zeph-channels 0.21.0

Multi-channel I/O adapters (CLI, Telegram, Discord, Slack) for Zeph
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
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
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
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
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
1842
1843
1844
1845
1846
1847
1848
1849
1850
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
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
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
// SPDX-FileCopyrightText: 2026 Andrei G <bug-ops>
// SPDX-License-Identifier: MIT OR Apache-2.0

//! Telegram channel adapter built on [teloxide](https://docs.rs/teloxide).
//!
//! This module exposes [`TelegramChannel`], which implements [`Channel`] by
//! wrapping a teloxide [`Dispatcher`] running in a background tokio task.
//!
//! # Key design decisions
//!
//! * **Access control** — messages from users not in `allowed_users` are
//!   dropped at the dispatcher level, before they reach the agent.
//! * **Streaming via edit-on-interval** — chunks are accumulated and the bot
//!   edits a single Telegram message at most once every 3 seconds, staying
//!   within Telegram's rate limits while still providing progressive output.
//! * **4096-byte chunking** — messages that exceed Telegram's limit are
//!   split at UTF-8 / newline boundaries via [`utf8_chunks`].
//! * **Elicitation** — MCP server input requests are forwarded to the user
//!   as sequential Telegram messages with per-field 120-second timeouts.
//!
//! [`Channel`]: zeph_core::channel::Channel
//! [`utf8_chunks`]: crate::markdown::utf8_chunks

use std::collections::HashMap;
use std::sync::atomic::{AtomicBool, Ordering};
use std::sync::{Arc, Mutex};
use std::time::{Duration, Instant};

use crate::markdown::markdown_to_telegram;
use crate::telegram_api_ext::{BotAccessSettings, GuestMessage, TelegramApiClient};
use axum::Router;
use axum::body::Body;
use axum::extract::State;
use axum::http::{Request, StatusCode};
use axum::response::{IntoResponse, Response};
use axum::routing::any;
use teloxide::prelude::*;
use teloxide::types::{BotCommand, ChatAction, MessageId, ParseMode};
use tokio::sync::mpsc;
use zeph_common::TaskSupervisor;
use zeph_core::channel::{
    Attachment, AttachmentKind, Channel, ChannelError, ChannelMessage, ElicitationField,
    ElicitationFieldType, ElicitationRequest, ElicitationResponse,
};

const MAX_MESSAGE_LEN: usize = 4096;
const MAX_IMAGE_BYTES: u32 = 20 * 1024 * 1024;
const MAX_TRACKED_CHATS: usize = 1000;

type BotReplyCounters = Arc<Mutex<HashMap<ChatId, u32>>>;

/// Telegram channel adapter using [teloxide](https://docs.rs/teloxide).
///
/// `TelegramChannel` bridges the Zeph agent loop with the Telegram Bot API.
/// It runs a teloxide [`Dispatcher`] in a background task that feeds incoming
/// messages into an internal [`mpsc`] channel; the agent calls [`recv`] to
/// receive them one at a time.
///
/// # Streaming output
///
/// LLM responses are streamed to Telegram via an edit-on-interval strategy:
/// chunks are accumulated in memory and the bot edits a single message every
/// three seconds.  This avoids hitting Telegram's rate limits while still
/// providing a progressive output experience.  [`flush_chunks`] performs one
/// final edit with the complete response text.
///
/// # Access control
///
/// The bot only accepts messages from usernames listed in `allowed_users`.
/// Messages from any other user are silently dropped.  Passing an empty
/// `allowed_users` list is treated as a configuration error and [`start`]
/// will return `Err`.
///
/// # Built-in commands
///
/// | Command | Behaviour |
/// |---------|-----------|
/// | `/start` | Sends a welcome message; not forwarded to the agent. |
/// | `/reset` | Forwarded to the agent as a regular message. |
/// | `/skills` | Forwarded to the agent as a regular message. |
/// | `/agent` | Forwarded to the agent as a regular message. |
///
/// # Examples
///
/// ```rust,no_run
/// use zeph_channels::telegram::TelegramChannel;
///
/// let token = std::env::var("TELEGRAM_BOT_TOKEN").unwrap();
/// let allowed = vec!["my_username".to_string()];
/// let channel = TelegramChannel::new(token, allowed)
///     .with_stream_interval(std::time::Duration::from_millis(2000))
///     .start()
///     .expect("failed to start telegram bot");
/// ```
///
/// [`recv`]: TelegramChannel::recv
/// [`flush_chunks`]: TelegramChannel::flush_chunks
/// [`start`]: TelegramChannel::start
pub struct TelegramChannel {
    bot: Bot,
    chat_id: Option<ChatId>,
    rx: mpsc::Receiver<IncomingMessage>,
    allowed_users: Vec<String>,
    accumulated: String,
    last_edit: Option<Instant>,
    message_id: Option<MessageId>,
    /// Minimum interval between streaming message edits. Defaults to 3 seconds.
    stream_interval: Duration,
    /// Raw Bot API 10.0 client for methods not yet covered by teloxide.
    api_ext: TelegramApiClient,
    /// Enable Guest Mode: respond to @mentions via `answerGuestQuery`.
    guest_mode: bool,
    /// Active guest query ID for the current response cycle.
    guest_query_id: Option<String>,
    /// Enable bot-to-bot communication.
    bot_to_bot: bool,
    /// Set to `true` by the background task when `setManagedBotAccessSettings` succeeds.
    /// Remains `false` if the API call fails, disabling bot-to-bot processing.
    bot_to_bot_active: Arc<AtomicBool>,
    /// Allowlist of bot usernames permitted when `bot_to_bot = true`. Empty = all bots.
    allowed_bots: Vec<String>,
    /// Maximum consecutive bot replies per chat before dropping.
    max_bot_chain_depth: u32,
    /// Per-chat consecutive bot reply counters for loop prevention.
    bot_reply_counters: BotReplyCounters,
    /// Optional supervisor used to register the Telegram listener task in the
    /// workspace-wide task registry with automatic restart on panic.
    supervisor: Option<TaskSupervisor>,
    /// Handle to the guest-mode axum proxy task. Kept alive for the lifetime
    /// of the channel; dropped when `TelegramChannel` is dropped.
    #[allow(dead_code)]
    guest_proxy_handle: Option<tokio::task::JoinHandle<()>>,
}

impl std::fmt::Debug for TelegramChannel {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.debug_struct("TelegramChannel")
            .field("chat_id", &self.chat_id)
            .field("allowed_users", &self.allowed_users)
            .field("accumulated_len", &self.accumulated.len())
            .field("stream_interval_ms", &self.stream_interval.as_millis())
            .field("guest_mode", &self.guest_mode)
            .field("bot_to_bot", &self.bot_to_bot)
            .field(
                "bot_to_bot_active",
                &self.bot_to_bot_active.load(Ordering::Relaxed),
            )
            .field("allowed_bots_count", &self.allowed_bots.len())
            .field("max_bot_chain_depth", &self.max_bot_chain_depth)
            .field("supervisor", &self.supervisor.is_some())
            .finish_non_exhaustive()
    }
}

#[derive(Debug)]
struct IncomingMessage {
    chat_id: ChatId,
    text: String,
    attachments: Vec<Attachment>,
    /// `Some(query_id)` when this message came from a `guest_message` update.
    guest_query_id: Option<String>,
    /// `true` when the sender is a Telegram bot (`from.is_bot = true`).
    is_from_bot: bool,
}

impl TelegramChannel {
    /// Create a new `TelegramChannel`.
    ///
    /// The channel is **not** active yet — no teloxide dispatcher is running
    /// and no updates will be received until [`start`] is called.
    ///
    /// # Arguments
    ///
    /// * `token` — Telegram bot token obtained from [@BotFather](https://t.me/BotFather).
    /// * `allowed_users` — Telegram usernames (without `@`) that are permitted
    ///   to interact with the bot.  Must not be empty when [`start`] is called.
    ///
    /// Use [`with_stream_interval`] to change the default 3-second streaming edit interval.
    ///
    /// [`start`]: TelegramChannel::start
    /// [`with_stream_interval`]: TelegramChannel::with_stream_interval
    #[must_use]
    pub fn new(token: impl Into<String>, allowed_users: Vec<String>) -> Self {
        let token = token.into();
        let bot = Bot::new(&token);
        let api_ext = TelegramApiClient::new(&token);
        let (_, rx) = mpsc::channel(64);
        Self {
            bot,
            chat_id: None,
            rx,
            allowed_users,
            accumulated: String::new(),
            last_edit: None,
            message_id: None,
            stream_interval: Duration::from_secs(3),
            api_ext,
            guest_mode: false,
            guest_query_id: None,
            bot_to_bot: false,
            bot_to_bot_active: Arc::new(AtomicBool::new(false)),
            allowed_bots: Vec::new(),
            max_bot_chain_depth: 1,
            bot_reply_counters: Arc::new(Mutex::new(HashMap::new())),
            supervisor: None,
            guest_proxy_handle: None,
        }
    }

    /// Set the minimum interval between streaming message edits.
    ///
    /// Values below 500 ms are clamped to 500 ms with a warning to prevent
    /// triggering Telegram's rate limits (30 edits/second per chat).
    ///
    /// Default: 3000 ms.
    #[must_use]
    pub fn with_stream_interval(mut self, interval: Duration) -> Self {
        const MIN_INTERVAL: Duration = Duration::from_millis(500);
        if interval < MIN_INTERVAL {
            tracing::warn!(
                requested_ms = interval.as_millis(),
                clamped_ms = MIN_INTERVAL.as_millis(),
                "stream_interval_ms is below the minimum safe value; clamping to 500ms to avoid Telegram rate limits"
            );
            self.stream_interval = MIN_INTERVAL;
        } else {
            self.stream_interval = interval;
        }
        self
    }

    /// Access the raw Bot API 10.0 client for methods not covered by teloxide.
    #[must_use]
    pub fn api_ext(&self) -> &TelegramApiClient {
        &self.api_ext
    }

    /// Attach a [`TaskSupervisor`] so the Telegram listener task is registered
    /// in the workspace-wide task registry with automatic restart on panic.
    ///
    /// The listener is spawned with
    /// `RestartPolicy::Restart { max: 5, base_delay: 2s }`.
    #[must_use]
    pub fn with_supervisor(mut self, supervisor: TaskSupervisor) -> Self {
        self.supervisor = Some(supervisor);
        self
    }

    /// Enable Bot API 10.0 Guest Mode — respond to @mentions via `answerGuestQuery`.
    ///
    /// When enabled, an HTTP proxy is started on loopback that intercepts `getUpdates`
    /// responses to extract `guest_message` updates before teloxide deserializes them.
    #[must_use]
    pub fn with_guest_mode(mut self, enabled: bool) -> Self {
        self.guest_mode = enabled;
        self
    }

    /// Enable bot-to-bot communication (Bot API 10.0).
    ///
    /// * `enabled` — whether to accept messages from other bots.
    /// * `allowed` — bot usernames (with `@` prefix) allowed to interact. Empty = all bots.
    /// * `max_depth` — maximum consecutive bot replies before dropping to prevent loops.
    #[must_use]
    pub fn with_bot_to_bot(mut self, enabled: bool, allowed: Vec<String>, max_depth: u32) -> Self {
        self.bot_to_bot = enabled;
        self.allowed_bots = allowed;
        self.max_bot_chain_depth = max_depth;
        self
    }

    /// Spawn a task that registers bot commands in the Telegram menu.
    fn register_commands(bot: Bot) {
        tokio::spawn(async move {
            let commands = vec![
                BotCommand::new("start", "Start a new conversation"),
                BotCommand::new("reset", "Reset conversation history"),
                BotCommand::new("skills", "List loaded skills"),
                BotCommand::new("agent", "Manage sub-agents (list/spawn/status/cancel)"),
            ];
            if let Err(e) = bot.set_my_commands(commands).await {
                tracing::warn!("failed to register bot commands: {e}");
            }
        });
    }

    /// Spawn the teloxide update listener and return `self` ready for use.
    ///
    /// # Errors
    ///
    /// Returns an error if the bot cannot be initialized.
    #[allow(clippy::too_many_lines)]
    pub fn start(mut self) -> Result<Self, ChannelError> {
        if self.allowed_users.is_empty() {
            tracing::error!("telegram.allowed_users is empty; refusing to start an open bot");
            return Err(ChannelError::Other(
                "telegram.allowed_users must not be empty".into(),
            ));
        }

        // Bot-to-Bot: register capability with Telegram (non-fatal, fire-and-forget).
        // bot_to_bot_active starts as false; the spawned task sets it to true only on success.
        // If setManagedBotAccessSettings fails, bot messages are silently dropped (spec FR-003).
        if self.bot_to_bot {
            let api_ext = self.api_ext.clone();
            let active_flag = self.bot_to_bot_active.clone();
            tokio::spawn(async move {
                let settings = BotAccessSettings {
                    allow_user_messages: true,
                    allow_bot_messages: true,
                };
                match api_ext.set_managed_bot_access_settings(&settings).await {
                    Ok(_) => {
                        active_flag.store(true, Ordering::Release);
                        tracing::info!(
                            "bot-to-bot communication enabled via setManagedBotAccessSettings"
                        );
                    }
                    Err(e) => {
                        tracing::warn!(
                            "setManagedBotAccessSettings failed: {e}; bot-to-bot disabled"
                        );
                    }
                }
            });
        }

        let (tx, rx) = mpsc::channel::<IncomingMessage>(64);
        self.rx = rx;

        // Guest Mode: spawn a transparent axum proxy on a local ephemeral port.
        // The proxy intercepts `getUpdates` responses to extract `guest_message` entries
        // before teloxide deserializes them (teloxide-core 0.13 discards unknown update kinds).
        // Redirecting `Bot` to the proxy via `set_api_url` avoids a second `getUpdates`
        // connection — no 409 Conflict risk (C3 fix).
        if self.guest_mode {
            let token = self.bot.token().to_owned();
            let (proxy_url, proxy_handle) =
                spawn_guest_proxy(&token, tx.clone(), self.allowed_users.clone())?;
            let proxy_url = reqwest::Url::parse(&proxy_url)
                .map_err(|e| ChannelError::Other(format!("guest proxy URL parse error: {e}")))?;
            self.bot = self.bot.clone().set_api_url(proxy_url);
            self.guest_proxy_handle = Some(proxy_handle);
        }

        let bot = self.bot.clone();
        let allowed = self.allowed_users.clone();
        let bot_to_bot = self.bot_to_bot;
        let bot_to_bot_active = self.bot_to_bot_active.clone();
        let allowed_bots = self.allowed_bots.clone();
        let max_bot_chain_depth = self.max_bot_chain_depth;
        let bot_reply_counters = self.bot_reply_counters.clone();

        Self::register_commands(bot.clone());

        let bot_for_factory = bot.clone();
        let allowed_for_factory = allowed.clone();
        let tx_for_factory = tx.clone();
        let bot_to_bot_active_for_factory = bot_to_bot_active.clone();
        let listener_factory = move || {
            let bot = bot_for_factory.clone();
            let allowed = allowed_for_factory.clone();
            let tx = tx_for_factory.clone();
            let bot_to_bot = bot_to_bot;
            let bot_to_bot_active = bot_to_bot_active_for_factory.clone();
            let allowed_bots = allowed_bots.clone();
            let max_bot_chain_depth = max_bot_chain_depth;
            let bot_reply_counters = bot_reply_counters.clone();
            async move {
                let handler = Update::filter_message().endpoint(move |msg: Message, bot: Bot| {
                    let tx = tx.clone();
                    let allowed = allowed.clone();
                    let allowed_bots = allowed_bots.clone();
                    let bot_reply_counters = bot_reply_counters.clone();
                    let bot_to_bot_active = bot_to_bot_active.clone();
                    async move {
                        handle_telegram_message(
                            bot,
                            msg,
                            tx,
                            allowed,
                            bot_to_bot,
                            bot_to_bot_active,
                            allowed_bots,
                            max_bot_chain_depth,
                            bot_reply_counters,
                        )
                        .await
                    }
                });

                Dispatcher::builder(bot, handler)
                    .enable_ctrlc_handler()
                    .build()
                    .dispatch()
                    .await;
            }
        };

        if let Some(sup) = &self.supervisor {
            sup.spawn(zeph_common::TaskDescriptor {
                name: "telegram_listener",
                restart: zeph_common::RestartPolicy::Restart {
                    max: 5,
                    base_delay: Duration::from_secs(2),
                },
                factory: listener_factory,
            });
        } else {
            tokio::spawn(listener_factory());
        }

        tracing::info!("telegram bot listener started");
        Ok(self)
    }

    /// Creates a `TelegramChannel` with an injectable sender for unit tests.
    ///
    /// The returned `Sender` allows injecting `IncomingMessage` values directly
    /// without a real Telegram bot token or live API access. The bot is
    /// initialized with a dummy token and `chat_id` is left unset; tests that
    /// exercise paths which call the bot API (e.g. `send()`, `confirm()`) must
    /// either avoid those code paths or configure a mock HTTP server via
    /// `Bot::set_api_url`.
    #[cfg(test)]
    fn new_test(allowed_users: Vec<String>) -> (Self, mpsc::Sender<IncomingMessage>) {
        let (tx, rx) = mpsc::channel(64);
        let channel = Self {
            bot: Bot::new("test_token"),
            chat_id: None,
            rx,
            allowed_users,
            accumulated: String::new(),
            last_edit: None,
            message_id: None,
            stream_interval: Duration::from_secs(3),
            api_ext: TelegramApiClient::new("test_token"),
            guest_mode: false,
            guest_query_id: None,
            bot_to_bot: false,
            bot_to_bot_active: Arc::new(AtomicBool::new(false)),
            allowed_bots: Vec::new(),
            max_bot_chain_depth: 1,
            bot_reply_counters: Arc::new(Mutex::new(HashMap::new())),
            supervisor: None,
            guest_proxy_handle: None,
        };
        (channel, tx)
    }

    fn is_command(text: &str) -> Option<&str> {
        let cmd = text.split_whitespace().next()?;
        if cmd.starts_with('/') {
            Some(cmd)
        } else {
            None
        }
    }

    fn should_send_update(&self) -> bool {
        match self.last_edit {
            None => true,
            Some(last) => last.elapsed() > self.stream_interval,
        }
    }

    async fn send_or_edit(&mut self) -> Result<(), ChannelError> {
        let Some(chat_id) = self.chat_id else {
            return Err(ChannelError::NoActiveSession);
        };

        let text = if self.accumulated.is_empty() {
            "..."
        } else {
            &self.accumulated
        };

        let formatted_text = markdown_to_telegram(text);

        if formatted_text.is_empty() {
            tracing::debug!("skipping send: formatted text is empty");
            return Ok(());
        }

        tracing::debug!("formatted_text (full): {}", formatted_text);

        match self.message_id {
            None => {
                tracing::debug!("sending new message (length: {})", formatted_text.len());
                let chunks = crate::markdown::utf8_chunks(&formatted_text, MAX_MESSAGE_LEN);
                for chunk in chunks {
                    let msg = self
                        .bot
                        .send_message(chat_id, chunk)
                        .parse_mode(ParseMode::MarkdownV2)
                        .await
                        .map_err(ChannelError::telegram)?;
                    self.message_id = Some(msg.id);
                    tracing::debug!("new message sent with id: {:?}", msg.id);
                }
            }
            Some(msg_id) => {
                tracing::debug!(
                    "editing message {:?} (length: {})",
                    msg_id,
                    formatted_text.len()
                );
                if formatted_text.len() <= MAX_MESSAGE_LEN {
                    let edit_result = self
                        .bot
                        .edit_message_text(chat_id, msg_id, &formatted_text)
                        .parse_mode(ParseMode::MarkdownV2)
                        .await;

                    if let Err(e) = edit_result {
                        let error_msg = e.to_string();

                        if error_msg.contains("message is not modified") {
                            tracing::debug!("message content unchanged, skipping edit");
                        } else if error_msg.contains("message to edit not found")
                            || error_msg.contains("MESSAGE_ID_INVALID")
                        {
                            tracing::warn!(
                                "Telegram edit failed (message_id stale?): {e}, sending new message"
                            );
                            self.message_id = None;
                            self.last_edit = None;

                            let msg = self
                                .bot
                                .send_message(chat_id, &formatted_text)
                                .parse_mode(ParseMode::MarkdownV2)
                                .await
                                .map_err(ChannelError::telegram)?;
                            self.message_id = Some(msg.id);
                        } else {
                            return Err(ChannelError::telegram(e));
                        }
                    } else {
                        tracing::debug!("message edited successfully");
                    }
                } else {
                    // Accumulated text exceeds limit: edit first chunk into existing message,
                    // send remaining chunks as new messages.
                    let chunks = crate::markdown::utf8_chunks(&formatted_text, MAX_MESSAGE_LEN);
                    let mut iter = chunks.into_iter();
                    if let Some(first) = iter.next() {
                        let edit_result = self
                            .bot
                            .edit_message_text(chat_id, msg_id, first)
                            .parse_mode(ParseMode::MarkdownV2)
                            .await;
                        if let Err(e) = edit_result {
                            let error_msg = e.to_string();
                            if !error_msg.contains("message is not modified") {
                                tracing::warn!("Telegram edit failed during split: {e}");
                            }
                        }
                    }
                    for chunk in iter {
                        let msg = self
                            .bot
                            .send_message(chat_id, chunk)
                            .parse_mode(ParseMode::MarkdownV2)
                            .await
                            .map_err(ChannelError::telegram)?;
                        self.message_id = Some(msg.id);
                        tracing::debug!("overflow chunk sent with id: {:?}", msg.id);
                    }
                }
            }
        }

        self.last_edit = Some(Instant::now());
        Ok(())
    }
}

/// Returns `true` when `username` appears in `allowed`.
///
/// An empty `allowed` list is treated as "no restriction" and always returns
/// `true`, but `start()` rejects empty lists before the listener is spawned,
/// so in practice `allowed` is never empty at call time.
fn is_user_authorized(username: Option<&str>, allowed: &[String]) -> bool {
    allowed.is_empty() || username.is_some_and(|u| allowed.iter().any(|a| a == u))
}

/// Returns `true` when `username` is in `allowed_bots` or the list is empty.
fn is_bot_authorized(username: Option<&str>, allowed_bots: &[String]) -> bool {
    allowed_bots.is_empty()
        || username.is_some_and(|u| {
            allowed_bots.iter().any(|a| {
                // Compare with or without "@" prefix
                a.trim_start_matches('@') == u.trim_start_matches('@')
            })
        })
}

/// Increments the consecutive bot-reply counter for `chat_id` and returns the new depth.
///
/// Evicts one arbitrary entry when at capacity to bound memory usage.
fn increment_bot_depth(counters: &BotReplyCounters, chat_id: ChatId) -> u32 {
    let mut map = counters.lock().expect("bot_reply_counters poisoned");
    if map.len() >= MAX_TRACKED_CHATS && !map.contains_key(&chat_id) {
        // Evict one arbitrary entry to bound memory. HashMap has no insertion order,
        // so we evict the first key returned by the iterator (effectively random).
        // This prevents a flood of unique chat IDs from resetting all legitimate counters.
        if let Some(&evict_key) = map.keys().next() {
            map.remove(&evict_key);
        }
    }
    let counter = map.entry(chat_id).or_insert(0);
    *counter += 1;
    *counter
}

/// Resets the bot-reply counter for `chat_id` on a non-bot message.
fn reset_bot_depth(counters: &BotReplyCounters, chat_id: ChatId) {
    counters
        .lock()
        .expect("bot_reply_counters poisoned")
        .remove(&chat_id);
}

/// Shared state for the guest-mode axum proxy.
#[derive(Clone)]
struct GuestProxyState {
    upstream: String,
    client: reqwest::Client,
    tx: mpsc::Sender<IncomingMessage>,
    allowed_users: Vec<String>,
}

/// Guest Mode transparent HTTP proxy.
///
/// Binds an ephemeral TCP port on `127.0.0.1`, proxies all requests from the
/// teloxide `Bot` to `https://api.telegram.org`. For `getUpdates` requests the
/// proxy injects `"guest_message"` into `allowed_updates` so Telegram delivers
/// these updates, extracts copies of `guest_message` entries from the response
/// to forward them to the agent, and returns the **original unmodified** response
/// to teloxide so its internal offset advances correctly — preventing duplicates.
/// This guarantees a single `getUpdates` connection per bot token — no 409 Conflict.
///
/// Returns the local URL (`http://127.0.0.1:<port>`) to pass to
/// `Bot::set_api_url()`, and a [`tokio::task::JoinHandle`] for the proxy task.
fn spawn_guest_proxy(
    token: &str,
    tx: mpsc::Sender<IncomingMessage>,
    allowed_users: Vec<String>,
) -> Result<(String, tokio::task::JoinHandle<()>), ChannelError> {
    let upstream = format!("https://api.telegram.org/bot{token}");
    let client = reqwest::Client::builder()
        .timeout(Duration::from_mins(1))
        .build()
        .map_err(|e| ChannelError::Other(format!("guest proxy client init failed: {e}")))?;

    let state = GuestProxyState {
        upstream,
        client,
        tx,
        allowed_users,
    };

    let listener = std::net::TcpListener::bind("127.0.0.1:0")
        .map_err(|e| ChannelError::Other(format!("guest proxy bind failed: {e}")))?;
    let local_addr = listener
        .local_addr()
        .map_err(|e| ChannelError::Other(format!("guest proxy addr failed: {e}")))?;
    let local_url = format!("http://127.0.0.1:{}/bot{token}", local_addr.port());

    let app = Router::new()
        .route("/{*path}", any(proxy_handler))
        .with_state(state);

    let handle = tokio::spawn(async move {
        let listener = tokio::net::TcpListener::from_std(listener)
            .expect("guest proxy: TcpListener conversion failed");
        if let Err(e) = axum::serve(listener, app).await {
            tracing::warn!("guest proxy axum serve error: {e}");
        }
    });

    Ok((local_url, handle))
}

/// Axum handler: proxies one request upstream and, on `getUpdates`, extracts
/// `guest_message` entries from the Telegram API response.
///
/// For `getUpdates` requests the handler:
/// 1. Injects `"guest_message"` into the outgoing `allowed_updates` array so
///    Telegram actually delivers these updates (teloxide omits them by default).
/// 2. Extracts a copy of any `guest_message` entries from the response and
///    forwards them to the agent via `tx`.
/// 3. Returns the **full** unmodified result array to teloxide — including the
///    original `guest_message` objects — so teloxide's internal offset advances
///    correctly and the same updates are never replayed.  Teloxide's
///    `Update::filter_message()` handler simply ignores unknown `UpdateKind`
///    variants, so the extra entries are silently dropped by the dispatcher.
async fn proxy_handler(State(state): State<GuestProxyState>, req: Request<Body>) -> Response {
    // Path structure from teloxide: /bot<TOKEN>/<method>
    // proxy_url passed to Bot::set_api_url is http://127.0.0.1:<port>/bot<TOKEN>
    // teloxide appends /<method> to base_url, so path is /bot<TOKEN>/<method>
    let path = req.uri().path();
    let query = req
        .uri()
        .query()
        .map(|q| format!("?{q}"))
        .unwrap_or_default();
    let method_part = path.splitn(4, '/').nth(3).unwrap_or("");
    let upstream_url = format!("{}/{method_part}{query}", state.upstream);

    let is_get_updates = method_part == "getUpdates";

    let method = req.method().clone();
    let headers = req.headers().clone();
    let mut body_bytes = match axum::body::to_bytes(req.into_body(), 4 * 1024 * 1024).await {
        Ok(b) => b,
        Err(e) => {
            tracing::warn!("guest proxy: failed to read request body: {e}");
            return StatusCode::BAD_REQUEST.into_response();
        }
    };

    // Inject "guest_message" into allowed_updates so Telegram delivers these updates.
    // teloxide only requests ["message"] by default; without this injection
    // Telegram never sends guest_message entries and Guest Mode is silently broken.
    if is_get_updates
        && let Ok(mut body_json) = serde_json::from_slice::<serde_json::Value>(&body_bytes)
    {
        let injected = serde_json::json!("guest_message");
        match body_json
            .get_mut("allowed_updates")
            .and_then(|v| v.as_array_mut())
        {
            Some(arr) => {
                if !arr.iter().any(|v| v.as_str() == Some("guest_message")) {
                    arr.push(injected);
                }
            }
            None => {
                body_json["allowed_updates"] = serde_json::json!(["message", "guest_message"]);
            }
        }
        if let Ok(b) = serde_json::to_vec(&body_json) {
            body_bytes = b.into();
        }
    }

    let mut upstream_req = state.client.request(method, &upstream_url);
    for (name, value) in &headers {
        upstream_req = upstream_req.header(name, value);
    }
    upstream_req = upstream_req.body(body_bytes);

    let upstream_resp = match upstream_req.send().await {
        Ok(r) => r,
        Err(e) => {
            tracing::warn!("guest proxy upstream error: {}", e.without_url());
            return StatusCode::BAD_GATEWAY.into_response();
        }
    };

    let status = upstream_resp.status();
    let resp_headers = upstream_resp.headers().clone();
    let resp_bytes = match upstream_resp.bytes().await {
        Ok(b) => b,
        Err(e) => {
            tracing::warn!(
                "guest proxy: failed to read upstream response: {}",
                e.without_url()
            );
            return StatusCode::BAD_GATEWAY.into_response();
        }
    };

    // On getUpdates success: extract guest_message entries and forward copies to the
    // agent, then return the original response unchanged to teloxide so its offset
    // advances correctly (no duplicate delivery on the next poll).
    if is_get_updates
        && status.is_success()
        && let Ok(json) = serde_json::from_slice::<serde_json::Value>(&resp_bytes)
        && let Some(result) = json.get("result").and_then(|r| r.as_array())
    {
        for update in result {
            if let Some(gm_val) = update.get("guest_message") {
                extract_and_forward_guest_message(gm_val, &state.tx, &state.allowed_users).await;
            }
        }
    }

    let mut response = Response::new(Body::from(resp_bytes));
    *response.status_mut() = status;
    for (name, value) in &resp_headers {
        response.headers_mut().insert(name, value.clone());
    }
    response
}

/// Parse a raw `guest_message` JSON value and forward it to the agent.
async fn extract_and_forward_guest_message(
    gm_val: &serde_json::Value,
    tx: &mpsc::Sender<IncomingMessage>,
    allowed_users: &[String],
) {
    let gm: GuestMessage = match serde_json::from_value(gm_val.clone()) {
        Ok(m) => m,
        Err(e) => {
            tracing::warn!("failed to parse guest_message: {e}");
            return;
        }
    };

    let caller_username = gm.guest_bot_caller_user.username.as_deref();
    if !is_user_authorized(caller_username, allowed_users) {
        tracing::warn!(
            username = ?caller_username,
            "guest_message from unauthorized user — dropped"
        );
        return;
    }

    let text = gm.text.unwrap_or_default();
    let chat_id = ChatId(gm.guest_bot_caller_chat.id);
    let _ = tx
        .send(IncomingMessage {
            chat_id,
            text,
            attachments: Vec::new(),
            guest_query_id: Some(gm.guest_query_id),
            is_from_bot: gm.guest_bot_caller_user.is_bot,
        })
        .await;
}

/// Extract the audio file identifier and byte size from a message, if present.
///
/// Prefers a voice note over an audio file when both are present.
fn extract_audio_attachment(msg: &Message) -> Option<(String, u32)> {
    msg.voice()
        .map(|v| (v.file.id.0.clone(), v.file.size))
        .or_else(|| msg.audio().map(|a| (a.file.id.0.clone(), a.file.size)))
}

/// Extract the largest photo's file identifier and byte size from a message.
///
/// Returns `None` when the message contains no photo or the largest available
/// size exceeds [`MAX_IMAGE_BYTES`].
fn extract_photo_attachment(msg: &Message) -> Option<(String, u32)> {
    let photos = msg.photo()?;
    let photo = photos.iter().max_by_key(|p| p.file.size)?;
    if photo.file.size > MAX_IMAGE_BYTES {
        tracing::warn!(
            size = photo.file.size,
            max = MAX_IMAGE_BYTES,
            "photo exceeds size limit, skipping"
        );
        return None;
    }
    Some((photo.file.id.0.clone(), photo.file.size))
}

/// Walk `reply_to_message` links to compute structural chain depth (spec FR-007).
///
/// Returns the number of `reply_to_message` hops from `msg` up to and including
/// the root, capped at `cap + 1` to bound recursion.  The Telegram API only
/// exposes one level of nesting in the payload, so in practice this returns 0
/// (no reply) or 1 (direct reply); the cap is a safeguard against future changes.
fn compute_chain_depth(msg: &Message, cap: u32) -> u32 {
    let mut depth = 0u32;
    let mut current = msg.reply_to_message();
    while current.is_some() && depth < cap + 1 {
        depth += 1;
        current = current.and_then(|m| m.reply_to_message());
    }
    depth
}

/// Process one incoming Telegram update inside the dispatcher endpoint.
///
/// Checks authorization, applies bot-to-bot policy (structural chain depth via
/// [`compute_chain_depth`] + consecutive-reply counter), downloads any media
/// attachments, and forwards the assembled [`IncomingMessage`] to the agent
/// via `tx`.  Returns `respond(())` in all branches so teloxide considers the
/// update handled.
#[allow(clippy::too_many_arguments)]
async fn handle_telegram_message(
    bot: Bot,
    msg: Message,
    tx: mpsc::Sender<IncomingMessage>,
    allowed: Vec<String>,
    bot_to_bot: bool,
    bot_to_bot_active: Arc<AtomicBool>,
    allowed_bots: Vec<String>,
    max_bot_chain_depth: u32,
    bot_reply_counters: BotReplyCounters,
) -> Result<(), teloxide::RequestError> {
    let sender_is_bot = msg.from.as_ref().is_some_and(|u| u.is_bot);

    if sender_is_bot {
        if !bot_to_bot || !bot_to_bot_active.load(Ordering::Acquire) {
            // Feature disabled — silently drop without logging
            return respond(());
        }
        let sender_username = msg.from.as_ref().and_then(|u| u.username.as_deref());
        if !is_bot_authorized(sender_username, &allowed_bots) {
            tracing::warn!(sender = ?sender_username, "rejected message from unauthorized bot");
            return respond(());
        }
        // Primary check: structural reply chain depth (spec FR-007).
        // Telegram API only exposes 1 level, so this catches direct reply loops.
        let chain_depth = compute_chain_depth(&msg, max_bot_chain_depth);
        if chain_depth >= max_bot_chain_depth {
            tracing::warn!(
                chain_depth,
                max = max_bot_chain_depth,
                message_id = msg.id.0,
                sender = ?sender_username,
                "dropping bot message: reply chain depth limit reached"
            );
            return respond(());
        }
        // Secondary check: consecutive bot replies in this chat (defense-in-depth).
        let consec_depth = increment_bot_depth(&bot_reply_counters, msg.chat.id);
        if consec_depth >= max_bot_chain_depth {
            tracing::warn!(
                consec_depth,
                max = max_bot_chain_depth,
                message_id = msg.id.0,
                sender = ?sender_username,
                "dropping bot message: consecutive reply limit reached"
            );
            return respond(());
        }
    } else {
        // Non-bot message: reset consecutive bot reply counter for this chat.
        reset_bot_depth(&bot_reply_counters, msg.chat.id);

        let username = msg.from.as_ref().and_then(|u| u.username.as_deref());
        if !is_user_authorized(username, &allowed) {
            tracing::warn!("rejected message from unauthorized user: {:?}", username);
            return respond(());
        }
    }

    let text = msg.text().unwrap_or_default().to_string();
    let mut attachments = Vec::new();

    if let Some((file_id, file_size)) = extract_audio_attachment(&msg) {
        match download_file(&bot, file_id, file_size).await {
            Ok(data) => {
                attachments.push(Attachment {
                    kind: AttachmentKind::Audio,
                    data,
                    filename: msg.audio().and_then(|a| a.file_name.clone()),
                });
            }
            Err(e) => {
                tracing::warn!("failed to download audio attachment: {e}");
            }
        }
    }

    if let Some((file_id, file_size)) = extract_photo_attachment(&msg) {
        match download_file(&bot, file_id, file_size).await {
            Ok(data) => {
                attachments.push(Attachment {
                    kind: AttachmentKind::Image,
                    data,
                    filename: None,
                });
            }
            Err(e) => {
                tracing::warn!("failed to download photo attachment: {e}");
            }
        }
    }

    if text.is_empty() && attachments.is_empty() {
        return respond(());
    }

    let _ = tx
        .send(IncomingMessage {
            chat_id: msg.chat.id,
            text,
            attachments,
            guest_query_id: None,
            is_from_bot: sender_is_bot,
        })
        .await;

    respond(())
}

async fn download_file(bot: &Bot, file_id: String, capacity: u32) -> Result<Vec<u8>, String> {
    use teloxide::net::Download;

    let file = bot
        .get_file(file_id.into())
        .await
        .map_err(|e| format!("get_file: {e}"))?;
    let mut buf: Vec<u8> = Vec::with_capacity(capacity as usize);
    bot.download_file(&file.path, &mut buf)
        .await
        .map_err(|e| format!("download_file: {e}"))?;
    Ok(buf)
}

impl Channel for TelegramChannel {
    /// Returns `false` — Telegram is a persistent remote channel with no
    /// meaningful concept of "session exit".
    fn supports_exit(&self) -> bool {
        false
    }

    /// Non-blocking receive: returns a buffered message if one is available.
    ///
    /// Updates `chat_id` when a message is returned so subsequent [`send`]
    /// calls know the destination.
    ///
    /// [`send`]: TelegramChannel::send
    fn try_recv(&mut self) -> Option<ChannelMessage> {
        self.rx.try_recv().ok().map(|incoming| {
            self.chat_id = Some(incoming.chat_id);
            let is_guest = incoming.guest_query_id.is_some();
            self.guest_query_id.clone_from(&incoming.guest_query_id);
            ChannelMessage {
                text: incoming.text,
                attachments: incoming.attachments,
                is_guest_context: is_guest,
                is_from_bot: incoming.is_from_bot,
            }
        })
    }

    /// Await the next user message from Telegram.
    ///
    /// The method loops internally to handle built-in commands:
    /// * `/start` — sends the welcome message and loops.
    /// * `/reset` and `/skills` — returned to the caller as regular messages.
    /// * Any other command or plain text — returned immediately.
    ///
    /// Resets the streaming state (`accumulated`, `last_edit`, `message_id`)
    /// so each user turn starts with a clean response buffer.
    ///
    /// Returns `Ok(None)` when the internal channel is closed (i.e. the
    /// dispatcher task has exited).
    ///
    /// # Errors
    ///
    /// Returns `Err` if sending the welcome reply for `/start` fails.
    #[cfg_attr(
        feature = "profiling",
        tracing::instrument(name = "channel.telegram.recv", skip_all, fields(msg_len = tracing::field::Empty))
    )]
    async fn recv(&mut self) -> Result<Option<ChannelMessage>, ChannelError> {
        loop {
            let Some(incoming) = self.rx.recv().await else {
                return Ok(None);
            };

            self.chat_id = Some(incoming.chat_id);

            // Reset streaming state for new response
            self.accumulated.clear();
            self.last_edit = None;
            self.message_id = None;

            let is_guest = incoming.guest_query_id.is_some();
            self.guest_query_id.clone_from(&incoming.guest_query_id);

            if let Some(cmd) = Self::is_command(&incoming.text) {
                match cmd {
                    "/start" => {
                        self.send("Welcome to Zeph! Send me a message to get started.")
                            .await?;
                        continue;
                    }
                    "/reset" => {
                        return Ok(Some(ChannelMessage {
                            text: "/reset".to_string(),
                            attachments: vec![],
                            is_guest_context: false,
                            is_from_bot: false,
                        }));
                    }
                    "/skills" => {
                        return Ok(Some(ChannelMessage {
                            text: "/skills".to_string(),
                            attachments: vec![],
                            is_guest_context: false,
                            is_from_bot: false,
                        }));
                    }
                    _ => {}
                }
            }

            return Ok(Some(ChannelMessage {
                text: incoming.text,
                attachments: incoming.attachments,
                is_guest_context: is_guest,
                is_from_bot: incoming.is_from_bot,
            }));
        }
    }

    /// Send a complete message to the active Telegram chat.
    ///
    /// The text is converted to `MarkdownV2` via [`markdown_to_telegram`]
    /// before sending.  Messages longer than 4096 bytes are split into
    /// multiple messages at UTF-8 / newline boundaries.
    ///
    /// # Errors
    ///
    /// Returns `Err(ChannelError::NoActiveSession)` if no active chat has been
    /// established yet (i.e. `recv` has never returned a message), or
    /// `Err(ChannelError::Telegram)` if the Telegram API call fails.
    #[cfg_attr(
        feature = "profiling",
        tracing::instrument(name = "channel.telegram.send", skip_all, fields(msg_len = %text.len()))
    )]
    async fn send(&mut self, text: &str) -> Result<(), ChannelError> {
        // Guest context: accumulate full response — flush_chunks calls answerGuestQuery once.
        if self.guest_query_id.is_some() {
            if !self.accumulated.is_empty() {
                self.accumulated.push('\n');
            }
            self.accumulated.push_str(text);
            return Ok(());
        }

        let Some(chat_id) = self.chat_id else {
            return Err(ChannelError::NoActiveSession);
        };

        let formatted_text = markdown_to_telegram(text);

        if formatted_text.is_empty() {
            tracing::debug!("skipping send: formatted text is empty");
            return Ok(());
        }

        if formatted_text.len() <= MAX_MESSAGE_LEN {
            self.bot
                .send_message(chat_id, &formatted_text)
                .parse_mode(ParseMode::MarkdownV2)
                .await
                .map_err(ChannelError::telegram)?;
        } else {
            let chunks = crate::markdown::utf8_chunks(&formatted_text, MAX_MESSAGE_LEN);
            for chunk in chunks {
                self.bot
                    .send_message(chat_id, chunk)
                    .parse_mode(ParseMode::MarkdownV2)
                    .await
                    .map_err(ChannelError::telegram)?;
            }
        }

        Ok(())
    }

    /// Append a streaming chunk to the response buffer.
    ///
    /// The chunk is accumulated in memory.  A Telegram edit is issued only
    /// when at least 3 seconds have elapsed since the last edit, which keeps
    /// the bot within Telegram's rate limits.
    ///
    /// Call [`flush_chunks`] after the stream ends to perform the final edit
    /// with the complete response text.
    ///
    /// # Errors
    ///
    /// Returns `Err` if the periodic Telegram edit fails.
    ///
    /// [`flush_chunks`]: TelegramChannel::flush_chunks
    async fn send_chunk(&mut self, chunk: &str) -> Result<(), ChannelError> {
        self.accumulated.push_str(chunk);
        tracing::debug!(
            "received chunk (size: {}, total: {})",
            chunk.len(),
            self.accumulated.len()
        );

        // In guest context, accumulate only — never call send_or_edit (NFR-005).
        // The full accumulated text is sent via answerGuestQuery in flush_chunks.
        if self.guest_query_id.is_some() {
            return Ok(());
        }

        if self.should_send_update() {
            tracing::debug!("sending update (should_send_update returned true)");
            self.send_or_edit().await?;
        }

        Ok(())
    }

    /// Finalise the streamed response with one last Telegram edit.
    ///
    /// Performs a final edit when a message has already been created
    /// (i.e. `message_id` is `Some`), then clears the accumulation
    /// buffer and resets streaming state so the channel is ready for the next
    /// agent turn.
    ///
    /// # Errors
    ///
    /// Returns `Err` if the final Telegram edit fails.
    async fn flush_chunks(&mut self) -> Result<(), ChannelError> {
        tracing::debug!(
            "flushing chunks (message_id: {:?}, accumulated: {} bytes)",
            self.message_id,
            self.accumulated.len()
        );

        // Guest context: send accumulated text via answerGuestQuery (single call).
        if let Some(query_id) = self.guest_query_id.take() {
            let full_text = std::mem::take(&mut self.accumulated);
            let text = full_text.trim();
            if text.len() > MAX_MESSAGE_LEN {
                tracing::warn!(
                    query_id,
                    bytes = text.len(),
                    max = MAX_MESSAGE_LEN,
                    "guest response exceeds 4096 bytes; Telegram truncates answerGuestQuery — consider shorter responses"
                );
            }
            if !text.is_empty()
                && let Err(e) = self
                    .api_ext
                    .answer_guest_query(&query_id, text, Some("HTML"))
                    .await
            {
                tracing::warn!(query_id, "answer_guest_query failed: {e}");
            }
            self.last_edit = None;
            self.message_id = None;
            return Ok(());
        }

        // Send if there is unsent accumulated text OR an existing message to finalize.
        // The `message_id.is_some()` guard alone would silently discard text that arrived
        // before the first interval elapsed (so `send_or_edit` was never called).
        if self.message_id.is_some() || !self.accumulated.is_empty() {
            self.send_or_edit().await?;
        }

        // Clear state for next response
        self.accumulated.clear();
        self.last_edit = None;
        self.message_id = None;

        Ok(())
    }

    /// Send a `typing…` chat action to Telegram.
    ///
    /// Silently succeeds (returns `Ok(())`) when no active chat exists yet so
    /// that the agent loop can call this unconditionally without checking state.
    ///
    /// # Errors
    ///
    /// Returns `Err` if the Telegram API call fails.
    async fn send_typing(&mut self) -> Result<(), ChannelError> {
        let Some(chat_id) = self.chat_id else {
            return Ok(());
        };
        self.bot
            .send_chat_action(chat_id, ChatAction::Typing)
            .await
            .map_err(ChannelError::telegram)?;
        Ok(())
    }

    /// Send a yes/no confirmation prompt to Telegram and await the reply.
    ///
    /// Sends `prompt` followed by instructions to reply `yes`.  Waits up to
    /// [`CONFIRM_TIMEOUT`] (30 s) for a response.  Returns `true` only when
    /// the user replies with the string `yes` (case-insensitive).
    ///
    /// Returns `Ok(false)` on timeout or channel close, never `Err` for those
    /// conditions.
    ///
    /// # Errors
    ///
    /// Returns `Err` if sending the prompt message fails.
    ///
    /// [`CONFIRM_TIMEOUT`]: crate::CONFIRM_TIMEOUT
    async fn confirm(&mut self, prompt: &str) -> Result<bool, ChannelError> {
        self.send(&format!(
            "{prompt}\nReply 'yes' to confirm (timeout: {}s).",
            crate::CONFIRM_TIMEOUT.as_secs()
        ))
        .await?;
        match tokio::time::timeout(crate::CONFIRM_TIMEOUT, self.rx.recv()).await {
            Ok(Some(incoming)) => Ok(incoming.text.trim().eq_ignore_ascii_case("yes")),
            Ok(None) => {
                tracing::warn!("confirm channel closed — denying secret request");
                Ok(false)
            }
            Err(_) => {
                tracing::warn!("confirm timed out after 30s — denied");
                Ok(false)
            }
        }
    }

    /// Collect structured input from the user on behalf of an MCP server.
    ///
    /// Sends an introductory message identifying the requesting MCP server,
    /// then prompts for each field sequentially via individual Telegram
    /// messages.  Each field has [`ELICITATION_TIMEOUT`] (120 s) to respond.
    ///
    /// The user can reply `/cancel` at any field prompt to abort the flow,
    /// which returns [`ElicitationResponse::Cancelled`].  Invalid values for a
    /// field return [`ElicitationResponse::Declined`] immediately.
    ///
    /// Enum fields use 1-based numeric selection to stay within Telegram's
    /// 64-byte callback-data limit.
    ///
    /// # Errors
    ///
    /// Returns `Err` if sending any of the prompt messages fails.
    ///
    /// [`ELICITATION_TIMEOUT`]: crate::ELICITATION_TIMEOUT
    /// [`ElicitationResponse::Cancelled`]: zeph_core::channel::ElicitationResponse::Cancelled
    /// [`ElicitationResponse::Declined`]: zeph_core::channel::ElicitationResponse::Declined
    async fn elicit(
        &mut self,
        request: ElicitationRequest,
    ) -> Result<ElicitationResponse, ChannelError> {
        let timeout = crate::ELICITATION_TIMEOUT;

        self.send(&format!(
            "*[MCP server '{}' is requesting input]*\n{}\n\n_Reply /cancel to cancel. \
             Timeout: {}s._",
            sanitize_markdown(&request.server_name),
            sanitize_markdown(&request.message),
            timeout.as_secs(),
        ))
        .await?;

        let mut values = serde_json::Map::new();
        for field in &request.fields {
            let prompt = build_telegram_field_prompt(field);
            self.send(&prompt).await?;

            let incoming = match tokio::time::timeout(timeout, self.rx.recv()).await {
                Ok(Some(msg)) => msg,
                Ok(None) => {
                    tracing::warn!(server = request.server_name, "elicitation channel closed");
                    return Ok(ElicitationResponse::Declined);
                }
                Err(_) => {
                    tracing::warn!(server = request.server_name, "elicitation timed out");
                    let _ = self
                        .send("Elicitation timed out — request cancelled.")
                        .await;
                    return Ok(ElicitationResponse::Cancelled);
                }
            };

            let text = incoming.text.trim().to_owned();

            if text.eq_ignore_ascii_case("/cancel") {
                let _ = self.send("Elicitation cancelled.").await;
                return Ok(ElicitationResponse::Cancelled);
            }

            let Some(value) = coerce_telegram_field(&text, &field.field_type) else {
                let _ = self
                    .send(&format!(
                        "Invalid value for '{}'. Declining.",
                        sanitize_markdown(&field.name)
                    ))
                    .await;
                return Ok(ElicitationResponse::Declined);
            };
            values.insert(field.name.clone(), value);
        }

        Ok(ElicitationResponse::Accepted(serde_json::Value::Object(
            values,
        )))
    }
}

/// Strip Markdown special characters to prevent format injection in Telegram messages.
///
/// Removes `*`, `_`, `[`, `]`, `` ` ``, and ANSI escape sequences (`\x1b`).
/// Used for untrusted strings coming from MCP server metadata (server names,
/// field descriptions) that are embedded in bot messages.
fn sanitize_markdown(s: &str) -> String {
    s.chars()
        .filter(|c| !matches!(c, '*' | '_' | '[' | ']' | '`' | '\x1b'))
        .collect()
}

/// Build a Telegram-formatted prompt string for a single elicitation field.
///
/// The prompt is tailored to the field type:
/// * `Boolean` — asks for `yes` or `no`.
/// * `Enum` — lists options with 1-based numeric indices to avoid Telegram's
///   64-byte callback-data limit.
/// * `Integer` / `Number` — asks for a numeric value.
/// * `String` — asks for free-form text.
fn build_telegram_field_prompt(field: &ElicitationField) -> String {
    let req = if field.required { " (required)" } else { "" };
    let name = sanitize_markdown(&field.name);
    match &field.field_type {
        ElicitationFieldType::Boolean => {
            format!("*{name}*{req}: Reply *yes* or *no*")
        }
        ElicitationFieldType::Enum(opts) => {
            // Use short numeric indexes to avoid Telegram 64-byte callback_data limit
            let list: String = opts
                .iter()
                .enumerate()
                .map(|(i, o)| format!("{}: {}", i + 1, sanitize_markdown(o)))
                .collect::<Vec<_>>()
                .join("\n");
            format!("*{name}*{req}: Reply with the number:\n{list}")
        }
        ElicitationFieldType::Integer => {
            format!("*{name}*{req}: Reply with an integer")
        }
        ElicitationFieldType::Number => {
            format!("*{name}*{req}: Reply with a number")
        }
        ElicitationFieldType::String => {
            format!("*{name}*{req}: Reply with text")
        }
    }
}

/// Coerce a raw Telegram reply to the JSON type required by an elicitation field.
///
/// Returns `None` when the input cannot be converted to the declared type.
///
/// Enum fields accept either a 1-based numeric index (as displayed by
/// [`build_telegram_field_prompt`]) or an exact case-insensitive match of the
/// option string.
fn coerce_telegram_field(text: &str, kind: &ElicitationFieldType) -> Option<serde_json::Value> {
    match kind {
        ElicitationFieldType::String => Some(serde_json::Value::String(text.to_owned())),
        ElicitationFieldType::Boolean => {
            if text.eq_ignore_ascii_case("yes") || text == "1" {
                Some(serde_json::Value::Bool(true))
            } else if text.eq_ignore_ascii_case("no") || text == "0" {
                Some(serde_json::Value::Bool(false))
            } else {
                None
            }
        }
        ElicitationFieldType::Integer => text
            .parse::<i64>()
            .ok()
            .map(|n| serde_json::Value::Number(n.into())),
        ElicitationFieldType::Number => text
            .parse::<f64>()
            .ok()
            .and_then(|n| serde_json::Number::from_f64(n).map(serde_json::Value::Number)),
        ElicitationFieldType::Enum(opts) => {
            // Accept numeric index (1-based) or exact match
            if let Ok(idx) = text.parse::<usize>()
                && idx >= 1
                && idx <= opts.len()
            {
                return Some(serde_json::Value::String(opts[idx - 1].clone()));
            }
            // Exact match (case-insensitive)
            opts.iter()
                .find(|o| o.eq_ignore_ascii_case(text))
                .map(|o| serde_json::Value::String(o.clone()))
        }
    }
}

#[cfg(test)]
mod tests {
    use std::time::Instant;

    use wiremock::matchers::any;
    use wiremock::{Mock, MockServer, ResponseTemplate};

    use super::*;

    // ---------------------------------------------------------------------------
    // Helpers
    // ---------------------------------------------------------------------------

    /// Minimal valid Telegram `sendMessage` / `editMessageText` response.
    fn tg_ok_message() -> serde_json::Value {
        serde_json::json!({
            "ok": true,
            "result": {
                "message_id": 42,
                "date": 1_700_000_000_i64,
                "chat": {"id": 1, "type": "private"}
            }
        })
    }

    /// Creates a `TelegramChannel` whose bot is pointed at `server` so that
    /// any API call the channel makes is intercepted rather than going to the
    /// real Telegram endpoint.
    async fn make_mocked_channel(
        server: &MockServer,
        allowed_users: Vec<String>,
    ) -> (TelegramChannel, mpsc::Sender<IncomingMessage>) {
        Mock::given(any())
            .respond_with(ResponseTemplate::new(200).set_body_json(tg_ok_message()))
            .mount(server)
            .await;

        let api_url = reqwest::Url::parse(&server.uri()).unwrap();
        let bot = Bot::new("test_token").set_api_url(api_url);
        let (tx, rx) = mpsc::channel(64);
        let channel = TelegramChannel {
            bot,
            chat_id: Some(ChatId(1)),
            rx,
            allowed_users,
            accumulated: String::new(),
            last_edit: None,
            message_id: None,
            stream_interval: Duration::from_secs(3),
            api_ext: TelegramApiClient::with_base_url(server.uri()),
            guest_mode: false,
            guest_query_id: None,
            bot_to_bot: false,
            bot_to_bot_active: Arc::new(AtomicBool::new(false)),
            allowed_bots: Vec::new(),
            max_bot_chain_depth: 1,
            bot_reply_counters: Arc::new(Mutex::new(HashMap::new())),
            supervisor: None,
            guest_proxy_handle: None,
        };
        (channel, tx)
    }

    fn plain_message(text: &str) -> IncomingMessage {
        IncomingMessage {
            chat_id: ChatId(1),
            text: text.to_string(),
            attachments: vec![],
            guest_query_id: None,
            is_from_bot: false,
        }
    }

    // ---------------------------------------------------------------------------
    // Pure-function unit tests (no async, no network)
    // ---------------------------------------------------------------------------

    #[test]
    fn is_user_authorized_empty_allowed_permits_all() {
        assert!(is_user_authorized(None, &[]));
        assert!(is_user_authorized(Some("anyone"), &[]));
    }

    #[test]
    fn is_user_authorized_known_user_is_permitted() {
        let allowed = vec!["alice".to_string(), "bob".to_string()];
        assert!(is_user_authorized(Some("alice"), &allowed));
        assert!(is_user_authorized(Some("bob"), &allowed));
    }

    #[test]
    fn is_user_authorized_unknown_user_is_rejected() {
        let allowed = vec!["alice".to_string()];
        assert!(!is_user_authorized(Some("eve"), &allowed));
        assert!(!is_user_authorized(None, &allowed));
    }

    #[test]
    fn is_command_detection() {
        assert_eq!(TelegramChannel::is_command("/start"), Some("/start"));
        assert_eq!(TelegramChannel::is_command("/reset now"), Some("/reset"));
        assert_eq!(TelegramChannel::is_command("hello"), None);
        assert_eq!(TelegramChannel::is_command(""), None);
    }

    #[test]
    fn should_send_update_first_chunk() {
        let channel = TelegramChannel::new("test_token".to_string(), Vec::new());
        assert!(channel.should_send_update());
    }

    #[test]
    fn should_send_update_time_threshold() {
        let mut channel = TelegramChannel::new("test_token".to_string(), Vec::new());
        channel.accumulated = "test".to_string();
        channel.last_edit = Some(Instant::now().checked_sub(Duration::from_secs(4)).unwrap());
        assert!(channel.should_send_update());
    }

    #[test]
    fn should_not_send_update_within_threshold() {
        let mut channel = TelegramChannel::new("test_token".to_string(), Vec::new());
        channel.last_edit = Some(
            Instant::now()
                .checked_sub(Duration::from_millis(500))
                .unwrap(),
        );
        assert!(!channel.should_send_update());
    }

    #[test]
    fn with_stream_interval_custom_interval_respected() {
        let mut channel = TelegramChannel::new("test_token".to_string(), Vec::new())
            .with_stream_interval(Duration::from_secs(2));
        // 1500ms elapsed < 2000ms interval — should NOT send
        channel.last_edit = Some(
            Instant::now()
                .checked_sub(Duration::from_millis(1500))
                .unwrap(),
        );
        assert!(!channel.should_send_update());
        // 2500ms elapsed > 2000ms interval — should send
        channel.last_edit = Some(
            Instant::now()
                .checked_sub(Duration::from_millis(2500))
                .unwrap(),
        );
        assert!(channel.should_send_update());
    }

    #[test]
    fn with_stream_interval_clamps_below_500ms() {
        let channel = TelegramChannel::new("test_token".to_string(), Vec::new())
            .with_stream_interval(Duration::from_millis(100));
        assert_eq!(channel.stream_interval, Duration::from_millis(500));
    }

    #[test]
    fn with_stream_interval_default_is_3s() {
        let channel = TelegramChannel::new("test_token", Vec::new());
        assert_eq!(channel.stream_interval, Duration::from_secs(3));
    }

    #[test]
    fn max_image_bytes_is_20_mib() {
        assert_eq!(MAX_IMAGE_BYTES, 20 * 1024 * 1024);
    }

    #[test]
    fn photo_size_limit_enforcement() {
        const { assert!(MAX_IMAGE_BYTES - 1 <= MAX_IMAGE_BYTES) };
        const { assert!(MAX_IMAGE_BYTES <= MAX_IMAGE_BYTES) };
        const { assert!(MAX_IMAGE_BYTES + 1 > MAX_IMAGE_BYTES) };
    }

    #[test]
    fn start_rejects_empty_allowed_users() {
        let result = TelegramChannel::new("test_token".to_string(), Vec::new()).start();
        assert!(result.is_err());
        assert!(matches!(result.unwrap_err(), ChannelError::Other(_)));
    }

    // ---------------------------------------------------------------------------
    // recv() — injectable sender tests (no network calls)
    // ---------------------------------------------------------------------------

    #[tokio::test]
    async fn recv_returns_channel_message_when_injected() {
        let (mut channel, tx) = TelegramChannel::new_test(vec![]);
        tx.send(plain_message("hello world")).await.unwrap();
        let msg = channel.recv().await.unwrap().unwrap();
        assert_eq!(msg.text, "hello world");
        assert!(msg.attachments.is_empty());
    }

    #[tokio::test]
    async fn recv_reset_command_routed_correctly() {
        let (mut channel, tx) = TelegramChannel::new_test(vec![]);
        tx.send(plain_message("/reset")).await.unwrap();
        let msg = channel.recv().await.unwrap().unwrap();
        assert_eq!(msg.text, "/reset");
    }

    #[tokio::test]
    async fn recv_skills_command_routed_correctly() {
        let (mut channel, tx) = TelegramChannel::new_test(vec![]);
        tx.send(plain_message("/skills")).await.unwrap();
        let msg = channel.recv().await.unwrap().unwrap();
        assert_eq!(msg.text, "/skills");
    }

    #[tokio::test]
    async fn recv_unknown_command_passed_through() {
        let (mut channel, tx) = TelegramChannel::new_test(vec![]);
        tx.send(plain_message("/unknown_cmd arg")).await.unwrap();
        let msg = channel.recv().await.unwrap().unwrap();
        assert_eq!(msg.text, "/unknown_cmd arg");
    }

    #[tokio::test]
    async fn recv_returns_none_when_sender_dropped() {
        let (mut channel, tx) = TelegramChannel::new_test(vec![]);
        drop(tx);
        let result = channel.recv().await.unwrap();
        assert!(result.is_none());
    }

    // ---------------------------------------------------------------------------
    // send_chunk() / flush_chunks() — accumulation (no network calls)
    // ---------------------------------------------------------------------------

    #[tokio::test]
    async fn send_chunk_accumulates_text_without_api_call() {
        let (mut channel, _tx) = TelegramChannel::new_test(vec![]);
        // Suppress the API call by setting last_edit within the 10-second threshold.
        channel.last_edit = Some(Instant::now());

        channel.send_chunk("hello").await.unwrap();
        channel.send_chunk(" world").await.unwrap();

        assert_eq!(channel.accumulated, "hello world");
    }

    #[tokio::test]
    async fn flush_chunks_clears_state_when_no_accumulated_and_no_message_id() {
        let (mut channel, _tx) = TelegramChannel::new_test(vec![]);
        // accumulated is empty and message_id is None — no API call expected.
        channel.last_edit = Some(Instant::now());

        channel.flush_chunks().await.unwrap();

        assert!(channel.accumulated.is_empty());
        assert!(channel.last_edit.is_none());
        assert!(channel.message_id.is_none());
    }

    // flush_chunks() must send accumulated text even when message_id is None.
    // Regression test for the data-loss bug where a short streaming response
    // (entire reply arrives before stream_interval elapses) was silently discarded.
    #[tokio::test]
    async fn flush_chunks_sends_when_accumulated_but_message_id_is_none() {
        let server = MockServer::start().await;
        let (mut channel, _tx) = make_mocked_channel(&server, vec![]).await;

        // Suppress the periodic send by setting last_edit to now (interval not elapsed).
        channel.last_edit = Some(Instant::now());
        channel.accumulated = "short reply".to_string();
        // message_id is None — the periodic path never fired.

        channel.flush_chunks().await.unwrap();

        // The mock server must have received exactly one sendMessage call.
        let requests = server.received_requests().await.unwrap();
        assert!(
            !requests.is_empty(),
            "flush_chunks must send accumulated text even when message_id is None"
        );
        assert!(channel.accumulated.is_empty());
    }

    // ---------------------------------------------------------------------------
    // recv(/start) — mock HTTP server required to intercept the welcome send()
    // ---------------------------------------------------------------------------

    #[tokio::test]
    async fn recv_start_consumed_internally_without_returning_to_caller() {
        let server = MockServer::start().await;
        let (mut channel, tx) = make_mocked_channel(&server, vec![]).await;

        // /start is consumed; recv() loops and waits for the next message.
        tx.send(plain_message("/start")).await.unwrap();
        tx.send(plain_message("hello after start")).await.unwrap();

        let msg = channel.recv().await.unwrap().unwrap();
        assert_eq!(msg.text, "hello after start");
    }

    // ---------------------------------------------------------------------------
    // flush_chunks() with message_id set — mock HTTP server required
    // ---------------------------------------------------------------------------

    #[tokio::test]
    async fn flush_chunks_calls_edit_and_clears_state_when_message_id_set() {
        let server = MockServer::start().await;
        let (mut channel, _tx) = make_mocked_channel(&server, vec![]).await;

        channel.accumulated = "partial response".to_string();
        channel.last_edit = Some(Instant::now());
        channel.message_id = Some(teloxide::types::MessageId(42));

        channel.flush_chunks().await.unwrap();

        assert!(channel.accumulated.is_empty());
        assert!(channel.last_edit.is_none());
        assert!(channel.message_id.is_none());
    }

    // ---------------------------------------------------------------------------
    // confirm() timeout / close / yes — tested at the rx+timeout level in
    // isolation (the same logic confirm() delegates to), avoiding the
    // send() REST call.  Full confirm() round-trips are covered by live
    // agent testing with a real (or mock) Telegram bot.
    // ---------------------------------------------------------------------------

    #[tokio::test]
    async fn confirm_timeout_logic_denies_on_timeout() {
        tokio::time::pause();
        let (_tx, mut rx) = mpsc::channel::<IncomingMessage>(1);
        let timeout_fut = tokio::time::timeout(crate::CONFIRM_TIMEOUT, rx.recv());
        tokio::time::advance(crate::CONFIRM_TIMEOUT + Duration::from_millis(1)).await;
        let result = timeout_fut.await;
        assert!(result.is_err(), "expected timeout Err, got recv result");
    }

    #[tokio::test]
    async fn confirm_close_logic_denies_on_channel_close() {
        let (tx, mut rx) = mpsc::channel::<IncomingMessage>(1);
        drop(tx);
        let result = tokio::time::timeout(crate::CONFIRM_TIMEOUT, rx.recv()).await;
        assert!(result.is_ok(), "should not time out");
        assert!(
            result.unwrap().is_none(),
            "closed channel should yield None"
        );
    }

    #[tokio::test]
    async fn confirm_yes_logic_accepts_yes_response() {
        let (tx, mut rx) = mpsc::channel::<IncomingMessage>(1);
        tx.send(plain_message("yes")).await.unwrap();
        let result = tokio::time::timeout(crate::CONFIRM_TIMEOUT, rx.recv())
            .await
            .unwrap()
            .unwrap();
        assert!(result.text.trim().eq_ignore_ascii_case("yes"));
    }

    #[tokio::test]
    async fn confirm_no_logic_denies_non_yes_response() {
        let (tx, mut rx) = mpsc::channel::<IncomingMessage>(1);
        tx.send(plain_message("no")).await.unwrap();
        let result = tokio::time::timeout(crate::CONFIRM_TIMEOUT, rx.recv())
            .await
            .unwrap()
            .unwrap();
        assert!(!result.text.trim().eq_ignore_ascii_case("yes"));
    }

    // ---------------------------------------------------------------------------
    // send_or_edit() — split at MAX_MESSAGE_LEN
    // ---------------------------------------------------------------------------

    #[tokio::test]
    async fn send_or_edit_splits_long_message_into_multiple_sends() {
        let server = MockServer::start().await;
        let (mut channel, _tx) = make_mocked_channel(&server, vec![]).await;

        // Build text that exceeds MAX_MESSAGE_LEN after markdown_to_telegram pass.
        // Plain ASCII repeated is safe: markdown_to_telegram won't expand it beyond itself.
        let long_text = "a".repeat(MAX_MESSAGE_LEN + 1);
        channel.accumulated = long_text;

        channel.send_or_edit().await.unwrap();

        // The mock server records every request; ≥2 sendMessage calls expected.
        let requests = server.received_requests().await.unwrap();
        assert!(
            requests.len() >= 2,
            "expected ≥2 API calls for oversized message, got {}",
            requests.len()
        );
    }

    #[tokio::test]
    async fn send_or_edit_single_message_when_within_limit() {
        let server = MockServer::start().await;
        let (mut channel, _tx) = make_mocked_channel(&server, vec![]).await;

        channel.accumulated = "short text".to_string();

        channel.send_or_edit().await.unwrap();

        let requests = server.received_requests().await.unwrap();
        assert_eq!(
            requests.len(),
            1,
            "expected exactly 1 API call for short message"
        );
        // message_id must be recorded after a successful send
        assert!(channel.message_id.is_some());
    }

    #[tokio::test]
    async fn send_or_edit_splits_when_edit_overflows() {
        let server = MockServer::start().await;
        let (mut channel, _tx) = make_mocked_channel(&server, vec![]).await;

        // Pre-set a message_id to trigger the edit branch.
        channel.message_id = Some(teloxide::types::MessageId(42));
        let long_text = "b".repeat(MAX_MESSAGE_LEN + 1);
        channel.accumulated = long_text;

        channel.send_or_edit().await.unwrap();

        // Expect: 1 editMessageText call + ≥1 sendMessage call for overflow chunks.
        let requests = server.received_requests().await.unwrap();
        assert!(
            requests.len() >= 2,
            "expected edit + at least 1 overflow send, got {}",
            requests.len()
        );
    }

    // ---------------------------------------------------------------------------
    // elicit() — happy path, timeout, /cancel, field-key sanitization
    // All tests that exercise elicit() need the mock server because elicit()
    // calls self.send() (which calls the Telegram Bot API) before reading rx.
    // ---------------------------------------------------------------------------

    #[tokio::test]
    async fn elicit_happy_path_string_field_returns_accepted() {
        let server = MockServer::start().await;
        let (mut channel, tx) = make_mocked_channel(&server, vec![]).await;

        let request = ElicitationRequest {
            server_name: "test-server".to_owned(),
            message: "Please provide your name".to_owned(),
            fields: vec![ElicitationField {
                name: "username".to_owned(),
                description: None,
                field_type: ElicitationFieldType::String,
                required: true,
            }],
        };

        // Send the answer before calling elicit() so it is buffered in the channel.
        tx.send(plain_message("alice")).await.unwrap();

        let response = channel.elicit(request).await.unwrap();

        match response {
            ElicitationResponse::Accepted(val) => {
                assert_eq!(val["username"], "alice");
            }
            other => panic!("expected Accepted, got {other:?}"),
        }
    }

    #[tokio::test]
    async fn elicit_field_key_uses_raw_name_not_sanitized() {
        let server = MockServer::start().await;
        let (mut channel, tx) = make_mocked_channel(&server, vec![]).await;

        // Field name contains a space — the old sanitize_field_key would strip it to "passphrase".
        let request = ElicitationRequest {
            server_name: "test-server".to_owned(),
            message: "Provide credentials".to_owned(),
            fields: vec![ElicitationField {
                name: "pass phrase".to_owned(),
                description: None,
                field_type: ElicitationFieldType::String,
                required: true,
            }],
        };

        tx.send(plain_message("hunter2")).await.unwrap();
        let response = channel.elicit(request).await.unwrap();

        match response {
            ElicitationResponse::Accepted(val) => {
                assert_eq!(
                    val["pass phrase"], "hunter2",
                    "raw field name must be the map key"
                );
                assert!(
                    val.get("passphrase").is_none(),
                    "sanitized key must not appear in response"
                );
            }
            other => panic!("expected Accepted, got {other:?}"),
        }
    }

    #[tokio::test]
    async fn elicit_cancel_command_returns_cancelled() {
        let server = MockServer::start().await;
        let (mut channel, tx) = make_mocked_channel(&server, vec![]).await;

        let request = ElicitationRequest {
            server_name: "test-server".to_owned(),
            message: "Provide a value".to_owned(),
            fields: vec![ElicitationField {
                name: "token".to_owned(),
                description: None,
                field_type: ElicitationFieldType::String,
                required: true,
            }],
        };

        tx.send(plain_message("/cancel")).await.unwrap();

        let response = channel.elicit(request).await.unwrap();
        assert!(
            matches!(response, ElicitationResponse::Cancelled),
            "expected Cancelled, got {response:?}"
        );
    }

    /// Verify the timeout branch of `elicit()` at the rx level, matching the
    /// same pattern used in `confirm_timeout_logic_denies_on_timeout`.
    #[tokio::test]
    async fn elicit_timeout_logic_cancels_on_timeout() {
        tokio::time::pause();
        let (_tx, mut rx) = mpsc::channel::<IncomingMessage>(1);
        let timeout_fut = tokio::time::timeout(crate::ELICITATION_TIMEOUT, rx.recv());
        tokio::time::advance(crate::ELICITATION_TIMEOUT + Duration::from_millis(1)).await;
        let result = timeout_fut.await;
        assert!(
            result.is_err(),
            "expected Err(Elapsed) for elicitation timeout, got recv result"
        );
    }

    // ---------------------------------------------------------------------------
    // is_bot_authorized — pure function tests
    // ---------------------------------------------------------------------------

    #[test]
    fn is_bot_authorized_empty_list_permits_all() {
        assert!(is_bot_authorized(None, &[]));
        assert!(is_bot_authorized(Some("@any_bot"), &[]));
    }

    #[test]
    fn is_bot_authorized_known_bot_permitted() {
        let allowed = vec!["@trusted".to_string()];
        assert!(is_bot_authorized(Some("@trusted"), &allowed));
        // Without @ prefix on the username side
        assert!(is_bot_authorized(Some("trusted"), &allowed));
    }

    #[test]
    fn is_bot_authorized_unknown_bot_rejected() {
        let allowed = vec!["@trusted".to_string()];
        assert!(!is_bot_authorized(Some("@evil"), &allowed));
        assert!(!is_bot_authorized(None, &allowed));
    }

    // ---------------------------------------------------------------------------
    // Bot depth counter — pure function tests
    // ---------------------------------------------------------------------------

    #[test]
    fn bot_depth_counter_increments_per_call() {
        let counters = BotReplyCounters::default();
        let chat = ChatId(1);
        assert_eq!(increment_bot_depth(&counters, chat), 1);
        assert_eq!(increment_bot_depth(&counters, chat), 2);
        assert_eq!(increment_bot_depth(&counters, chat), 3);
    }

    #[test]
    fn bot_depth_counter_resets_on_human_message() {
        let counters = BotReplyCounters::default();
        let chat = ChatId(42);
        increment_bot_depth(&counters, chat);
        increment_bot_depth(&counters, chat);
        reset_bot_depth(&counters, chat);
        assert_eq!(increment_bot_depth(&counters, chat), 1);
    }

    #[test]
    fn bot_depth_evicts_one_entry_when_at_capacity() {
        let counters: BotReplyCounters = Arc::new(Mutex::new(HashMap::new()));
        // Fill to capacity
        {
            let mut map = counters.lock().unwrap();
            for i in 0..1000_i64 {
                map.insert(ChatId(i), 1);
            }
        }
        // New chat should trigger single-entry eviction, not clear all
        let new_chat = ChatId(1001);
        increment_bot_depth(&counters, new_chat);
        let map = counters.lock().unwrap();
        // Total entries must remain <= MAX_TRACKED_CHATS (one evicted, one added)
        assert_eq!(map.len(), MAX_TRACKED_CHATS);
        // New entry must be present
        assert!(map.contains_key(&new_chat));
    }

    // ---------------------------------------------------------------------------
    // Guest context — recv() stores guest_query_id and sets is_guest_context
    // ---------------------------------------------------------------------------

    #[tokio::test]
    async fn guest_context_stored_on_recv() {
        let (mut channel, tx) = TelegramChannel::new_test(vec!["alice".to_string()]);
        tx.send(IncomingMessage {
            chat_id: ChatId(1),
            text: "hello".to_string(),
            attachments: vec![],
            guest_query_id: Some("qid123".to_string()),
            is_from_bot: false,
        })
        .await
        .unwrap();
        let msg = channel.recv().await.unwrap().unwrap();
        assert!(msg.is_guest_context);
        assert_eq!(channel.guest_query_id.as_deref(), Some("qid123"));
    }

    #[tokio::test]
    async fn send_chunk_does_not_call_api_in_guest_context() {
        let (mut channel, _tx) = TelegramChannel::new_test(vec![]);
        channel.guest_query_id = Some("qid".to_string());
        // Simulate interval elapsed so send_or_edit would fire if not guarded
        channel.last_edit = None;

        channel.send_chunk("part1").await.unwrap();
        channel.send_chunk(" part2").await.unwrap();

        // Accumulated text must be present (guard ran, no API call was possible anyway
        // since there is no mock server — if send_or_edit were called, it would panic)
        assert_eq!(channel.accumulated, "part1 part2");
    }

    #[tokio::test]
    async fn flush_chunks_routes_to_answer_guest_query() {
        use wiremock::matchers::{method, path};

        let server = MockServer::start().await;

        // Mount a scoped mock specifically for answerGuestQuery.
        // SentGuestMessage deserializes {message_id: i64, chat_id: i64}.
        let answer_mock = Mock::given(method("POST"))
            .and(path("/answerGuestQuery"))
            .respond_with(ResponseTemplate::new(200).set_body_json(serde_json::json!({
                "ok": true,
                "result": {
                    "message_id": 1,
                    "chat_id": 2
                }
            })));
        let answer_handle = server.register_as_scoped(answer_mock).await;

        let (mut channel, _tx) = make_mocked_channel(&server, vec![]).await;
        channel.guest_query_id = Some("qid42".to_string());
        channel.accumulated = "response text".to_string();

        channel.flush_chunks().await.unwrap();

        assert_eq!(
            answer_handle.received_requests().await.len(),
            1,
            "answerGuestQuery must be called exactly once"
        );
        assert!(channel.guest_query_id.is_none());
        assert!(channel.accumulated.is_empty());
    }

    #[tokio::test]
    async fn bot_to_bot_false_drops_bot_messages() {
        // With bot_to_bot=false, messages from bots must be silently dropped.
        let (tx, mut rx) = mpsc::channel::<IncomingMessage>(8);
        let bot = Bot::new("test_token");
        let bot_reply_counters: BotReplyCounters = Arc::new(Mutex::new(HashMap::new()));
        let bot_to_bot_active = Arc::new(AtomicBool::new(false));

        let msg_json = serde_json::json!({
            "message_id": 1,
            "date": 1_700_000_000_i64,
            "chat": {"id": 1, "type": "private"},
            "from": {
                "id": 999,
                "is_bot": true,
                "first_name": "EvilBot",
                "username": "evil_bot"
            },
            "text": "bot says hi"
        });
        let msg: teloxide::types::Message = serde_json::from_value(msg_json).unwrap();

        handle_telegram_message(
            bot,
            msg,
            tx.clone(),
            vec!["human".to_string()],
            false, // bot_to_bot disabled
            bot_to_bot_active,
            vec![],
            3,
            bot_reply_counters,
        )
        .await
        .unwrap();

        drop(tx);
        // Channel must be empty: bot message was dropped before reaching tx.send().
        assert!(
            rx.recv().await.is_none(),
            "bot message must be dropped when bot_to_bot=false"
        );
    }

    // ---------------------------------------------------------------------------
    // with_supervisor — verifies that the Telegram listener task is registered
    // ---------------------------------------------------------------------------

    /// Confirm that `with_supervisor()` stores the supervisor so that after
    /// `start()` the listener task appears in the supervisor registry.
    ///
    /// The test uses `new_test()` (no real bot token / network) combined with
    /// a real `TaskSupervisor` running under tokio. Because `start()` spawns a
    /// task that calls `bot.set_my_commands()` (`register_commands`) and then
    /// runs the teloxide dispatcher, we point the bot at a wiremock server
    /// that accepts all requests with HTTP 200 so the task does not immediately
    /// panic due to a network error.
    #[tokio::test]
    async fn with_supervisor_registers_listener_task() {
        use tokio_util::sync::CancellationToken;

        let cancel = CancellationToken::new();
        let sup = zeph_common::TaskSupervisor::new(cancel.clone());

        // new_test creates a channel with a real Bot pointed at a dummy URL.
        // The bot won't be called because we immediately check the registry
        // before the spawned dispatcher has time to make a request.
        let (channel, _tx) = TelegramChannel::new_test(vec!["user".to_string()]);
        let channel = channel.with_supervisor(sup.clone());

        // start() spawns the telegram_listener task via supervisor.
        channel
            .start()
            .expect("start() must succeed with non-empty allowed_users");

        // Give the tokio runtime one yield so the supervisor's state is updated.
        tokio::task::yield_now().await;

        let snapshot = sup.snapshot();
        let names: Vec<&str> = snapshot.iter().map(|s| s.name.as_ref()).collect();
        assert!(
            names.contains(&"telegram_listener"),
            "expected 'telegram_listener' in supervisor snapshot, got: {names:?}"
        );

        cancel.cancel();
    }
}