qql-edge 0.2.0

Embedded QQL executor backed by qdrant-edge and optional local FastEmbed inference
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
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
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
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
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
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
1145
1146
1147
1148
1149
1150
1151
1152
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
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
use std::collections::HashMap;
use std::path::PathBuf;
use std::sync::OnceLock;
use std::sync::{Arc, Mutex};

use async_trait::async_trait;
use fastembed::{
    Bgem3Embedding, Bgem3InitOptions, Bgem3Model, EmbeddingModel, ImageEmbedding,
    ImageEmbeddingModel, ImageInitOptions, InitOptionsWithLength, RerankInitOptions, RerankerModel,
    SparseInitOptions, SparseModel, SparseTextEmbedding, TextEmbedding, TextRerank,
};

use qql_core::error::QqlError;
use qql_embed::{Embedder, JointEmbeddingOutput, SparseVector};

fn err(msg: impl Into<std::borrow::Cow<'static, str>>) -> QqlError {
    QqlError::execution("QQL-EDGE-EMBED", msg, None)
}

/// Public description of a local ONNX embedding model.
#[derive(Debug, Clone)]
pub struct EmbeddingModelInfo {
    /// Stable enum-style name, e.g. `"BGESmallENV15"`.
    pub name: String,
    /// HuggingFace / Xenova model code, e.g. `"Xenova/bge-small-en-v1.5"`.
    pub model_code: String,
    /// Output dimension of dense vectors.
    pub dim: usize,
    /// Short human description from fastembed.
    pub description: String,
    /// True when this entry is a multivector / ColBERT-capable offline model.
    pub multi: bool,
    /// True when this entry is an image / CLIP vision model.
    pub image: bool,
}

/// Options for constructing a [`FastEmbedder`].
#[derive(Debug, Clone, Default)]
pub struct FastEmbedderOptions {
    /// Dense model name. Accepts enum names (`BGESmallENV15`), HF codes
    /// (`Xenova/bge-small-en-v1.5`), or short aliases (`bge-small-en-v1.5`).
    /// For CLIP text use `ClipVitB32` / `Qdrant/clip-ViT-B-32-text`.
    /// `None` → default `BGESmallENV15` (384-d).
    pub model: Option<String>,
    /// Offline sparse model (SPLADE or BGE-M3 via `SparseTextEmbedding`).
    /// Accepts `splade`, `SPLADEPPV1`, `Qdrant/Splade_PP_en_v1`, `bge-m3`,
    /// `BGEM3`, `BAAI/bge-m3`. When set, `embed_sparse` uses real ONNX
    /// inference. `None` → local BM25 hashing for sparse requests.
    pub sparse_model: Option<String>,
    /// Offline multivector model. Accepts `bge-m3`, `BGEM3Q`,
    /// `gpahal/bge-m3-onnx-int8`. When set, `embed_multi` runs via BGE-M3 ColBERT.
    pub multi_model: Option<String>,
    /// Offline image / CLIP vision model. Accepts `ClipVitB32`,
    /// `Qdrant/clip-ViT-B-32-vision`, `clip-vision`. Pairs with dense CLIP text.
    pub image_model: Option<String>,
    /// Offline cross-encoder reranker (`bge-reranker-base`, `BGERerankerBase`, …).
    pub reranker_model: Option<String>,
    /// Override model cache directory. `None` → fastembed default
    /// (`FASTEMBED_CACHE_DIR` / `HF_HOME` / `./.fastembed_cache`).
    pub cache_dir: Option<PathBuf>,
    /// Show HuggingFace download progress (default: `false` for bindings —
    /// progress bars on a Node/Python stderr are noise).
    pub show_download_progress: bool,
}

struct DenseSlot {
    model: Arc<Mutex<TextEmbedding>>,
    model_name: String,
    model_code: String,
    dim: usize,
}

struct MultiSlot {
    model: Arc<Mutex<Bgem3Embedding>>,
    model_name: String,
    model_code: String,
    /// Per-token ColBERT dimension (1024 for BGE-M3).
    dim: usize,
}

struct ImageSlot {
    model: Arc<Mutex<ImageEmbedding>>,
    model_name: String,
    model_code: String,
    dim: usize,
}

struct RerankSlot {
    model: Arc<Mutex<TextRerank>>,
    model_name: String,
    model_code: String,
}

struct SparseSlot {
    model: Arc<Mutex<SparseTextEmbedding>>,
    model_name: String,
    model_code: String,
}

pub struct FastEmbedder {
    dense: DenseSlot,
    sparse: Option<SparseSlot>,
    multi: Option<MultiSlot>,
    image: Option<ImageSlot>,
    reranker: Option<RerankSlot>,
}

type CacheKey = (String, String);
type CachedModel<T> = Arc<Mutex<T>>;
type ModelCache<T> = Mutex<HashMap<CacheKey, CachedModel<T>>>;

static DENSE_CACHE: OnceLock<ModelCache<TextEmbedding>> = OnceLock::new();
static SPARSE_CACHE: OnceLock<ModelCache<SparseTextEmbedding>> = OnceLock::new();
static MULTI_CACHE: OnceLock<ModelCache<Bgem3Embedding>> = OnceLock::new();
static IMAGE_CACHE: OnceLock<ModelCache<ImageEmbedding>> = OnceLock::new();
static RERANK_CACHE: OnceLock<ModelCache<TextRerank>> = OnceLock::new();

fn cache_dir_key(dir: Option<&PathBuf>) -> String {
    dir.map(|p| p.display().to_string()).unwrap_or_default()
}

fn dense_cache() -> &'static ModelCache<TextEmbedding> {
    DENSE_CACHE.get_or_init(|| Mutex::new(HashMap::new()))
}

fn sparse_cache() -> &'static ModelCache<SparseTextEmbedding> {
    SPARSE_CACHE.get_or_init(|| Mutex::new(HashMap::new()))
}

fn multi_cache() -> &'static ModelCache<Bgem3Embedding> {
    MULTI_CACHE.get_or_init(|| Mutex::new(HashMap::new()))
}

fn image_cache() -> &'static ModelCache<ImageEmbedding> {
    IMAGE_CACHE.get_or_init(|| Mutex::new(HashMap::new()))
}

fn rerank_cache() -> &'static ModelCache<TextRerank> {
    RERANK_CACHE.get_or_init(|| Mutex::new(HashMap::new()))
}

impl FastEmbedder {
    pub fn try_new(options: InitOptionsWithLength<EmbeddingModel>) -> Result<Self, QqlError> {
        let cache_dir = if options.cache_dir != std::path::PathBuf::new() {
            Some(options.cache_dir.clone())
        } else {
            None
        };
        Self::try_with_options(FastEmbedderOptions {
            model: Some(format!("{:?}", options.model_name)),
            sparse_model: None,
            multi_model: None,
            image_model: None,
            reranker_model: None,
            cache_dir,
            show_download_progress: options.show_download_progress,
        })
    }

    /// Construct from high-level options (dense + optional multi, cache, progress).
    pub fn try_with_options(opts: FastEmbedderOptions) -> Result<Self, QqlError> {
        let dense_model = match opts.model.as_deref() {
            None | Some("") => EmbeddingModel::default(),
            Some(name) => resolve_embedding_model(name)?,
        };
        let mut dense_init = InitOptionsWithLength::new(dense_model.clone());
        if let Some(ref dir) = opts.cache_dir {
            dense_init = dense_init.with_cache_dir(dir.clone());
        }
        dense_init = dense_init.with_show_download_progress(opts.show_download_progress);

        let dense_name = format!("{:?}", dense_model);
        let dense_info = TextEmbedding::get_model_info(&dense_model).map_err(|e| {
            err(format!(
                "fastembed has no model info for '{dense_name}': {e}"
            ))
        })?;
        let dense_dim = dense_info.dim;
        let dense_code = dense_info.model_code.clone();

        let cache_key = cache_dir_key(opts.cache_dir.as_ref());
        let dense_handle = dense_cache()
            .lock()
            .map_err(|e| err(format!("fastembed model cache poisoned: {e}")))?
            .get(&(dense_name.clone(), cache_key.clone()))
            .cloned();
        let dense_model_arc = if let Some(model) = dense_handle {
            model
        } else {
            let model = Arc::new(Mutex::new(
                TextEmbedding::try_new(dense_init)
                    .map_err(|e| err(format!("fastembed dense init failed: {e}")))?,
            ));
            let mut cache = dense_cache()
                .lock()
                .map_err(|e| err(format!("fastembed model cache poisoned: {e}")))?;
            Arc::clone(
                cache
                    .entry((dense_name.clone(), cache_key.clone()))
                    .or_insert_with(|| Arc::clone(&model)),
            )
        };

        let sparse = match opts.sparse_model.as_deref() {
            None | Some("") => None,
            Some(name) => {
                let sp = resolve_sparse_model(name)?;
                let info = SparseTextEmbedding::get_model_info(&sp);
                let sparse_name = format!("{:?}", sp);
                let sparse_code = info.model_code.clone();

                let sparse_handle = sparse_cache()
                    .lock()
                    .map_err(|e| err(format!("fastembed sparse cache poisoned: {e}")))?
                    .get(&(sparse_name.clone(), cache_key.clone()))
                    .cloned();
                let sparse_arc = if let Some(model) = sparse_handle {
                    model
                } else {
                    let mut sparse_init = SparseInitOptions::new(sp);
                    if let Some(ref dir) = opts.cache_dir {
                        sparse_init = sparse_init.with_cache_dir(dir.clone());
                    }
                    sparse_init =
                        sparse_init.with_show_download_progress(opts.show_download_progress);
                    let model = Arc::new(Mutex::new(
                        SparseTextEmbedding::try_new(sparse_init).map_err(|e| {
                            err(format!("fastembed SparseTextEmbedding init failed: {e}"))
                        })?,
                    ));
                    let mut cache = sparse_cache()
                        .lock()
                        .map_err(|e| err(format!("fastembed sparse cache poisoned: {e}")))?;
                    Arc::clone(
                        cache
                            .entry((sparse_name.clone(), cache_key.clone()))
                            .or_insert_with(|| Arc::clone(&model)),
                    )
                };
                Some(SparseSlot {
                    model: sparse_arc,
                    model_name: sparse_name,
                    model_code: sparse_code,
                })
            }
        };

        let multi = match opts.multi_model.as_deref() {
            None | Some("") => None,
            Some(name) => {
                let bge = resolve_multi_model(name)?;
                let multi_info = Bgem3Embedding::get_model_info(&bge);
                let multi_name = format!("{:?}", bge);
                let multi_code = multi_info.model_code.clone();
                // BGE-M3 ColBERT token dim matches dense output dim (1024).
                let multi_dim = multi_info.dim;

                let multi_handle = multi_cache()
                    .lock()
                    .map_err(|e| err(format!("fastembed multi cache poisoned: {e}")))?
                    .get(&(multi_name.clone(), cache_key.clone()))
                    .cloned();
                let multi_arc = if let Some(model) = multi_handle {
                    model
                } else {
                    let mut multi_init = Bgem3InitOptions::new(bge);
                    if let Some(ref dir) = opts.cache_dir {
                        multi_init = multi_init.with_cache_dir(dir.clone());
                    }
                    multi_init =
                        multi_init.with_show_download_progress(opts.show_download_progress);
                    let model =
                        Arc::new(Mutex::new(Bgem3Embedding::try_new(multi_init).map_err(
                            |e| err(format!("fastembed multi (BGE-M3) init failed: {e}")),
                        )?));
                    let mut cache = multi_cache()
                        .lock()
                        .map_err(|e| err(format!("fastembed multi cache poisoned: {e}")))?;
                    Arc::clone(
                        cache
                            .entry((multi_name.clone(), cache_key.clone()))
                            .or_insert_with(|| Arc::clone(&model)),
                    )
                };
                Some(MultiSlot {
                    model: multi_arc,
                    model_name: multi_name,
                    model_code: multi_code,
                    dim: multi_dim,
                })
            }
        };

        let image = match opts.image_model.as_deref() {
            None | Some("") => None,
            Some(name) => {
                let img = resolve_image_model(name)?;
                let info = ImageEmbedding::get_model_info(&img);
                let image_name = format!("{:?}", img);
                let image_code = info.model_code.clone();
                let image_dim = info.dim;

                let image_handle = image_cache()
                    .lock()
                    .map_err(|e| err(format!("fastembed image cache poisoned: {e}")))?
                    .get(&(image_name.clone(), cache_key.clone()))
                    .cloned();
                let image_arc = if let Some(model) = image_handle {
                    model
                } else {
                    let mut image_init = ImageInitOptions::new(img);
                    if let Some(ref dir) = opts.cache_dir {
                        image_init = image_init.with_cache_dir(dir.clone());
                    }
                    image_init =
                        image_init.with_show_download_progress(opts.show_download_progress);
                    let model = Arc::new(Mutex::new(ImageEmbedding::try_new(image_init).map_err(
                        |e| err(format!("fastembed image (CLIP vision) init failed: {e}")),
                    )?));
                    let mut cache = image_cache()
                        .lock()
                        .map_err(|e| err(format!("fastembed image cache poisoned: {e}")))?;
                    Arc::clone(
                        cache
                            .entry((image_name.clone(), cache_key.clone()))
                            .or_insert_with(|| Arc::clone(&model)),
                    )
                };
                Some(ImageSlot {
                    model: image_arc,
                    model_name: image_name,
                    model_code: image_code,
                    dim: image_dim,
                })
            }
        };

        let reranker = match opts.reranker_model.as_deref() {
            None | Some("") => None,
            Some(name) => {
                let rm = resolve_reranker_model(name)?;
                let info = TextRerank::get_model_info(&rm);
                let rerank_name = format!("{:?}", rm);
                let rerank_code = info.model_code.clone();
                let handle = rerank_cache()
                    .lock()
                    .map_err(|e| err(format!("fastembed rerank cache poisoned: {e}")))?
                    .get(&(rerank_name.clone(), cache_key.clone()))
                    .cloned();
                let arc = if let Some(model) = handle {
                    model
                } else {
                    let mut init = RerankInitOptions::new(rm);
                    if let Some(ref dir) = opts.cache_dir {
                        init = init.with_cache_dir(dir.clone());
                    }
                    init = init.with_show_download_progress(opts.show_download_progress);
                    let model =
                        Arc::new(Mutex::new(TextRerank::try_new(init).map_err(|e| {
                            err(format!("fastembed TextRerank init failed: {e}"))
                        })?));
                    let mut cache = rerank_cache()
                        .lock()
                        .map_err(|e| err(format!("fastembed rerank cache poisoned: {e}")))?;
                    Arc::clone(
                        cache
                            .entry((rerank_name.clone(), cache_key.clone()))
                            .or_insert_with(|| Arc::clone(&model)),
                    )
                };
                Some(RerankSlot {
                    model: arc,
                    model_name: rerank_name,
                    model_code: rerank_code,
                })
            }
        };

        Ok(Self {
            dense: DenseSlot {
                model: dense_model_arc,
                model_name: dense_name,
                model_code: dense_code,
                dim: dense_dim,
            },
            sparse,
            multi,
            image,
            reranker,
        })
    }

    pub fn try_default() -> Result<Self, QqlError> {
        Self::try_with_options(FastEmbedderOptions::default())
    }

    /// Build from a dense model name string. See [`resolve_embedding_model`].
    pub fn try_from_model(model: &str) -> Result<Self, QqlError> {
        Self::try_with_options(FastEmbedderOptions {
            model: Some(model.to_string()),
            ..Default::default()
        })
    }

    pub fn model_name(&self) -> &str {
        &self.dense.model_name
    }

    pub fn model_code(&self) -> &str {
        &self.dense.model_code
    }

    pub fn dimension(&self) -> usize {
        self.dense.dim
    }

    pub fn multi_model_name(&self) -> Option<&str> {
        self.multi.as_ref().map(|m| m.model_name.as_str())
    }

    pub fn multi_model_code(&self) -> Option<&str> {
        self.multi.as_ref().map(|m| m.model_code.as_str())
    }

    pub fn multi_dimension(&self) -> Option<usize> {
        self.multi.as_ref().map(|m| m.dim)
    }

    pub fn has_multi(&self) -> bool {
        self.multi.is_some()
    }

    pub fn image_model_name(&self) -> Option<&str> {
        self.image.as_ref().map(|m| m.model_name.as_str())
    }

    pub fn image_model_code(&self) -> Option<&str> {
        self.image.as_ref().map(|m| m.model_code.as_str())
    }

    pub fn image_dimension(&self) -> Option<usize> {
        self.image.as_ref().map(|m| m.dim)
    }

    pub fn has_image(&self) -> bool {
        self.image.is_some()
    }

    pub fn has_reranker(&self) -> bool {
        self.reranker.is_some()
    }

    pub fn sparse_model_name(&self) -> Option<&str> {
        self.sparse.as_ref().map(|s| s.model_name.as_str())
    }

    pub fn sparse_model_code(&self) -> Option<&str> {
        self.sparse.as_ref().map(|s| s.model_code.as_str())
    }

    pub fn has_sparse(&self) -> bool {
        self.sparse.is_some()
    }

    pub fn reranker_model_code(&self) -> Option<&str> {
        self.reranker.as_ref().map(|m| m.model_code.as_str())
    }

    fn accepts_reranker_model(&self, requested: &str) -> bool {
        let Some(ref r) = self.reranker else {
            return false;
        };
        let req = requested.trim();
        if req.is_empty() || req.eq_ignore_ascii_case("default") {
            return true;
        }
        req.eq_ignore_ascii_case(&r.model_name)
            || req.eq_ignore_ascii_case(&r.model_code)
            || short_alias_matches(req, &r.model_code)
            || is_reranker_alias(req)
    }

    /// Whether a QQL `USING MODEL '…'` / `MODEL '…'` string refers to this embedder.
    /// Empty / `"default"` always match (host did not pin a model).
    pub fn accepts_model(&self, requested: &str) -> bool {
        let r = requested.trim();
        if r.is_empty() || r.eq_ignore_ascii_case("default") {
            return true;
        }
        if r.eq_ignore_ascii_case(&self.dense.model_name)
            || r.eq_ignore_ascii_case(&self.dense.model_code)
            || short_alias_matches(r, &self.dense.model_code)
        {
            return true;
        }
        if let Some(ref sparse) = self.sparse {
            if r.eq_ignore_ascii_case(&sparse.model_name)
                || r.eq_ignore_ascii_case(&sparse.model_code)
                || short_alias_matches(r, &sparse.model_code)
                || is_sparse_alias(r)
            {
                return true;
            }
        }
        if let Some(ref multi) = self.multi {
            if r.eq_ignore_ascii_case(&multi.model_name)
                || r.eq_ignore_ascii_case(&multi.model_code)
                || short_alias_matches(r, &multi.model_code)
                || is_multi_alias(r)
            {
                return true;
            }
        }
        if let Some(ref image) = self.image {
            if r.eq_ignore_ascii_case(&image.model_name)
                || r.eq_ignore_ascii_case(&image.model_code)
                || short_alias_matches(r, &image.model_code)
                || is_image_alias(r)
            {
                return true;
            }
        }
        false
    }

    fn accepts_image_model(&self, requested: &str) -> bool {
        let Some(ref image) = self.image else {
            return false;
        };
        let r = requested.trim();
        if r.is_empty() || r.eq_ignore_ascii_case("default") {
            return true;
        }
        r.eq_ignore_ascii_case(&image.model_name)
            || r.eq_ignore_ascii_case(&image.model_code)
            || short_alias_matches(r, &image.model_code)
            || is_image_alias(r)
    }

    fn accepts_dense_model(&self, requested: &str) -> bool {
        let r = requested.trim();
        if r.is_empty() || r.eq_ignore_ascii_case("default") {
            return true;
        }
        r.eq_ignore_ascii_case(&self.dense.model_name)
            || r.eq_ignore_ascii_case(&self.dense.model_code)
            || short_alias_matches(r, &self.dense.model_code)
    }

    fn accepts_sparse_model(&self, requested: &str) -> bool {
        let Some(ref sparse) = self.sparse else {
            return false;
        };
        let r = requested.trim();
        if r.is_empty() || r.eq_ignore_ascii_case("default") {
            return true;
        }
        r.eq_ignore_ascii_case(&sparse.model_name)
            || r.eq_ignore_ascii_case(&sparse.model_code)
            || short_alias_matches(r, &sparse.model_code)
            || is_sparse_alias(r)
    }

    fn accepts_multi_model(&self, requested: &str) -> bool {
        let Some(ref multi) = self.multi else {
            return false;
        };
        let r = requested.trim();
        if r.is_empty() || r.eq_ignore_ascii_case("default") {
            return true;
        }
        r.eq_ignore_ascii_case(&multi.model_name)
            || r.eq_ignore_ascii_case(&multi.model_code)
            || short_alias_matches(r, &multi.model_code)
            || is_multi_alias(r)
    }
}

impl std::fmt::Debug for FastEmbedder {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.debug_struct("FastEmbedder")
            .field("model_name", &self.dense.model_name)
            .field("model_code", &self.dense.model_code)
            .field("dim", &self.dense.dim)
            .field(
                "sparse_model",
                &self.sparse.as_ref().map(|m| m.model_code.as_str()),
            )
            .field(
                "multi_model",
                &self.multi.as_ref().map(|m| m.model_code.as_str()),
            )
            .field("multi_dim", &self.multi.as_ref().map(|m| m.dim))
            .field(
                "image_model",
                &self.image.as_ref().map(|m| m.model_code.as_str()),
            )
            .field("image_dim", &self.image.as_ref().map(|m| m.dim))
            .finish()
    }
}

/// List dense text models, offline sparse models (SPLADE / BGE-M3),
/// multi (BGE-M3 / ColBERT), and image models (CLIP vision, …) that
/// fastembed can load.
pub fn list_embedding_models() -> Vec<EmbeddingModelInfo> {
    let mut models: Vec<EmbeddingModelInfo> = TextEmbedding::list_supported_models()
        .into_iter()
        .map(|m| EmbeddingModelInfo {
            name: format!("{:?}", m.model),
            model_code: m.model_code,
            dim: m.dim,
            description: m.description,
            multi: false,
            image: false,
        })
        .collect();
    for m in SparseTextEmbedding::list_supported_models() {
        models.push(EmbeddingModelInfo {
            name: format!("{:?}", m.model),
            model_code: m.model_code,
            dim: m.dim,
            description: format!("{} (sparse / SPLADE)", m.description),
            multi: false,
            image: false,
        });
    }
    for m in Bgem3Embedding::list_supported_models() {
        models.push(EmbeddingModelInfo {
            name: format!("{:?}", m.model),
            model_code: m.model_code,
            dim: m.dim,
            description: format!("{} (multivector / ColBERT via BGE-M3)", m.description),
            multi: true,
            image: false,
        });
    }
    for m in ImageEmbedding::list_supported_models() {
        models.push(EmbeddingModelInfo {
            name: format!("{:?}", m.model),
            model_code: m.model_code,
            dim: m.dim,
            description: format!("{} (image / CLIP vision)", m.description),
            multi: false,
            image: true,
        });
    }
    models
}

/// Resolve a user-facing dense model string to an [`EmbeddingModel`].
///
/// Accepts, case-insensitively:
/// - enum Debug names: `BGESmallENV15`, `AllMiniLML6V2`
/// - HuggingFace model codes: `Xenova/bge-small-en-v1.5`
/// - short aliases: `bge-small-en-v1.5`, `all-minilm-l6-v2`
pub fn resolve_embedding_model(name: &str) -> Result<EmbeddingModel, QqlError> {
    let name = name.trim();
    if name.is_empty() {
        return Ok(EmbeddingModel::default());
    }

    // 1. Debug / FromStr name
    if let Ok(m) = name.parse::<EmbeddingModel>() {
        return Ok(m);
    }

    // 2. Exact model_code / short slug (case-insensitive)
    for info in TextEmbedding::list_supported_models() {
        if info.model_code.eq_ignore_ascii_case(name) {
            return Ok(info.model);
        }
        if short_alias_matches(name, &info.model_code) {
            return Ok(info.model);
        }
        if let Some(slug) = info.model_code.rsplit('/').next() {
            if slug.eq_ignore_ascii_case(name) {
                return Ok(info.model);
            }
        }
    }

    // Suggest a few options so callers aren't left guessing
    let mut suggestions: Vec<String> = TextEmbedding::list_supported_models()
        .into_iter()
        .map(|m| format!("{:?} ({}, {}-d)", m.model, m.model_code, m.dim))
        .take(6)
        .collect();
    suggestions.sort();
    Err(err(format!(
        "unknown embedding model '{name}'. Use list_embedding_models() for the full list. Examples: {}",
        suggestions.join("; ")
    )))
}

/// Resolve offline multi model (`bge-m3`, HF code, etc.).
pub fn resolve_multi_model(name: &str) -> Result<Bgem3Model, QqlError> {
    let name = name.trim();
    if name.is_empty() || is_multi_alias(name) {
        return Ok(Bgem3Model::default());
    }
    if let Ok(m) = name.parse::<Bgem3Model>() {
        return Ok(m);
    }
    for info in Bgem3Embedding::list_supported_models() {
        if info.model_code.eq_ignore_ascii_case(name)
            || short_alias_matches(name, &info.model_code)
            || format!("{:?}", info.model).eq_ignore_ascii_case(name)
        {
            return Ok(info.model);
        }
    }
    Err(err(format!(
        "unknown multi embedding model '{name}'. Offline multi uses BGE-M3 \
         (e.g. 'bge-m3', 'BGEM3Q', 'gpahal/bge-m3-onnx-int8')"
    )))
}

fn is_multi_alias(name: &str) -> bool {
    matches!(
        name.to_ascii_lowercase().as_str(),
        "bge-m3" | "bgem3" | "bgem3q" | "colbert" | "multi" | "multivector" | "late-interaction"
    )
}

/// Resolve offline image model (CLIP vision, etc.).
pub fn resolve_image_model(name: &str) -> Result<ImageEmbeddingModel, QqlError> {
    let name = name.trim();
    if name.is_empty() || is_image_alias(name) {
        return Ok(ImageEmbeddingModel::default());
    }
    if let Ok(m) = name.parse::<ImageEmbeddingModel>() {
        return Ok(m);
    }
    for info in ImageEmbedding::list_supported_models() {
        if info.model_code.eq_ignore_ascii_case(name)
            || short_alias_matches(name, &info.model_code)
            || format!("{:?}", info.model).eq_ignore_ascii_case(name)
        {
            return Ok(info.model);
        }
    }
    Err(err(format!(
        "unknown image embedding model '{name}'. Offline image uses CLIP vision \
         (e.g. 'ClipVitB32', 'Qdrant/clip-ViT-B-32-vision', 'clip-vision')"
    )))
}

fn is_image_alias(name: &str) -> bool {
    matches!(
        name.to_ascii_lowercase().as_str(),
        "clip"
            | "clip-vision"
            | "clip_vision"
            | "clip-vit-b-32"
            | "clip-vit-b-32-vision"
            | "image"
            | "vision"
    )
}

/// Resolve offline cross-encoder model id.
pub fn resolve_reranker_model(name: &str) -> Result<RerankerModel, QqlError> {
    let name = name.trim();
    if name.is_empty() || is_reranker_alias(name) {
        return Ok(RerankerModel::default());
    }
    if let Ok(m) = name.parse::<RerankerModel>() {
        return Ok(m);
    }
    for info in TextRerank::list_supported_models() {
        if info.model_code.eq_ignore_ascii_case(name)
            || short_alias_matches(name, &info.model_code)
            || format!("{:?}", info.model).eq_ignore_ascii_case(name)
        {
            return Ok(info.model);
        }
    }
    Err(err(format!(
        "unknown reranker model '{name}'. Examples: bge-reranker-base, BGERerankerBase, \
         BAAI/bge-reranker-base, jina-reranker-v1-turbo-en"
    )))
}

fn is_reranker_alias(name: &str) -> bool {
    matches!(
        name.to_ascii_lowercase().as_str(),
        "rerank"
            | "reranker"
            | "cross-encoder"
            | "cross_encoder"
            | "bge-reranker"
            | "bge-reranker-base"
    )
}

/// Resolve offline sparse model (SPLADE, BGE-M3 sparse).
pub fn resolve_sparse_model(name: &str) -> Result<SparseModel, QqlError> {
    let name = name.trim();
    if name.is_empty()
        || matches!(
            name.to_ascii_lowercase().as_str(),
            "splade" | "spladeppv1" | "splade_pp_en_v1" | "sparse" | "bm25"
        )
    {
        return Ok(SparseModel::default());
    }
    if matches!(name.to_ascii_lowercase().as_str(), "bge-m3" | "bgem3") {
        return Ok(SparseModel::BGEM3);
    }
    if let Ok(m) = name.parse::<SparseModel>() {
        return Ok(m);
    }
    for info in SparseTextEmbedding::list_supported_models() {
        if info.model_code.eq_ignore_ascii_case(name)
            || short_alias_matches(name, &info.model_code)
            || format!("{:?}", info.model).eq_ignore_ascii_case(name)
        {
            return Ok(info.model);
        }
    }
    Err(err(format!(
        "unknown sparse embedding model '{name}'. Offline sparse supports \
         'splade' (SPLADEPPV1) and 'bge-m3' (BGEM3)"
    )))
}

fn is_sparse_alias(name: &str) -> bool {
    matches!(
        name.to_ascii_lowercase().as_str(),
        "splade" | "spladeppv1" | "splade_pp_en_v1" | "sparse" | "bm25" | "bge-m3" | "bgem3"
    )
}

fn short_alias_matches(requested: &str, model_code: &str) -> bool {
    let req = requested.trim().trim_matches('"');
    let code = model_code;
    if req.eq_ignore_ascii_case(code) {
        return true;
    }
    // strip org prefix: "Xenova/bge-small-en-v1.5" ↔ "bge-small-en-v1.5"
    if let Some(slug) = code.rsplit('/').next() {
        if req.eq_ignore_ascii_case(slug) {
            return true;
        }
    }
    // Strip common suffixes people omit when referring to a converted model.
    [
        "-onnx-q",
        "-onnx",
        "-q4_k_m",
        "-q8_0",
        "-onnx-int8",
        "-int8",
    ]
    .iter()
    .any(|suffix| {
        code.strip_suffix(suffix)
            .is_some_and(|base| req.eq_ignore_ascii_case(base))
            || code
                .rsplit('/')
                .next()
                .and_then(|slug| slug.strip_suffix(suffix))
                .is_some_and(|base| req.eq_ignore_ascii_case(base))
    })
}

#[async_trait]
impl Embedder for FastEmbedder {
    fn dimension(&self) -> Option<usize> {
        Some(self.dense.dim)
    }

    fn multi_dimension(&self) -> Option<usize> {
        self.multi.as_ref().map(|m| m.dim)
    }

    fn accepts_model(&self, model: &str) -> bool {
        self.accepts_model(model)
    }

    // image_dimension is not on Embedder trait; use dimension() for dense CLIP text.
    // Image dim available via FastEmbedder::image_dimension().

    async fn embed_dense(&self, text: &str, model: &str) -> Result<Vec<f32>, QqlError> {
        if !self.accepts_dense_model(model) {
            // Multi-only model id on dense path is a clear mistake.
            if self.accepts_multi_model(model) {
                return Err(err(format!(
                    "model '{model}' is the multi/ColBERT model on this edge embedder; \
                     use it with AS MULTI / multivector RERANK, not dense embedding. \
                     Dense model is '{}' ({}).",
                    self.dense.model_name, self.dense.model_code
                )));
            }
            return Err(err(format!(
                "local embedder is locked to dense model '{}' ({}); cannot satisfy USING MODEL '{model}'. \
                 Create the executor with model='{model}' (or omit MODEL to use the locked one).",
                self.dense.model_name, self.dense.model_code
            )));
        }

        let model = self.dense.model.clone();
        let texts = vec![text.to_string()];

        let mut embeddings = tokio::task::spawn_blocking(move || {
            let mut model = model
                .lock()
                .map_err(|e| err(format!("fastembed mutex poisoned: {e}")))?;
            model
                .embed(texts, None)
                .map_err(|e| err(format!("fastembed failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        embeddings
            .pop()
            .ok_or_else(|| err("fastembed returned empty result"))
    }

    async fn embed_dense_batch(
        &self,
        texts: &[String],
        model: &str,
    ) -> Result<Vec<Vec<f32>>, QqlError> {
        if texts.is_empty() {
            return Ok(vec![]);
        }
        if !self.accepts_dense_model(model) {
            if self.accepts_multi_model(model) {
                return Err(err(format!(
                    "model '{model}' is the multi/ColBERT model on this edge embedder; \
                     use it with AS MULTI / multivector RERANK, not dense embedding. \
                     Dense model is '{}' ({}).",
                    self.dense.model_name, self.dense.model_code
                )));
            }
            return Err(err(format!(
                "local embedder is locked to dense model '{}' ({}); cannot satisfy USING MODEL '{model}'. \
                 Create the executor with model='{model}' (or omit MODEL to use the locked one).",
                self.dense.model_name, self.dense.model_code
            )));
        }

        let model = self.dense.model.clone();
        let batch = texts.to_vec();

        let embeddings = tokio::task::spawn_blocking(move || {
            let mut model = model
                .lock()
                .map_err(|e| err(format!("fastembed mutex poisoned: {e}")))?;
            model
                .embed(batch, None)
                .map_err(|e| err(format!("fastembed batch failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        Ok(embeddings)
    }

    async fn embed_sparse(&self, text: &str, model: &str) -> Result<SparseVector, QqlError> {
        let Some(ref sparse) = self.sparse else {
            // No sparse model configured: accept default/empty models for local
            // BM25, but reject explicit non-default models.
            if !model.is_empty() && !model.eq_ignore_ascii_case("default") {
                return Err(qql_embed::sparse_model_unsupported_error(model));
            }
            return Ok(qql_embed::sparse::build_query_default(text));
        };

        if !(self.accepts_sparse_model(model)
            || model.is_empty()
            || model.eq_ignore_ascii_case("default"))
        {
            return Err(err(format!(
                "local sparse embedder is locked to '{}' ({}); cannot satisfy MODEL '{model}'",
                sparse.model_name, sparse.model_code
            )));
        }

        let model_arc = sparse.model.clone();
        let texts = vec![text.to_string()];

        let embeddings = tokio::task::spawn_blocking(move || {
            let mut model = model_arc
                .lock()
                .map_err(|e| err(format!("fastembed sparse mutex poisoned: {e}")))?;
            model
                .embed(texts, None)
                .map_err(|e| err(format!("fastembed SparseTextEmbedding failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        let emb = embeddings
            .into_iter()
            .next()
            .map(|e| SparseVector {
                indices: e.indices.iter().map(|&i| i as u32).collect(),
                values: e.values.clone(),
            })
            .ok_or_else(|| err("fastembed SparseTextEmbedding returned no result"))?;

        Ok(emb)
    }

    /// Single-pass BGE-M3 joint embedding: one `Bgem3Embedding::embed` call
    /// yields dense, sparse, and ColBERT together. Falls back to the default
    /// three-call implementation when no BGE-M3 multi model is configured.
    async fn embed_joint(&self, text: &str, model: &str) -> Result<JointEmbeddingOutput, QqlError> {
        let Some(ref multi) = self.multi else {
            // No BGE-M3 model: delegate to default per-call impl (non-optimal
            // but correct — no error suppression).
            let dense = self.embed_dense(text, model).await?;
            let sparse = self.embed_sparse(text, model).await?;
            let multi_vec = self.embed_multi(text, model).await?;
            return Ok(JointEmbeddingOutput {
                dense: Some(dense),
                sparse: Some(sparse),
                multi: Some(multi_vec),
            });
        };

        if !(self.accepts_multi_model(model)
            || model.is_empty()
            || model.eq_ignore_ascii_case("default"))
        {
            return Err(err(format!(
                "local joint embedder uses BGE-M3 '{}' ({}); cannot satisfy MODEL '{model}'",
                multi.model_name, multi.model_code
            )));
        }

        let model_arc = multi.model.clone();
        let texts = vec![text.to_string()];

        let output = tokio::task::spawn_blocking(move || {
            let mut m = model_arc
                .lock()
                .map_err(|e| err(format!("fastembed joint mutex poisoned: {e}")))?;
            m.embed(texts, None)
                .map_err(|e| err(format!("fastembed BGE-M3 joint failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        let dense = output.dense.into_iter().next();
        let sparse = output.sparse.into_iter().next().map(|e| SparseVector {
            indices: e.indices.iter().map(|&i| i as u32).collect(),
            values: e.values.clone(),
        });
        let colbert = output.colbert.into_iter().next();

        Ok(JointEmbeddingOutput {
            dense,
            sparse,
            multi: colbert,
        })
    }

    async fn embed_multi(&self, text: &str, model: &str) -> Result<Vec<Vec<f32>>, QqlError> {
        let Some(ref multi) = self.multi else {
            return Err(qql_embed::multi_unsupported_error(model));
        };
        if !self.accepts_multi_model(model) && !self.accepts_dense_model(model) {
            // Accept dense model id only when multi is configured and model is defaulted;
            // explicit wrong model still errors.
            if !model.is_empty() && !model.eq_ignore_ascii_case("default") {
                return Err(err(format!(
                    "local multi embedder is locked to '{}' ({}); cannot satisfy MODEL '{model}'",
                    multi.model_name, multi.model_code
                )));
            }
        }

        let model_arc = multi.model.clone();
        let texts = vec![text.to_string()];

        let mut output = tokio::task::spawn_blocking(move || {
            let mut model = model_arc
                .lock()
                .map_err(|e| err(format!("fastembed multi mutex poisoned: {e}")))?;
            model
                .embed(texts, None)
                .map_err(|e| err(format!("fastembed multi (BGE-M3) failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        output
            .colbert
            .pop()
            .filter(|rows| !rows.is_empty())
            .ok_or_else(|| err("fastembed multi returned empty ColBERT bag"))
    }

    async fn embed_multi_batch(
        &self,
        texts: &[String],
        model: &str,
    ) -> Result<Vec<Vec<Vec<f32>>>, QqlError> {
        if texts.is_empty() {
            return Ok(vec![]);
        }
        let Some(ref multi) = self.multi else {
            return Err(qql_embed::multi_unsupported_error(model));
        };
        if !(self.accepts_multi_model(model)
            || model.is_empty()
            || model.eq_ignore_ascii_case("default"))
        {
            return Err(err(format!(
                "local multi embedder is locked to '{}' ({}); cannot satisfy MODEL '{model}'",
                multi.model_name, multi.model_code
            )));
        }

        let model_arc = multi.model.clone();
        let batch = texts.to_vec();

        let output = tokio::task::spawn_blocking(move || {
            let mut model = model_arc
                .lock()
                .map_err(|e| err(format!("fastembed multi mutex poisoned: {e}")))?;
            model
                .embed(batch, None)
                .map_err(|e| err(format!("fastembed multi batch failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        if output.colbert.iter().any(|rows| rows.is_empty()) {
            return Err(err("fastembed multi returned an empty ColBERT bag"));
        }
        Ok(output.colbert)
    }

    async fn embed_image(&self, source: &str, model: &str) -> Result<Vec<f32>, QqlError> {
        let Some(ref image) = self.image else {
            return Err(qql_embed::image_unsupported_error(model));
        };
        if !(self.accepts_image_model(model)
            || model.is_empty()
            || model.eq_ignore_ascii_case("default"))
        {
            return Err(err(format!(
                "local image embedder is locked to '{}' ({}); cannot satisfy MODEL '{model}'",
                image.model_name, image.model_code
            )));
        }

        let model_arc = image.model.clone();
        let path = source.to_string();

        let mut embeddings = tokio::task::spawn_blocking(move || {
            let mut model = model_arc
                .lock()
                .map_err(|e| err(format!("fastembed image mutex poisoned: {e}")))?;
            model
                .embed(vec![path], None)
                .map_err(|e| err(format!("fastembed image embed failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        embeddings
            .pop()
            .ok_or_else(|| err("fastembed image returned empty result"))
    }

    async fn embed_image_batch(
        &self,
        sources: &[String],
        model: &str,
    ) -> Result<Vec<Vec<f32>>, QqlError> {
        if sources.is_empty() {
            return Ok(vec![]);
        }
        let Some(ref image) = self.image else {
            return Err(qql_embed::image_unsupported_error(model));
        };
        if !(self.accepts_image_model(model)
            || model.is_empty()
            || model.eq_ignore_ascii_case("default"))
        {
            return Err(err(format!(
                "local image embedder is locked to '{}' ({}); cannot satisfy MODEL '{model}'",
                image.model_name, image.model_code
            )));
        }

        let model_arc = image.model.clone();
        let batch = sources.to_vec();

        let embeddings = tokio::task::spawn_blocking(move || {
            let mut model = model_arc
                .lock()
                .map_err(|e| err(format!("fastembed image mutex poisoned: {e}")))?;
            model
                .embed(batch, None)
                .map_err(|e| err(format!("fastembed image batch failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        Ok(embeddings)
    }

    async fn rerank_pairs(
        &self,
        query: &str,
        documents: &[String],
        model: &str,
    ) -> Result<Vec<f32>, QqlError> {
        if documents.is_empty() {
            return Ok(vec![]);
        }
        let Some(ref reranker) = self.reranker else {
            return Err(qql_embed::cross_rerank_unsupported_error(model));
        };
        if !(self.accepts_reranker_model(model)
            || model.is_empty()
            || model.eq_ignore_ascii_case("default"))
        {
            return Err(err(format!(
                "local reranker is locked to '{}' ({}); cannot satisfy MODEL '{model}'",
                reranker.model_name, reranker.model_code
            )));
        }

        let model_arc = reranker.model.clone();
        let q = query.to_string();
        let docs = documents.to_vec();

        let ranked = tokio::task::spawn_blocking(move || {
            let mut model = model_arc
                .lock()
                .map_err(|e| err(format!("fastembed rerank mutex poisoned: {e}")))?;
            model
                .rerank(q, docs, false, None)
                .map_err(|e| err(format!("fastembed TextRerank failed: {e}")))
        })
        .await
        .map_err(|e| err(format!("spawn_blocking failed: {e}")))??;

        // Unpermute to original document order.
        let mut scores = vec![0.0f32; documents.len()];
        for item in ranked {
            if item.index < scores.len() {
                scores[item.index] = item.score;
            }
        }
        Ok(scores)
    }
}

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

    #[test]
    fn cache_dir_key_empty_for_none() {
        assert_eq!(cache_dir_key(None), "");
    }

    #[test]
    fn cache_dir_key_preserves_path() {
        let p = PathBuf::from("/tmp/my_cache");
        assert_eq!(cache_dir_key(Some(&p)), "/tmp/my_cache");
    }

    #[test]
    fn resolve_sparse_model_splade_default() {
        let m = resolve_sparse_model("").unwrap();
        assert_eq!(m, SparseModel::SPLADEPPV1);
    }

    #[test]
    fn resolve_sparse_model_by_alias() {
        let m = resolve_sparse_model("splade").unwrap();
        assert_eq!(m, SparseModel::SPLADEPPV1);
    }

    #[test]
    fn resolve_sparse_model_by_enum_name() {
        let m = resolve_sparse_model("SPLADEPPV1").unwrap();
        assert_eq!(m, SparseModel::SPLADEPPV1);
    }

    #[test]
    fn resolve_sparse_model_by_model_code() {
        let m = resolve_sparse_model("Qdrant/Splade_PP_en_v1").unwrap();
        assert_eq!(m, SparseModel::SPLADEPPV1);
    }

    #[test]
    fn resolve_sparse_model_bgem3() {
        let m = resolve_sparse_model("bge-m3").unwrap();
        assert_eq!(m, SparseModel::BGEM3);
        let m = resolve_sparse_model("BGEM3").unwrap();
        assert_eq!(m, SparseModel::BGEM3);
        let m = resolve_sparse_model("BAAI/bge-m3").unwrap();
        assert_eq!(m, SparseModel::BGEM3);
    }

    #[test]
    fn resolve_sparse_model_unknown_errors() {
        let e = resolve_sparse_model("nonexistent_model").unwrap_err();
        assert!(e.message.contains("nonexistent_model"));
    }

    #[test]
    fn is_sparse_alias_matches() {
        assert!(is_sparse_alias("splade"));
        assert!(is_sparse_alias("SPLADE"));
        assert!(is_sparse_alias("bge-m3"));
        assert!(is_sparse_alias("bgem3"));
        assert!(!is_sparse_alias("unknown"));
    }

    #[test]
    fn options_default_has_no_sparse_model() {
        let opts = FastEmbedderOptions::default();
        assert!(opts.sparse_model.is_none());
        assert!(opts.model.is_none());
    }

    #[test]
    fn options_with_sparse_model() {
        let opts = FastEmbedderOptions {
            sparse_model: Some("splade".into()),
            ..Default::default()
        };
        assert_eq!(opts.sparse_model.as_deref(), Some("splade"));
    }
}