res = client.get(get_url).send().await.unwrap();
            if res.status() == StatusCode::OK {
                ok_count += 1;
            }
        }
        assert_eq!(ok_count, 1, "exactly one worker should store the response");

        ctx.shutdown().await;
    }
1285
1286
}

1287
1288
1289
1290
#[cfg(test)]
mod error_tests {
    use super::*;

1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
    #[tokio::test]
    async fn test_404_not_found() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18401,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // Test unknown endpoint
        let req = Request::builder()
            .method("GET")
            .uri("/unknown_endpoint")
            .body(Body::empty())
            .unwrap();
1310

1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::NOT_FOUND);

        // Test POST to unknown endpoint
        let req = Request::builder()
            .method("POST")
            .uri("/api/v2/generate")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(
                serde_json::to_string(&json!({"text": "test"})).unwrap(),
            ))
            .unwrap();
1323

1324
1325
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::NOT_FOUND);
1326

1327
1328
        ctx.shutdown().await;
    }
1329

1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
    #[tokio::test]
    async fn test_method_not_allowed() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18402,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // GET request to POST-only endpoint
        let req = Request::builder()
            .method("GET")
            .uri("/generate")
            .body(Body::empty())
            .unwrap();
1349

1350
1351
1352
        let resp = app.clone().oneshot(req).await.unwrap();
        // Note: Axum returns 405 for wrong methods on matched routes
        assert_eq!(resp.status(), StatusCode::METHOD_NOT_ALLOWED);
1353

1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
        // POST request to GET-only endpoint
        let req = Request::builder()
            .method("POST")
            .uri("/health")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from("{}"))
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::METHOD_NOT_ALLOWED);

        ctx.shutdown().await;
1366
1367
    }

1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
    #[tokio::test]
    async fn test_payload_too_large() {
        // Create context with small payload limit
        let config = RouterConfig {
            mode: RoutingMode::Regular {
                worker_urls: vec![],
            },
            policy: PolicyConfig::Random,
            host: "127.0.0.1".to_string(),
            port: 3010,
            max_payload_size: 1024, // 1KB limit
            request_timeout_secs: 600,
            worker_startup_timeout_secs: 1,
            worker_startup_check_interval_secs: 1,
            dp_aware: false,
            api_key: None,
            discovery: None,
            metrics: None,
            log_dir: None,
            log_level: None,
            request_id_headers: None,
            max_concurrent_requests: 64,
1390
1391
1392
            queue_size: 0,
            queue_timeout_secs: 60,
            rate_limit_tokens_per_second: None,
1393
            cors_allowed_origins: vec![],
1394
            retry: RetryConfig::default(),
1395
            circuit_breaker: CircuitBreakerConfig::default(),
1396
1397
            disable_retries: false,
            disable_circuit_breaker: false,
1398
            health_check: sglang_router_rs::config::HealthCheckConfig::default(),
1399
            enable_igw: false,
1400
1401
1402
            connection_mode: ConnectionMode::Http,
            model_path: None,
            tokenizer_path: None,
1403
            history_backend: sglang_router_rs::config::HistoryBackend::Memory,
1404
1405
1406
1407
1408
1409
        };

        let ctx = TestContext::new_with_config(
            config,
            vec![MockWorkerConfig {
                port: 18403,
1410
1411
1412
1413
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
1414
1415
1416
            }],
        )
        .await;
1417

1418
1419
1420
        // Note: The server would have payload size middleware configured
        // but we cannot test it directly through the test app
        // This test is kept for documentation purposes
1421

1422
1423
        ctx.shutdown().await;
    }
1424

1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
    #[tokio::test]
    async fn test_invalid_json_payload() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18404,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // Send invalid JSON
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from("{invalid json}"))
            .unwrap();
1445

1446
1447
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);
1448

1449
1450
1451
1452
1453
1454
1455
        // Send empty body
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::empty())
            .unwrap();
1456

1457
1458
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);
1459

1460
1461
        ctx.shutdown().await;
    }
1462

1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
    #[tokio::test]
    async fn test_missing_required_fields() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18405,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // Missing messages in chat completion
        let payload = json!({
            "model": "test-model"
            // missing "messages"
1480
        });
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493

        let req = Request::builder()
            .method("POST")
            .uri("/v1/chat/completions")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        // Axum validates JSON schema - returns 422 for validation errors
        assert_eq!(resp.status(), StatusCode::UNPROCESSABLE_ENTITY);

        ctx.shutdown().await;
1494
1495
    }

1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
    #[tokio::test]
    async fn test_invalid_model() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18406,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "model": "invalid-model-name-that-does-not-exist",
            "messages": [{"role": "user", "content": "Hello"}],
            "stream": false
        });
1514

1515
1516
1517
1518
1519
1520
        let req = Request::builder()
            .method("POST")
            .uri("/v1/chat/completions")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
1521

1522
1523
1524
        let resp = app.oneshot(req).await.unwrap();
        // Mock worker accepts any model, but real implementation might return 400
        assert!(resp.status().is_success() || resp.status() == StatusCode::BAD_REQUEST);
1525

1526
1527
1528
        ctx.shutdown().await;
    }
}
1529

1530
1531
1532
#[cfg(test)]
mod cache_tests {
    use super::*;
1533

1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
    #[tokio::test]
    async fn test_flush_cache() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18501,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let req = Request::builder()
            .method("POST")
            .uri("/flush_cache")
            .body(Body::empty())
            .unwrap();
1552

1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        // The response might be empty or contain a message
        let body_bytes = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        if !body_bytes.is_empty() {
            if let Ok(body) = serde_json::from_slice::<serde_json::Value>(&body_bytes) {
                // Check that we got a successful response with expected fields
                assert!(body.is_object());
                assert!(body.get("message").is_some() || body.get("status").is_some());
            }
        }

        ctx.shutdown().await;
1569
1570
    }

1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
    #[tokio::test]
    async fn test_get_loads() {
        let ctx = TestContext::new(vec![
            MockWorkerConfig {
                port: 18502,
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
            },
            MockWorkerConfig {
                port: 18503,
1583
1584
1585
1586
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
1587
1588
1589
            },
        ])
        .await;
1590

1591
        let app = ctx.create_app().await;
1592

1593
1594
1595
1596
1597
        let req = Request::builder()
            .method("GET")
            .uri("/get_loads")
            .body(Body::empty())
            .unwrap();
1598

1599
1600
        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
1601

1602
1603
1604
1605
        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();
1606

1607
1608
1609
1610
        // Verify the response contains load information
        assert!(body_json.is_object());
        // The exact structure depends on the implementation
        // but should contain worker load information
1611

1612
1613
        ctx.shutdown().await;
    }
1614

1615
1616
1617
    #[tokio::test]
    async fn test_flush_cache_no_workers() {
        let ctx = TestContext::new(vec![]).await;
1618

1619
        let app = ctx.create_app().await;
1620

1621
1622
1623
1624
1625
        let req = Request::builder()
            .method("POST")
            .uri("/flush_cache")
            .body(Body::empty())
            .unwrap();
1626

1627
1628
1629
1630
1631
        let resp = app.oneshot(req).await.unwrap();
        // Should either succeed (no-op) or return service unavailable
        assert!(
            resp.status() == StatusCode::OK || resp.status() == StatusCode::SERVICE_UNAVAILABLE
        );
1632

1633
        ctx.shutdown().await;
1634
1635
1636
1637
    }
}

#[cfg(test)]
1638
mod load_balancing_tests {
1639
1640
    use super::*;

1641
1642
1643
1644
1645
1646
    #[tokio::test]
    async fn test_request_distribution() {
        // Create multiple workers
        let ctx = TestContext::new(vec![
            MockWorkerConfig {
                port: 18601,
1647
1648
1649
1650
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
            },
            MockWorkerConfig {
                port: 18602,
                worker_type: WorkerType::Regular,
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
            },
        ])
        .await;
1661

1662
        let app = ctx.create_app().await;
1663

1664
1665
1666
1667
1668
1669
1670
        // Send multiple requests and track distribution
        let mut request_count = 0;
        for i in 0..10 {
            let payload = json!({
                "text": format!("Request {}", i),
                "stream": false
            });
1671

1672
1673
1674
1675
1676
1677
            let req = Request::builder()
                .method("POST")
                .uri("/generate")
                .header(CONTENT_TYPE, "application/json")
                .body(Body::from(serde_json::to_string(&payload).unwrap()))
                .unwrap();
1678

1679
1680
1681
            let resp = app.clone().oneshot(req).await.unwrap();
            if resp.status() == StatusCode::OK {
                request_count += 1;
1682
            }
1683
        }
1684

1685
1686
        // With random policy, all requests should succeed
        assert_eq!(request_count, 10);
1687

1688
1689
1690
        ctx.shutdown().await;
    }
}
1691

1692
1693
1694
#[cfg(test)]
mod pd_mode_tests {
    use super::*;
1695

1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
    #[tokio::test]
    async fn test_pd_mode_routing() {
        // Create PD mode configuration with prefill and decode workers
        let mut prefill_worker = MockWorker::new(MockWorkerConfig {
            port: 18701,
            worker_type: WorkerType::Prefill,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        });
1706

1707
1708
1709
1710
1711
1712
        let mut decode_worker = MockWorker::new(MockWorkerConfig {
            port: 18702,
            worker_type: WorkerType::Decode,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
1713
1714
        });

1715
1716
        let prefill_url = prefill_worker.start().await.unwrap();
        let decode_url = decode_worker.start().await.unwrap();
1717

1718
1719
1720
1721
1722
        tokio::time::sleep(tokio::time::Duration::from_millis(200)).await;

        // Extract port from prefill URL
        let prefill_port = prefill_url
            .split(':')
1723
            .next_back()
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
            .and_then(|p| p.trim_end_matches('/').parse::<u16>().ok())
            .unwrap_or(9000);

        let config = RouterConfig {
            mode: RoutingMode::PrefillDecode {
                prefill_urls: vec![(prefill_url, Some(prefill_port))],
                decode_urls: vec![decode_url],
                prefill_policy: None,
                decode_policy: None,
            },
            policy: PolicyConfig::Random,
            host: "127.0.0.1".to_string(),
            port: 3011,
            max_payload_size: 256 * 1024 * 1024,
            request_timeout_secs: 600,
            worker_startup_timeout_secs: 1,
            worker_startup_check_interval_secs: 1,
            discovery: None,
            metrics: None,
            log_dir: None,
            dp_aware: false,
            api_key: None,
            log_level: None,
            request_id_headers: None,
            max_concurrent_requests: 64,
1749
1750
1751
            queue_size: 0,
            queue_timeout_secs: 60,
            rate_limit_tokens_per_second: None,
1752
            cors_allowed_origins: vec![],
1753
            retry: RetryConfig::default(),
1754
            circuit_breaker: CircuitBreakerConfig::default(),
1755
1756
            disable_retries: false,
            disable_circuit_breaker: false,
1757
            health_check: sglang_router_rs::config::HealthCheckConfig::default(),
1758
            enable_igw: false,
1759
1760
1761
            connection_mode: ConnectionMode::Http,
            model_path: None,
            tokenizer_path: None,
1762
            history_backend: sglang_router_rs::config::HistoryBackend::Memory,
1763
1764
        };

1765
1766
1767
        // Create app context
        let app_context = common::create_test_context(config);

1768
        // Create router - this might fail due to health check issues
1769
        let router_result = RouterFactory::create_router(&app_context).await;
1770
1771
1772
1773
1774
1775
1776

        // Clean up workers
        prefill_worker.stop().await;
        decode_worker.stop().await;

        // For now, just verify the configuration was attempted
        assert!(router_result.is_err() || router_result.is_ok());
1777
1778
1779
1780
    }
}

#[cfg(test)]
1781
mod request_id_tests {
1782
1783
    use super::*;

1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
    #[tokio::test]
    async fn test_request_id_generation() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18901,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // Test 1: Request without any request ID header should generate one
        let payload = json!({
            "text": "Test request",
            "stream": false
        });
1802

1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        // Check that response has x-request-id header
        let request_id = resp.headers().get("x-request-id");
        assert!(
            request_id.is_some(),
            "Response should have x-request-id header"
        );

        let id_value = request_id.unwrap().to_str().unwrap();
        assert!(
            id_value.starts_with("gnt-"),
            "Generate endpoint should have gnt- prefix"
        );
        assert!(
            id_value.len() > 4,
            "Request ID should have content after prefix"
        );

        // Test 2: Request with custom x-request-id should preserve it
        let custom_id = "custom-request-id-123";
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .header("x-request-id", custom_id)
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);
1842

1843
1844
1845
1846
1847
1848
1849
1850
        let response_id = resp.headers().get("x-request-id");
        assert!(response_id.is_some());
        assert_eq!(response_id.unwrap(), custom_id);

        // Test 3: Different endpoints should have different prefixes
        let chat_payload = json!({
            "messages": [{"role": "user", "content": "Hello"}],
            "model": "test-model"
1851
        });
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887

        let req = Request::builder()
            .method("POST")
            .uri("/v1/chat/completions")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&chat_payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let request_id = resp.headers().get("x-request-id");
        assert!(request_id.is_some());
        assert!(request_id
            .unwrap()
            .to_str()
            .unwrap()
            .starts_with("chatcmpl-"));

        // Test 4: Alternative request ID headers should be recognized
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .header("x-correlation-id", "correlation-123")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let response_id = resp.headers().get("x-request-id");
        assert!(response_id.is_some());
        assert_eq!(response_id.unwrap(), "correlation-123");

        ctx.shutdown().await;
1888
1889
    }

1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
    #[tokio::test]
    async fn test_request_id_with_custom_headers() {
        // Create config with custom request ID headers
        let config = RouterConfig {
            mode: RoutingMode::Regular {
                worker_urls: vec![],
            },
            policy: PolicyConfig::Random,
            host: "127.0.0.1".to_string(),
            port: 3002,
            max_payload_size: 256 * 1024 * 1024,
            request_timeout_secs: 600,
            worker_startup_timeout_secs: 1,
            worker_startup_check_interval_secs: 1,
            discovery: None,
            metrics: None,
            dp_aware: false,
            api_key: None,
            log_dir: None,
            log_level: None,
            request_id_headers: Some(vec!["custom-id".to_string(), "trace-id".to_string()]),
            max_concurrent_requests: 64,
1912
1913
1914
            queue_size: 0,
            queue_timeout_secs: 60,
            rate_limit_tokens_per_second: None,
1915
            cors_allowed_origins: vec![],
1916
            retry: RetryConfig::default(),
1917
            circuit_breaker: CircuitBreakerConfig::default(),
1918
1919
            disable_retries: false,
            disable_circuit_breaker: false,
1920
            health_check: sglang_router_rs::config::HealthCheckConfig::default(),
1921
            enable_igw: false,
1922
1923
1924
            connection_mode: ConnectionMode::Http,
            model_path: None,
            tokenizer_path: None,
1925
            history_backend: sglang_router_rs::config::HistoryBackend::Memory,
1926
        };
1927

1928
1929
1930
1931
1932
        let ctx = TestContext::new_with_config(
            config,
            vec![MockWorkerConfig {
                port: 18902,
                worker_type: WorkerType::Regular,
1933
1934
1935
                health_status: HealthStatus::Healthy,
                response_delay_ms: 0,
                fail_rate: 0.0,
1936
1937
1938
            }],
        )
        .await;
1939

1940
        let app = ctx.create_app().await;
1941

1942
1943
1944
1945
        let payload = json!({
            "text": "Test request",
            "stream": false
        });
1946

1947
1948
1949
1950
1951
1952
1953
1954
        // Test custom header is recognized
        let req = Request::builder()
            .method("POST")
            .uri("/generate")
            .header(CONTENT_TYPE, "application/json")
            .header("custom-id", "my-custom-id")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();
1955

1956
1957
1958
1959
1960
1961
1962
1963
        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let response_id = resp.headers().get("x-request-id");
        assert!(response_id.is_some());
        assert_eq!(response_id.unwrap(), "my-custom-id");

        ctx.shutdown().await;
1964
1965
    }
}
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294

#[cfg(test)]
mod rerank_tests {
    use super::*;
    // Note: RerankRequest and RerankResult are available for future use

    #[tokio::test]
    async fn test_rerank_success() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18105,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "query": "machine learning algorithms",
            "documents": [
                "Introduction to machine learning concepts",
                "Deep learning neural networks tutorial"
            ],
            "model": "test-rerank-model",
            "top_k": 2,
            "return_documents": true,
            "rid": "test-request-123"
        });

        let req = Request::builder()
            .method("POST")
            .uri("/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();

        // Verify response structure
        assert!(body_json.get("results").is_some());
        assert!(body_json.get("model").is_some());
        assert_eq!(body_json["model"], "test-rerank-model");

        let results = body_json["results"].as_array().unwrap();
        assert_eq!(results.len(), 2);

        // Verify results are sorted by score (highest first)
        assert!(results[0]["score"].as_f64().unwrap() >= results[1]["score"].as_f64().unwrap());

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_rerank_with_top_k() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18106,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "query": "test query",
            "documents": [
                "Document 1",
                "Document 2",
                "Document 3"
            ],
            "model": "test-model",
            "top_k": 1,
            "return_documents": true
        });

        let req = Request::builder()
            .method("POST")
            .uri("/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();

        // Should only return top_k results
        let results = body_json["results"].as_array().unwrap();
        assert_eq!(results.len(), 1);

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_rerank_without_documents() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18107,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "query": "test query",
            "documents": ["Document 1", "Document 2"],
            "model": "test-model",
            "return_documents": false
        });

        let req = Request::builder()
            .method("POST")
            .uri("/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();

        // Documents should be null when return_documents is false
        let results = body_json["results"].as_array().unwrap();
        for result in results {
            assert!(result.get("document").is_none());
        }

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_rerank_worker_failure() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18108,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 1.0, // Always fail
        }])
        .await;

        let app = ctx.create_app().await;

        let payload = json!({
            "query": "test query",
            "documents": ["Document 1"],
            "model": "test-model"
        });

        let req = Request::builder()
            .method("POST")
            .uri("/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        // Should return the worker's error response
        assert_eq!(resp.status(), StatusCode::INTERNAL_SERVER_ERROR);

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_v1_rerank_compatibility() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18110,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // Test V1 API format (simplified input)
        let payload = json!({
            "query": "machine learning algorithms",
            "documents": [
                "Introduction to machine learning concepts",
                "Deep learning neural networks tutorial",
                "Statistical learning theory basics"
            ]
        });

        let req = Request::builder()
            .method("POST")
            .uri("/v1/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::OK);

        let body = axum::body::to_bytes(resp.into_body(), usize::MAX)
            .await
            .unwrap();
        let body_json: serde_json::Value = serde_json::from_slice(&body).unwrap();

        // Verify response structure
        assert!(body_json.get("results").is_some());
        assert!(body_json.get("model").is_some());

        // V1 API should use default model name
        assert_eq!(body_json["model"], "default");

        let results = body_json["results"].as_array().unwrap();
        assert_eq!(results.len(), 3); // All documents should be returned

        // Verify results are sorted by score (highest first)
        assert!(results[0]["score"].as_f64().unwrap() >= results[1]["score"].as_f64().unwrap());
        assert!(results[1]["score"].as_f64().unwrap() >= results[2]["score"].as_f64().unwrap());

        // V1 API should return documents by default
        for result in results {
            assert!(result.get("document").is_some());
        }

        ctx.shutdown().await;
    }

    #[tokio::test]
    async fn test_rerank_invalid_request() {
        let ctx = TestContext::new(vec![MockWorkerConfig {
            port: 18111,
            worker_type: WorkerType::Regular,
            health_status: HealthStatus::Healthy,
            response_delay_ms: 0,
            fail_rate: 0.0,
        }])
        .await;

        let app = ctx.create_app().await;

        // Test empty query string (validation should fail)
        let payload = json!({
            "query": "",
            "documents": ["Document 1", "Document 2"],
            "model": "test-model"
        });

        let req = Request::builder()
            .method("POST")
            .uri("/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);

        // Test query with only whitespace (validation should fail)
        let payload = json!({
            "query": "   ",
            "documents": ["Document 1", "Document 2"],
            "model": "test-model"
        });

        let req = Request::builder()
            .method("POST")
            .uri("/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);

        // Test empty documents list (validation should fail)
        let payload = json!({
            "query": "test query",
            "documents": [],
            "model": "test-model"
        });

        let req = Request::builder()
            .method("POST")
            .uri("/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.clone().oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);

        // Test invalid top_k (validation should fail)
        let payload = json!({
            "query": "test query",
            "documents": ["Document 1", "Document 2"],
            "model": "test-model",
            "top_k": 0
        });

        let req = Request::builder()
            .method("POST")
            .uri("/rerank")
            .header(CONTENT_TYPE, "application/json")
            .body(Body::from(serde_json::to_string(&payload).unwrap()))
            .unwrap();

        let resp = app.oneshot(req).await.unwrap();
        assert_eq!(resp.status(), StatusCode::BAD_REQUEST);

        ctx.shutdown().await;
    }
}