test_config.py 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759
  1. """Tests for configuration loading."""
  2. import importlib.util
  3. import tempfile
  4. import tomllib
  5. from pathlib import Path
  6. from stt_daemon.config import load_config
  7. REPO_ROOT = Path(__file__).resolve().parents[1]
  8. LAYA_REV = "970dc8c5f63d7b886a68409493f37d569424f933"
  9. def test_load_config_defaults():
  10. """Test that load_config returns defaults when no file exists."""
  11. with tempfile.TemporaryDirectory() as tmpdir:
  12. # Change to temp dir and load (no config.toml exists)
  13. import os
  14. old_cwd = os.getcwd()
  15. try:
  16. os.chdir(tmpdir)
  17. cfg = load_config()
  18. assert cfg.audio.backend == "parec"
  19. assert cfg.audio.device == "VirtualMicSink.monitor"
  20. assert cfg.audio.sample_rate == 16000
  21. assert cfg.vad.aggressiveness == 2
  22. assert cfg.whisper.model == "small"
  23. assert cfg.whisper.language is None
  24. assert cfg.output.stdout is True
  25. assert cfg.output.jsonl is False
  26. finally:
  27. os.chdir(old_cwd)
  28. def test_load_config_from_file():
  29. """Test loading config from a TOML file."""
  30. config_toml = """
  31. [audio]
  32. backend = "sounddevice"
  33. device = "hw:0"
  34. sample_rate = 44100
  35. [vad]
  36. aggressiveness = 3
  37. pre_roll_ms = 500
  38. [whisper]
  39. model = "base"
  40. device = "cuda"
  41. language = "es"
  42. [output]
  43. stdout = false
  44. jsonl = true
  45. jsonl_dir = "logs"
  46. """
  47. with tempfile.TemporaryDirectory() as tmpdir:
  48. config_path = Path(tmpdir) / "config.toml"
  49. config_path.write_text(config_toml)
  50. cfg = load_config(str(config_path))
  51. assert cfg.audio.backend == "sounddevice"
  52. assert cfg.audio.device == "hw:0"
  53. assert cfg.audio.sample_rate == 44100
  54. assert cfg.vad.aggressiveness == 3
  55. assert cfg.vad.pre_roll_ms == 500
  56. assert cfg.whisper.model == "base"
  57. assert cfg.whisper.device == "cuda"
  58. assert cfg.whisper.language == "es"
  59. assert cfg.output.stdout is False
  60. assert cfg.output.jsonl is True
  61. assert cfg.output.jsonl_dir == "logs"
  62. def test_language_normalization_auto():
  63. """Test that language 'auto' is normalized to None."""
  64. config_toml = """
  65. [whisper]
  66. language = "auto"
  67. """
  68. with tempfile.TemporaryDirectory() as tmpdir:
  69. config_path = Path(tmpdir) / "config.toml"
  70. config_path.write_text(config_toml)
  71. cfg = load_config(str(config_path))
  72. assert cfg.whisper.language is None
  73. def test_language_normalization_empty_string():
  74. """Test that language '' is normalized to None."""
  75. config_toml = """
  76. [whisper]
  77. language = ""
  78. """
  79. with tempfile.TemporaryDirectory() as tmpdir:
  80. config_path = Path(tmpdir) / "config.toml"
  81. config_path.write_text(config_toml)
  82. cfg = load_config(str(config_path))
  83. assert cfg.whisper.language is None
  84. def test_language_normalization_none():
  85. """Test that language 'none' is normalized to None."""
  86. config_toml = """
  87. [whisper]
  88. language = "none"
  89. """
  90. with tempfile.TemporaryDirectory() as tmpdir:
  91. config_path = Path(tmpdir) / "config.toml"
  92. config_path.write_text(config_toml)
  93. cfg = load_config(str(config_path))
  94. assert cfg.whisper.language is None
  95. def test_language_explicit_en():
  96. """Test that explicit language codes are preserved."""
  97. config_toml = """
  98. [whisper]
  99. language = "en"
  100. """
  101. with tempfile.TemporaryDirectory() as tmpdir:
  102. config_path = Path(tmpdir) / "config.toml"
  103. config_path.write_text(config_toml)
  104. cfg = load_config(str(config_path))
  105. assert cfg.whisper.language == "en"
  106. def test_partial_config():
  107. """Test that partial configs merge with defaults."""
  108. config_toml = """
  109. [audio]
  110. device = "custom_device"
  111. """
  112. with tempfile.TemporaryDirectory() as tmpdir:
  113. config_path = Path(tmpdir) / "config.toml"
  114. config_path.write_text(config_toml)
  115. cfg = load_config(str(config_path))
  116. # Overridden
  117. assert cfg.audio.device == "custom_device"
  118. # Defaults
  119. assert cfg.audio.backend == "parec"
  120. assert cfg.audio.sample_rate == 16000
  121. assert cfg.vad.aggressiveness == 2
  122. def test_chunking_and_confidence_fields():
  123. """New chunking/gate fields load from TOML."""
  124. config_toml = """
  125. [vad]
  126. chunk_pause_ms = 500
  127. soft_max_s = 8.0
  128. rms_floor = 0.01
  129. trailing_trim_ms = 150
  130. [whisper]
  131. no_speech_prob_max = 0.7
  132. avg_logprob_min = -0.8
  133. """
  134. with tempfile.TemporaryDirectory() as tmpdir:
  135. config_path = Path(tmpdir) / "config.toml"
  136. config_path.write_text(config_toml)
  137. cfg = load_config(str(config_path))
  138. assert cfg.vad.chunk_pause_ms == 500
  139. assert cfg.vad.soft_max_s == 8.0
  140. assert cfg.vad.rms_floor == 0.01
  141. assert cfg.vad.trailing_trim_ms == 150
  142. assert cfg.whisper.no_speech_prob_max == 0.7
  143. assert cfg.whisper.avg_logprob_min == -0.8
  144. def test_chunking_defaults():
  145. """New fields have sane defaults."""
  146. with tempfile.TemporaryDirectory() as tmpdir:
  147. import os
  148. old_cwd = os.getcwd()
  149. try:
  150. os.chdir(tmpdir)
  151. cfg = load_config()
  152. finally:
  153. os.chdir(old_cwd)
  154. assert cfg.vad.chunk_pause_ms == 350
  155. assert cfg.vad.soft_max_s == 10.0
  156. assert cfg.vad.rms_floor == 0.008
  157. assert cfg.vad.trailing_trim_ms == 200
  158. assert cfg.whisper.no_speech_prob_max == 0.6
  159. assert cfg.whisper.avg_logprob_min == -1.0
  160. def test_output_defaults_disable_audio_saving():
  161. with tempfile.TemporaryDirectory() as tmpdir:
  162. import os
  163. old_cwd = os.getcwd()
  164. try:
  165. os.chdir(tmpdir)
  166. cfg = load_config()
  167. finally:
  168. os.chdir(old_cwd)
  169. assert cfg.output.save_audio is False
  170. assert cfg.output.audio_dir == "audio"
  171. def test_output_save_audio_fields():
  172. config_toml = """
  173. [output]
  174. save_audio = true
  175. audio_dir = "clips"
  176. """
  177. with tempfile.TemporaryDirectory() as tmpdir:
  178. config_path = Path(tmpdir) / "config.toml"
  179. config_path.write_text(config_toml)
  180. cfg = load_config(str(config_path))
  181. assert cfg.output.save_audio is True
  182. assert cfg.output.audio_dir == "clips"
  183. def test_daemon_defaults():
  184. with tempfile.TemporaryDirectory() as tmpdir:
  185. import os
  186. old_cwd = os.getcwd()
  187. try:
  188. os.chdir(tmpdir)
  189. cfg = load_config()
  190. finally:
  191. os.chdir(old_cwd)
  192. assert cfg.daemon.queue_size == 50
  193. assert cfg.daemon.backlog_warn_at == 5
  194. def test_daemon_config_fields():
  195. config_toml = """
  196. [daemon]
  197. queue_size = 20
  198. backlog_warn_at = 3
  199. """
  200. with tempfile.TemporaryDirectory() as tmpdir:
  201. config_path = Path(tmpdir) / "config.toml"
  202. config_path.write_text(config_toml)
  203. cfg = load_config(str(config_path))
  204. assert cfg.daemon.queue_size == 20
  205. assert cfg.daemon.backlog_warn_at == 3
  206. def test_unknown_sections_warn_and_are_ignored(caplog):
  207. """A typo'd section like [assitant] must warn, not silently use defaults."""
  208. config_toml = """
  209. [assitant]
  210. require_viewer = false
  211. """
  212. with tempfile.TemporaryDirectory() as tmpdir:
  213. config_path = Path(tmpdir) / "config.toml"
  214. config_path.write_text(config_toml)
  215. with caplog.at_level("WARNING", logger="stt_daemon.config"):
  216. cfg = load_config(str(config_path))
  217. assert "assitant" in caplog.text
  218. assert cfg.assistant.require_viewer is True
  219. def test_cpu_threads_default_is_auto():
  220. """cpu_threads defaults to 0 (CTranslate2's automatic choice)."""
  221. with tempfile.TemporaryDirectory() as tmpdir:
  222. import os
  223. old_cwd = os.getcwd()
  224. try:
  225. os.chdir(tmpdir)
  226. cfg = load_config()
  227. finally:
  228. os.chdir(old_cwd)
  229. assert cfg.whisper.cpu_threads == 0
  230. def test_cpu_threads_config_field():
  231. config_toml = """
  232. [whisper]
  233. cpu_threads = 16
  234. """
  235. with tempfile.TemporaryDirectory() as tmpdir:
  236. config_path = Path(tmpdir) / "config.toml"
  237. config_path.write_text(config_toml)
  238. cfg = load_config(str(config_path))
  239. assert cfg.whisper.cpu_threads == 16
  240. def test_language_strategy_defaults_to_static():
  241. """language_strategy defaults to 'static'."""
  242. with tempfile.TemporaryDirectory() as tmpdir:
  243. import os
  244. old_cwd = os.getcwd()
  245. try:
  246. os.chdir(tmpdir)
  247. cfg = load_config()
  248. finally:
  249. os.chdir(old_cwd)
  250. assert cfg.whisper.language_strategy == "static"
  251. def test_language_strategy_sticky_is_parsed():
  252. """language_strategy 'sticky' is parsed correctly."""
  253. config_toml = """
  254. [whisper]
  255. language_strategy = "sticky"
  256. """
  257. with tempfile.TemporaryDirectory() as tmpdir:
  258. config_path = Path(tmpdir) / "config.toml"
  259. config_path.write_text(config_toml)
  260. cfg = load_config(str(config_path))
  261. assert cfg.whisper.language_strategy == "sticky"
  262. def test_unknown_language_strategy_falls_back_to_static(caplog):
  263. """An unknown language_strategy value falls back to 'static' with a warning."""
  264. config_toml = """
  265. [whisper]
  266. language_strategy = "bogus"
  267. """
  268. with tempfile.TemporaryDirectory() as tmpdir:
  269. config_path = Path(tmpdir) / "config.toml"
  270. config_path.write_text(config_toml)
  271. with caplog.at_level("WARNING", logger="stt_daemon.config"):
  272. cfg = load_config(str(config_path))
  273. assert cfg.whisper.language_strategy == "static"
  274. assert "Unknown language_strategy" in caplog.text
  275. assert "bogus" in caplog.text
  276. def test_sticky_redetect_logprob_defaults_to_minus_0_8():
  277. """sticky_redetect_logprob defaults to -0.8."""
  278. with tempfile.TemporaryDirectory() as tmpdir:
  279. import os
  280. old_cwd = os.getcwd()
  281. try:
  282. os.chdir(tmpdir)
  283. cfg = load_config()
  284. finally:
  285. os.chdir(old_cwd)
  286. assert cfg.whisper.sticky_redetect_logprob == -0.8
  287. def test_sticky_redetect_logprob_is_parsed():
  288. """sticky_redetect_logprob is parsed from TOML."""
  289. config_toml = """
  290. [whisper]
  291. sticky_redetect_logprob = -0.5
  292. """
  293. with tempfile.TemporaryDirectory() as tmpdir:
  294. config_path = Path(tmpdir) / "config.toml"
  295. config_path.write_text(config_toml)
  296. cfg = load_config(str(config_path))
  297. assert cfg.whisper.sticky_redetect_logprob == -0.5
  298. def test_llm_context_and_compaction_defaults():
  299. with tempfile.TemporaryDirectory() as tmpdir:
  300. import os
  301. old_cwd = os.getcwd()
  302. try:
  303. os.chdir(tmpdir)
  304. cfg = load_config()
  305. finally:
  306. os.chdir(old_cwd)
  307. assert cfg.llm.max_context_tokens == 131072
  308. assert cfg.llm.compaction_safety_margin == 256
  309. assert cfg.llm.compaction_enabled is True
  310. assert cfg.llm.compaction_trigger_ratio == 0.8
  311. assert cfg.llm.compaction_keep_turns == 6
  312. assert cfg.llm.compaction_max_tokens == 512
  313. assert cfg.llm.compaction_timeout_s == 30.0
  314. assert cfg.llm.compaction_input_tokens == 8192
  315. assert cfg.llm.inference_queue_size == 8
  316. assert cfg.llm.prompt_budget == 131072 - 1024 - 256
  317. assert not hasattr(cfg.llm, "max_prompt_tokens")
  318. def test_llm_context_and_compaction_fields_parsed():
  319. config_toml = """
  320. [llm]
  321. max_context_tokens = 8192
  322. max_tokens = 2048
  323. compaction_safety_margin = 128
  324. compaction_enabled = false
  325. compaction_trigger_ratio = 0.5
  326. compaction_keep_turns = 2
  327. compaction_max_tokens = 256
  328. compaction_timeout_s = 12.5
  329. compaction_input_tokens = 4096
  330. inference_queue_size = 3
  331. """
  332. with tempfile.TemporaryDirectory() as tmpdir:
  333. config_path = Path(tmpdir) / "config.toml"
  334. config_path.write_text(config_toml)
  335. cfg = load_config(str(config_path))
  336. assert cfg.llm.max_context_tokens == 8192
  337. assert cfg.llm.compaction_safety_margin == 128
  338. assert cfg.llm.compaction_enabled is False
  339. assert cfg.llm.compaction_trigger_ratio == 0.5
  340. assert cfg.llm.compaction_keep_turns == 2
  341. assert cfg.llm.compaction_max_tokens == 256
  342. assert cfg.llm.compaction_timeout_s == 12.5
  343. assert cfg.llm.compaction_input_tokens == 4096
  344. assert cfg.llm.inference_queue_size == 3
  345. assert cfg.llm.prompt_budget == 8192 - 2048 - 128
  346. def test_max_prompt_tokens_is_deprecated_and_ignored(caplog):
  347. config_toml = """
  348. [llm]
  349. max_prompt_tokens = 50
  350. """
  351. with tempfile.TemporaryDirectory() as tmpdir:
  352. config_path = Path(tmpdir) / "config.toml"
  353. config_path.write_text(config_toml)
  354. with caplog.at_level("WARNING", logger="stt_daemon.config"):
  355. cfg = load_config(str(config_path))
  356. assert "max_prompt_tokens" in caplog.text
  357. assert "deprecated" in caplog.text
  358. assert cfg.llm.prompt_budget == 131072 - 1024 - 256
  359. def test_backend_defaults_to_faster_whisper():
  360. """backend defaults to 'faster-whisper'."""
  361. with tempfile.TemporaryDirectory() as tmpdir:
  362. import os
  363. old_cwd = os.getcwd()
  364. try:
  365. os.chdir(tmpdir)
  366. cfg = load_config()
  367. finally:
  368. os.chdir(old_cwd)
  369. assert cfg.whisper.backend == "faster-whisper"
  370. def test_backend_pywhispercpp_is_parsed():
  371. """backend 'pywhispercpp' is parsed from TOML."""
  372. config_toml = """
  373. [whisper]
  374. backend = "pywhispercpp"
  375. """
  376. with tempfile.TemporaryDirectory() as tmpdir:
  377. config_path = Path(tmpdir) / "config.toml"
  378. config_path.write_text(config_toml)
  379. cfg = load_config(str(config_path))
  380. assert cfg.whisper.backend == "pywhispercpp"
  381. def test_backend_is_normalized():
  382. """Backend names are normalized: stripped, lowercased, underscores to dashes."""
  383. config_toml = """
  384. [whisper]
  385. backend = " Faster_Whisper "
  386. """
  387. with tempfile.TemporaryDirectory() as tmpdir:
  388. config_path = Path(tmpdir) / "config.toml"
  389. config_path.write_text(config_toml)
  390. cfg = load_config(str(config_path))
  391. assert cfg.whisper.backend == "faster-whisper"
  392. def test_unknown_backend_falls_back_to_faster_whisper(caplog):
  393. """An unknown backend value falls back to 'faster-whisper' with a warning."""
  394. config_toml = """
  395. [whisper]
  396. backend = "bogus"
  397. """
  398. with tempfile.TemporaryDirectory() as tmpdir:
  399. config_path = Path(tmpdir) / "config.toml"
  400. config_path.write_text(config_toml)
  401. with caplog.at_level("WARNING", logger="stt_daemon.config"):
  402. cfg = load_config(str(config_path))
  403. assert cfg.whisper.backend == "faster-whisper"
  404. assert "Unknown whisper backend" in caplog.text
  405. assert "bogus" in caplog.text
  406. def test_triage_defaults():
  407. """The [triage] section defaults match the plan's surviving-key list."""
  408. with tempfile.TemporaryDirectory() as tmpdir:
  409. import os
  410. old_cwd = os.getcwd()
  411. try:
  412. os.chdir(tmpdir)
  413. cfg = load_config()
  414. finally:
  415. os.chdir(old_cwd)
  416. assert cfg.triage.enabled is False
  417. assert cfg.triage.questions_file == "questions.json"
  418. assert cfg.triage.window == 500
  419. assert cfg.triage.state_char_budget == 3000
  420. assert cfg.triage.predict_max_tokens == 1024
  421. assert cfg.triage.confidence_field == "answer_confidence"
  422. assert cfg.triage.escalate_threshold == 0.6
  423. assert cfg.triage.choice_min_confidence == 0.85
  424. assert cfg.triage.extract_enabled is True
  425. assert cfg.triage.extract_max_tokens == 512
  426. assert cfg.triage.extract_timeout_s == 30.0
  427. assert cfg.triage.device == ""
  428. assert cfg.triage.preload == ["multilingual"]
  429. assert cfg.triage.default_checkpoint == "multilingual"
  430. def test_triage_config_fields_parsed():
  431. """Every surviving [triage] key is applied from TOML."""
  432. config_toml = """
  433. [triage]
  434. enabled = true
  435. questions_file = "custom_questions.json"
  436. window = 4
  437. state_char_budget = 800
  438. predict_max_tokens = 768
  439. confidence_field = "confidence"
  440. escalate_threshold = 0.7
  441. choice_min_confidence = 0.9
  442. extract_enabled = false
  443. extract_max_tokens = 256
  444. extract_timeout_s = 12.5
  445. device = "cuda"
  446. preload = ["multilingual", "english"]
  447. default_checkpoint = "english"
  448. """
  449. with tempfile.TemporaryDirectory() as tmpdir:
  450. config_path = Path(tmpdir) / "config.toml"
  451. config_path.write_text(config_toml)
  452. cfg = load_config(str(config_path))
  453. assert cfg.triage.enabled is True
  454. assert cfg.triage.questions_file == "custom_questions.json"
  455. assert cfg.triage.window == 4
  456. assert cfg.triage.state_char_budget == 800
  457. assert cfg.triage.predict_max_tokens == 768
  458. assert cfg.triage.confidence_field == "confidence"
  459. assert cfg.triage.escalate_threshold == 0.7
  460. assert cfg.triage.choice_min_confidence == 0.9
  461. assert cfg.triage.extract_enabled is False
  462. assert cfg.triage.extract_max_tokens == 256
  463. assert cfg.triage.extract_timeout_s == 12.5
  464. assert cfg.triage.device == "cuda"
  465. assert cfg.triage.preload == ["multilingual", "english"]
  466. assert cfg.triage.default_checkpoint == "english"
  467. def test_unknown_triag_section_warns_and_keeps_defaults(caplog):
  468. """A typo'd [triag] section must warn and never raise; triage keeps defaults."""
  469. config_toml = """
  470. [triag]
  471. enabled = true
  472. window = 3
  473. """
  474. with tempfile.TemporaryDirectory() as tmpdir:
  475. config_path = Path(tmpdir) / "config.toml"
  476. config_path.write_text(config_toml)
  477. with caplog.at_level("WARNING", logger="stt_daemon.config"):
  478. cfg = load_config(str(config_path))
  479. assert "triag" in caplog.text
  480. assert cfg.triage.enabled is False
  481. assert cfg.triage.window == 500
  482. def test_triage_config_has_no_struck_keys():
  483. """The v5 keys are deliberately gone from TriageConfig (superseded/removed)."""
  484. from stt_daemon.config import TriageConfig
  485. struck = {"socket", "model", "predict_timeout_s", "checkpoint", "subfolder", "max_len"}
  486. for name in struck:
  487. assert not hasattr(TriageConfig, name), f"TriageConfig must not expose {name!r}"
  488. def test_llm_triage_system_prompt_is_parsed():
  489. """[llm].triage_system_prompt overrides the JSON-only extractor default."""
  490. config_toml = """
  491. [llm]
  492. triage_system_prompt = "ONLY JSON, nothing else."
  493. """
  494. with tempfile.TemporaryDirectory() as tmpdir:
  495. config_path = Path(tmpdir) / "config.toml"
  496. config_path.write_text(config_toml)
  497. cfg = load_config(str(config_path))
  498. assert cfg.llm.triage_system_prompt == "ONLY JSON, nothing else."
  499. def test_no_stt_laya_script():
  500. """Triage is in-process only: no stt-laya console script may ever be added."""
  501. with (REPO_ROOT / "pyproject.toml").open("rb") as fh:
  502. data = tomllib.load(fh)
  503. scripts = data["project"]["scripts"]
  504. assert set(scripts) == {"stt-daemon", "stt-assistant", "stt-ctl", "stt-tui"}
  505. assert not any("laya" in name for name in scripts)
  506. def test_layasvc_module_not_importable():
  507. """The retired external laya service package must not exist (in-process lane)."""
  508. assert importlib.util.find_spec("stt_daemon.layasvc") is None
  509. def test_triage_extra_pins_laya_revision():
  510. """The triage extra pins laya to the exact verified upstream revision."""
  511. with (REPO_ROOT / "pyproject.toml").open("rb") as fh:
  512. data = tomllib.load(fh)
  513. assert data["project"]["optional-dependencies"]["triage"] == ["laya"]
  514. source = data["tool"]["uv"]["sources"]["laya"]
  515. assert source["git"] == "https://github.com/NandhaKishorM/laya.git"
  516. assert source["rev"] == LAYA_REV
  517. def test_questions_json_gitignored():
  518. """The user's private questions.json is never committed (like config.toml)."""
  519. lines = (REPO_ROOT / ".gitignore").read_text().splitlines()
  520. assert "questions.json" in lines
  521. def test_speaker_defaults():
  522. """The [speaker] section defaults match the plan's 14-key list."""
  523. with tempfile.TemporaryDirectory() as tmpdir:
  524. import os
  525. old_cwd = os.getcwd()
  526. try:
  527. os.chdir(tmpdir)
  528. cfg = load_config()
  529. finally:
  530. os.chdir(old_cwd)
  531. assert cfg.speaker.enabled is False
  532. assert cfg.speaker.model_path is None
  533. assert cfg.speaker.provider == "cpu"
  534. assert cfg.speaker.threshold_id == 0.60
  535. assert cfg.speaker.threshold_new == 0.45
  536. assert cfg.speaker.min_enroll_s == 2.0
  537. assert cfg.speaker.rms_floor == 0.0
  538. assert cfg.speaker.margin == 0.04
  539. assert cfg.speaker.mint_cooldown_s == 60.0
  540. assert cfg.speaker.confirm_s == 12.0
  541. assert cfg.speaker.max_prototypes == 8
  542. assert cfg.speaker.provisional_ttl_h == 24.0
  543. assert cfg.speaker.store_dir == "speakers"
  544. assert cfg.speaker.join_timeout_s == 2.0
  545. def test_speaker_config_fields_parsed():
  546. """Every [speaker] key is applied from TOML."""
  547. config_toml = """
  548. [speaker]
  549. enabled = true
  550. model_path = "models/nemo_en_titanet_small.onnx"
  551. provider = "cuda"
  552. threshold_id = 0.7
  553. threshold_new = 0.5
  554. min_enroll_s = 3.0
  555. rms_floor = 0.01
  556. margin = 0.05
  557. mint_cooldown_s = 30.0
  558. confirm_s = 8.0
  559. max_prototypes = 4
  560. provisional_ttl_h = 12.0
  561. store_dir = "voiceprints"
  562. join_timeout_s = 1.5
  563. """
  564. with tempfile.TemporaryDirectory() as tmpdir:
  565. config_path = Path(tmpdir) / "config.toml"
  566. config_path.write_text(config_toml)
  567. cfg = load_config(str(config_path))
  568. assert cfg.speaker.enabled is True
  569. assert cfg.speaker.model_path == "models/nemo_en_titanet_small.onnx"
  570. assert cfg.speaker.provider == "cuda"
  571. assert cfg.speaker.threshold_id == 0.7
  572. assert cfg.speaker.threshold_new == 0.5
  573. assert cfg.speaker.min_enroll_s == 3.0
  574. assert cfg.speaker.rms_floor == 0.01
  575. assert cfg.speaker.margin == 0.05
  576. assert cfg.speaker.mint_cooldown_s == 30.0
  577. assert cfg.speaker.confirm_s == 8.0
  578. assert cfg.speaker.max_prototypes == 4
  579. assert cfg.speaker.provisional_ttl_h == 12.0
  580. assert cfg.speaker.store_dir == "voiceprints"
  581. assert cfg.speaker.join_timeout_s == 1.5
  582. def test_unknown_speaker_key_is_ignored():
  583. """An unknown key inside [speaker] is silently ignored (per-field .get)."""
  584. config_toml = """
  585. [speaker]
  586. enabled = true
  587. bogus_key = 42
  588. """
  589. with tempfile.TemporaryDirectory() as tmpdir:
  590. config_path = Path(tmpdir) / "config.toml"
  591. config_path.write_text(config_toml)
  592. cfg = load_config(str(config_path))
  593. assert cfg.speaker.enabled is True
  594. assert not hasattr(cfg.speaker, "bogus_key")
  595. assert cfg.speaker.provider == "cpu"
  596. def test_unknown_speaker_provider_falls_back_to_cpu(caplog):
  597. """An unknown speaker provider falls back to 'cpu' with a warning."""
  598. config_toml = """
  599. [speaker]
  600. provider = "bogus"
  601. """
  602. with tempfile.TemporaryDirectory() as tmpdir:
  603. config_path = Path(tmpdir) / "config.toml"
  604. config_path.write_text(config_toml)
  605. with caplog.at_level("WARNING", logger="stt_daemon.config"):
  606. cfg = load_config(str(config_path))
  607. assert cfg.speaker.provider == "cpu"
  608. assert "Unknown speaker provider" in caplog.text
  609. assert "bogus" in caplog.text