| 1 | import 'package:dio/dio.dart'; |
| 2 | import 'package:flutter_test/flutter_test.dart'; |
| 3 | import 'package:localmind/core/models/enums.dart'; |
| 4 | import 'package:localmind/features/servers/data/models/server.dart'; |
| 5 | import 'package:localmind/features/servers/data/server_api_service.dart'; |
| 6 | import 'dart:convert'; |
| 7 | |
| 8 | class TestInterceptor extends Interceptor { |
| 9 | final dynamic responseData; |
| 10 | TestInterceptor(this.responseData); |
| 11 | |
| 12 | @override |
| 13 | void onRequest(RequestOptions options, RequestInterceptorHandler handler) { |
| 14 | handler.resolve( |
| 15 | Response(requestOptions: options, data: responseData, statusCode: 200), |
| 16 | ); |
| 17 | } |
| 18 | } |
| 19 | |
| 20 | class RequestRecordingInterceptor extends Interceptor { |
| 21 | final List<RequestOptions> requests = []; |
| 22 | final int statusCode; |
| 23 | RequestRecordingInterceptor({this.statusCode = 200}); |
| 24 | |
| 25 | @override |
| 26 | void onRequest(RequestOptions options, RequestInterceptorHandler handler) { |
| 27 | requests.add(options); |
| 28 | handler.resolve( |
| 29 | Response( |
| 30 | requestOptions: options, |
| 31 | statusCode: statusCode, |
| 32 | data: {'status': 'ok'}, |
| 33 | ), |
| 34 | ); |
| 35 | } |
| 36 | } |
| 37 | |
| 38 | class FailingInterceptor extends Interceptor { |
| 39 | @override |
| 40 | void onRequest(RequestOptions options, RequestInterceptorHandler handler) { |
| 41 | handler.reject( |
| 42 | DioException( |
| 43 | requestOptions: options, |
| 44 | type: DioExceptionType.connectionError, |
| 45 | message: 'connection refused', |
| 46 | ), |
| 47 | ); |
| 48 | } |
| 49 | } |
| 50 | |
| 51 | class RoutingInterceptor extends Interceptor { |
| 52 | RoutingInterceptor(this.responses); |
| 53 | |
| 54 | /// Maps a request key ("METHOD /path" or "METHOD /path:body-name") to a |
| 55 | /// mock response body. Keys match against the request's path suffix so |
| 56 | /// tests don't need a baseUrl. |
| 57 | final Map<String, dynamic> responses; |
| 58 | |
| 59 | @override |
| 60 | void onRequest(RequestOptions options, RequestInterceptorHandler handler) { |
| 61 | final key = _matchKey(options); |
| 62 | final body = key == null ? null : responses[key]; |
| 63 | if (body == null) { |
| 64 | handler.reject( |
| 65 | DioException( |
| 66 | requestOptions: options, |
| 67 | type: DioExceptionType.badResponse, |
| 68 | message: 'No mock for ${options.method} ${options.path}', |
| 69 | ), |
| 70 | ); |
| 71 | return; |
| 72 | } |
| 73 | handler.resolve( |
| 74 | Response(requestOptions: options, statusCode: 200, data: body), |
| 75 | ); |
| 76 | } |
| 77 | |
| 78 | String? _matchKey(RequestOptions options) { |
| 79 | final method = options.method.toUpperCase(); |
| 80 | final path = options.path; |
| 81 | final base = '$method $path'; |
| 82 | final suffix = '$method ${_suffix(path)}'; |
| 83 | for (final candidate in responses.keys) { |
| 84 | if (candidate == base || candidate == suffix) return candidate; |
| 85 | } |
| 86 | final name = _bodyName(options.data); |
| 87 | if (name == null) return null; |
| 88 | final baseKey = '$base:$name'; |
| 89 | if (responses.containsKey(baseKey)) return baseKey; |
| 90 | return '$suffix:$name'; |
| 91 | } |
| 92 | |
| 93 | String _suffix(String path) { |
| 94 | final idx = path.indexOf('/api/'); |
| 95 | if (idx < 0) return path; |
| 96 | return path.substring(idx); |
| 97 | } |
| 98 | |
| 99 | String? _bodyName(dynamic data) { |
| 100 | if (data is Map) { |
| 101 | return data['model']?.toString() ?? data['name']?.toString(); |
| 102 | } |
| 103 | if (data is String && data.isNotEmpty) { |
| 104 | try { |
| 105 | final decoded = jsonDecode(data); |
| 106 | if (decoded is Map) { |
| 107 | return decoded['model']?.toString() ?? decoded['name']?.toString(); |
| 108 | } |
| 109 | } catch (_) { |
| 110 | // Not JSON — leave name null. |
| 111 | } |
| 112 | } |
| 113 | return null; |
| 114 | } |
| 115 | } |
| 116 | |
| 117 | void main() { |
| 118 | group('ServerApiService - OpenAI/llama.cpp model list parsing', () { |
| 119 | late Server testServer; |
| 120 | |
| 121 | setUp(() { |
| 122 | testServer = Server( |
| 123 | id: 'test-openai', |
| 124 | name: 'Test OpenAI Compatible', |
| 125 | type: ServerType.openAICompatible, |
| 126 | host: 'localhost', |
| 127 | port: 8080, |
| 128 | createdAt: DateTime.now(), |
| 129 | lastConnectedAt: DateTime.now(), |
| 130 | ); |
| 131 | }); |
| 132 | |
| 133 | test( |
| 134 | 'parses models successfully using fallback data array (llama.cpp / OpenAI standard)', |
| 135 | () async { |
| 136 | final mockData = { |
| 137 | "object": "list", |
| 138 | "data": [ |
| 139 | { |
| 140 | "id": "meta-llama/Llama-3-8B-Instruct", |
| 141 | "object": "model", |
| 142 | "created": 1677610602, |
| 143 | "owned_by": "openai", |
| 144 | "meta": { |
| 145 | "n_ctx_train": 8192, |
| 146 | "n_params": 8000000000, |
| 147 | "size": 4800000000, |
| 148 | }, |
| 149 | }, |
| 150 | ], |
| 151 | }; |
| 152 | |
| 153 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 154 | final service = ServerApiService(dio); |
| 155 | |
| 156 | final models = await service.fetchModels(testServer); |
| 157 | |
| 158 | expect(models, hasLength(1)); |
| 159 | final model = models.first; |
| 160 | expect(model.id, "meta-llama/Llama-3-8B-Instruct"); |
| 161 | expect(model.name, "Meta Llama/Llama 3 8B Instruct"); |
| 162 | expect(model.contextLength, 8192); |
| 163 | expect(model.parameterCount, 8.0); // 8000000000 / 1000000000 |
| 164 | expect(model.fileSize, 4800000000); |
| 165 | expect(model.serverType, ServerType.openAICompatible); |
| 166 | expect(model.serverId, testServer.id); |
| 167 | }, |
| 168 | ); |
| 169 | |
| 170 | test( |
| 171 | 'parses models successfully using custom models array (LM Studio)', |
| 172 | () async { |
| 173 | final mockData = { |
| 174 | "models": [ |
| 175 | { |
| 176 | "key": "lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF", |
| 177 | "display_name": "Llama 3 8B Instruct", |
| 178 | "description": "LM Studio test model", |
| 179 | "params_string": "8B", |
| 180 | "max_context_length": 4096, |
| 181 | "size_bytes": 5000000000, |
| 182 | "quantization": {"name": "Q4_K_M"}, |
| 183 | "architecture": "llama", |
| 184 | }, |
| 185 | ], |
| 186 | }; |
| 187 | |
| 188 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 189 | final service = ServerApiService(dio); |
| 190 | |
| 191 | final models = await service.fetchModels(testServer); |
| 192 | |
| 193 | expect(models, hasLength(1)); |
| 194 | final model = models.first; |
| 195 | expect(model.id, "lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF"); |
| 196 | expect(model.name, "Llama 3 8B Instruct"); |
| 197 | expect(model.description, "LM Studio test model"); |
| 198 | expect(model.parameterCount, 8.0); |
| 199 | expect(model.contextLength, 4096); |
| 200 | expect(model.fileSize, 5000000000); |
| 201 | expect(model.quantization, "Q4_K_M"); |
| 202 | expect(model.architecture, "llama"); |
| 203 | }, |
| 204 | ); |
| 205 | |
| 206 | test('parses model capabilities from LM Studio models array', () async { |
| 207 | final mockData = { |
| 208 | "models": [ |
| 209 | { |
| 210 | "key": "google/gemma-4-e2b", |
| 211 | "display_name": "Gemma 4 E2B", |
| 212 | "capabilities": {"vision": true, "trained_for_tool_use": true}, |
| 213 | }, |
| 214 | { |
| 215 | "key": "mistralai/ministral-3-14b-reasoning", |
| 216 | "display_name": "Ministral 3 14B Reasoning", |
| 217 | "capabilities": { |
| 218 | "vision": true, |
| 219 | "trained_for_tool_use": true, |
| 220 | "reasoning": { |
| 221 | "allowed_options": ["off", "on"], |
| 222 | "default": "on", |
| 223 | }, |
| 224 | }, |
| 225 | }, |
| 226 | { |
| 227 | "key": "mradermacher/kimi-vl-a3b-thinking-2506-i1", |
| 228 | "display_name": "Kimi VL A3B Thinking 2506 I1", |
| 229 | "capabilities": {"vision": false, "trained_for_tool_use": false}, |
| 230 | }, |
| 231 | ], |
| 232 | }; |
| 233 | |
| 234 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 235 | final service = ServerApiService(dio); |
| 236 | |
| 237 | final models = await service.fetchModels(testServer); |
| 238 | |
| 239 | expect(models, hasLength(3)); |
| 240 | |
| 241 | expect(models[0].supportsVision, isTrue); |
| 242 | expect(models[0].supportsToolUse, isTrue); |
| 243 | expect(models[0].supportsReasoning, isFalse); |
| 244 | |
| 245 | expect(models[1].supportsVision, isTrue); |
| 246 | expect(models[1].supportsToolUse, isTrue); |
| 247 | expect(models[1].supportsReasoning, isTrue); |
| 248 | |
| 249 | expect(models[2].supportsVision, isFalse); |
| 250 | expect(models[2].supportsToolUse, isFalse); |
| 251 | expect(models[2].supportsReasoning, isFalse); |
| 252 | }); |
| 253 | |
| 254 | test( |
| 255 | 'only infers standard reasoning-effort models without capabilities', |
| 256 | () async { |
| 257 | final mockData = { |
| 258 | "object": "list", |
| 259 | "data": [ |
| 260 | {"id": "o1-mini", "object": "model"}, |
| 261 | {"id": "o3-mini", "object": "model"}, |
| 262 | {"id": "openai/gpt-5-mini", "object": "model"}, |
| 263 | {"id": "deepseek-reasoner", "object": "model"}, |
| 264 | { |
| 265 | "id": "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B", |
| 266 | "object": "model", |
| 267 | }, |
| 268 | {"id": "qwq-32b", "object": "model"}, |
| 269 | {"id": "meta-llama/Llama-3.1-8B-Instruct", "object": "model"}, |
| 270 | ], |
| 271 | }; |
| 272 | |
| 273 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 274 | final service = ServerApiService(dio); |
| 275 | |
| 276 | final models = await service.fetchModels(testServer); |
| 277 | |
| 278 | expect(models, hasLength(7)); |
| 279 | |
| 280 | // o1-mini |
| 281 | expect(models[0].supportsReasoning, isTrue); |
| 282 | expect(models[0].supportedReasoningEfforts, ['low', 'medium', 'high']); |
| 283 | expect(models[0].defaultReasoningEffort, 'medium'); |
| 284 | expect(models[0].reasoningMandatory, isFalse); |
| 285 | |
| 286 | // o3-mini |
| 287 | expect(models[1].supportsReasoning, isTrue); |
| 288 | expect(models[1].supportedReasoningEfforts, ['low', 'medium', 'high']); |
| 289 | expect(models[1].defaultReasoningEffort, 'medium'); |
| 290 | |
| 291 | // gpt-5-mini |
| 292 | expect(models[2].supportsReasoning, isTrue); |
| 293 | expect(models[2].supportedReasoningEfforts, ['low', 'medium', 'high']); |
| 294 | |
| 295 | // Provider-specific reasoning models need an explicit capability |
| 296 | // payload so LocalMind does not send unsupported request fields. |
| 297 | expect(models[3].supportsReasoning, isFalse); |
| 298 | expect(models[4].supportsReasoning, isFalse); |
| 299 | expect(models[5].supportsReasoning, isFalse); |
| 300 | |
| 301 | expect(models[6].supportsReasoning, isFalse); |
| 302 | expect(models[6].supportedReasoningEfforts, isNull); |
| 303 | }, |
| 304 | ); |
| 305 | |
| 306 | test('parses LM Studio allowed_options/default/mandatory', () async { |
| 307 | final mockData = { |
| 308 | "models": [ |
| 309 | { |
| 310 | "key": "meta/muse-glimmer", |
| 311 | "display_name": "Muse Glimmer", |
| 312 | "capabilities": { |
| 313 | "reasoning": { |
| 314 | "allowed_options": ["low", "medium", "high", "xhigh"], |
| 315 | "default": "medium", |
| 316 | }, |
| 317 | }, |
| 318 | }, |
| 319 | { |
| 320 | "key": "google/gemma-4-e2b", |
| 321 | "display_name": "Gemma 4 E2B", |
| 322 | "capabilities": { |
| 323 | "reasoning": { |
| 324 | "allowed_options": ["off", "on"], |
| 325 | "default": "on", |
| 326 | }, |
| 327 | }, |
| 328 | }, |
| 329 | { |
| 330 | "key": "deepseek-r1", |
| 331 | "display_name": "DeepSeek R1", |
| 332 | "capabilities": { |
| 333 | "reasoning": { |
| 334 | "allowed_options": ["on"], |
| 335 | "default": "on", |
| 336 | }, |
| 337 | }, |
| 338 | }, |
| 339 | { |
| 340 | "key": "plain-model", |
| 341 | "display_name": "Plain", |
| 342 | "capabilities": {"vision": false}, |
| 343 | }, |
| 344 | ], |
| 345 | }; |
| 346 | |
| 347 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 348 | final service = ServerApiService(dio); |
| 349 | |
| 350 | final models = await service.fetchModels(testServer); |
| 351 | expect(models, hasLength(4)); |
| 352 | final byId = {for (final m in models) m.id: m}; |
| 353 | |
| 354 | final glimmer = byId['meta/muse-glimmer']!; |
| 355 | expect(glimmer.supportsReasoning, isTrue); |
| 356 | expect(glimmer.reasoningMandatory, isTrue); |
| 357 | expect(glimmer.supportedReasoningEfforts, [ |
| 358 | 'low', |
| 359 | 'medium', |
| 360 | 'high', |
| 361 | 'xhigh', |
| 362 | ]); |
| 363 | expect(glimmer.defaultReasoningEffort, 'medium'); |
| 364 | |
| 365 | final gemma = byId['google/gemma-4-e2b']!; |
| 366 | expect(gemma.supportsReasoning, isTrue); |
| 367 | expect(gemma.reasoningMandatory, isFalse); |
| 368 | expect(gemma.supportedReasoningEfforts, ['off', 'on']); |
| 369 | |
| 370 | final r1 = byId['deepseek-r1']!; |
| 371 | expect(r1.supportsReasoning, isTrue); |
| 372 | expect(r1.reasoningMandatory, isTrue); |
| 373 | |
| 374 | expect(byId['plain-model']!.supportsReasoning, isFalse); |
| 375 | expect(byId['plain-model']!.supportedReasoningEfforts, isNull); |
| 376 | }); |
| 377 | |
| 378 | test( |
| 379 | 'handles running models with fallback data array (llama.cpp)', |
| 380 | () async { |
| 381 | final mockData = { |
| 382 | "object": "list", |
| 383 | "data": [ |
| 384 | {"id": "active-model", "object": "model"}, |
| 385 | ], |
| 386 | }; |
| 387 | |
| 388 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 389 | final service = ServerApiService(dio); |
| 390 | |
| 391 | final lmStudioServer = testServer.copyWith(type: ServerType.lmStudio); |
| 392 | final running = await service.fetchRunningModels(lmStudioServer); |
| 393 | |
| 394 | expect(running, contains("active-model")); |
| 395 | }, |
| 396 | ); |
| 397 | |
| 398 | test( |
| 399 | 'handles running models with custom models array and loaded instances (LM Studio)', |
| 400 | () async { |
| 401 | final mockData = { |
| 402 | "models": [ |
| 403 | { |
| 404 | "key": "active-model-lm", |
| 405 | "loaded_instances": [ |
| 406 | {"id": "inst_1"}, |
| 407 | ], |
| 408 | }, |
| 409 | {"key": "inactive-model-lm", "loaded_instances": []}, |
| 410 | ], |
| 411 | }; |
| 412 | |
| 413 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 414 | final service = ServerApiService(dio); |
| 415 | |
| 416 | final lmStudioServer = testServer.copyWith(type: ServerType.lmStudio); |
| 417 | final running = await service.fetchRunningModels(lmStudioServer); |
| 418 | |
| 419 | expect(running, contains("inst_1")); |
| 420 | expect(running, isNot(contains("inactive-model-lm"))); |
| 421 | }, |
| 422 | ); |
| 423 | |
| 424 | test('handles custom string quantization value gracefully', () async { |
| 425 | final mockData = { |
| 426 | "data": [ |
| 427 | {"id": "model-with-string-quant", "quantization": "Q5_K_M"}, |
| 428 | ], |
| 429 | }; |
| 430 | |
| 431 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 432 | final service = ServerApiService(dio); |
| 433 | |
| 434 | final models = await service.fetchModels(testServer); |
| 435 | expect(models.first.quantization, "Q5_K_M"); |
| 436 | }); |
| 437 | |
| 438 | test( |
| 439 | 'handles actual user llama.cpp endpoint output successfully with Map architecture', |
| 440 | () async { |
| 441 | final mockData = { |
| 442 | "data": [ |
| 443 | { |
| 444 | "id": "Gemma-4-31B:IQ4_XS", |
| 445 | "aliases": [], |
| 446 | "tags": [], |
| 447 | "object": "model", |
| 448 | "owned_by": "llamacpp", |
| 449 | "created": 1779851034, |
| 450 | "status": { |
| 451 | "value": "unloaded", |
| 452 | "args": [ |
| 453 | "D:\\llamacpp\\llama-server.exe", |
| 454 | "--host", |
| 455 | "127.0.0.1", |
| 456 | "--jinja", |
| 457 | "--mlock", |
| 458 | "--no-mmap", |
| 459 | "--port", |
| 460 | "0", |
| 461 | "--alias", |
| 462 | "Gemma-4-31B:IQ4_XS", |
| 463 | "--ctx-size", |
| 464 | "65536", |
| 465 | "--cache-type-k", |
| 466 | "q4_0", |
| 467 | "--cache-type-v", |
| 468 | "q4_0", |
| 469 | "--flash-attn", |
| 470 | "true", |
| 471 | "--model", |
| 472 | "./models/gemma-4-31B-it-llmfan-i1-IQ4_XS.gguf", |
| 473 | "--parallel", |
| 474 | "1", |
| 475 | "--reasoning", |
| 476 | "off", |
| 477 | ], |
| 478 | "preset": |
| 479 | "[Gemma-4-31B:IQ4_XS]\njinja = true\nmlock = 1\nmmap = 0\nctx-size = 65536\ncache-type-k = q4_0\ncache-type-v = q4_0\nflash-attn = true\nmodel = ./models/gemma-4-31B-it-llmfan-i1-IQ4_XS.gguf\nparallel = 1\nreasoning = off\n\n", |
| 480 | }, |
| 481 | "architecture": { |
| 482 | "input_modalities": ["text"], |
| 483 | "output_modalities": ["text"], |
| 484 | }, |
| 485 | }, |
| 486 | ], |
| 487 | "object": "list", |
| 488 | }; |
| 489 | |
| 490 | final dio = Dio()..interceptors.add(TestInterceptor(mockData)); |
| 491 | final service = ServerApiService(dio); |
| 492 | |
| 493 | final models = await service.fetchModels(testServer); |
| 494 | expect(models, hasLength(1)); |
| 495 | expect(models.first.id, "Gemma-4-31B:IQ4_XS"); |
| 496 | expect(models.first.architecture, isNull); |
| 497 | }, |
| 498 | ); |
| 499 | }); |
| 500 | |
| 501 | group('ServerApiService - Ollama load behavior', () { |
| 502 | test( |
| 503 | 'loadModel is a no-op for Ollama (does not POST /api/generate)', |
| 504 | () async { |
| 505 | final ollamaServer = Server( |
| 506 | id: 'test-ollama', |
| 507 | name: 'Test Ollama', |
| 508 | type: ServerType.ollama, |
| 509 | host: 'localhost', |
| 510 | port: 11434, |
| 511 | createdAt: DateTime.now(), |
| 512 | lastConnectedAt: DateTime.now(), |
| 513 | ); |
| 514 | |
| 515 | final dio = Dio()..interceptors.add(RequestRecordingInterceptor()); |
| 516 | final service = ServerApiService(dio); |
| 517 | |
| 518 | await service.loadModel(ollamaServer, 'llama3.2:latest'); |
| 519 | await service.loadModelWithInstanceId(ollamaServer, 'llama3.2:latest'); |
| 520 | |
| 521 | final recorder = |
| 522 | dio.interceptors.firstWhere((i) => i is RequestRecordingInterceptor) |
| 523 | as RequestRecordingInterceptor; |
| 524 | expect( |
| 525 | recorder.requests, |
| 526 | isEmpty, |
| 527 | reason: |
| 528 | 'Ollama should not issue an explicit load request; ' |
| 529 | 'calling /api/generate without a prompt can trigger an ' |
| 530 | 'auto-pull of the model from the Ollama registry.', |
| 531 | ); |
| 532 | }, |
| 533 | ); |
| 534 | |
| 535 | test( |
| 536 | 'unloadModel for Ollama posts keep_alive=0 and a valid payload', |
| 537 | () async { |
| 538 | final ollamaServer = Server( |
| 539 | id: 'test-ollama', |
| 540 | name: 'Test Ollama', |
| 541 | type: ServerType.ollama, |
| 542 | host: 'localhost', |
| 543 | port: 11434, |
| 544 | createdAt: DateTime.now(), |
| 545 | lastConnectedAt: DateTime.now(), |
| 546 | ); |
| 547 | |
| 548 | final dio = Dio()..interceptors.add(RequestRecordingInterceptor()); |
| 549 | final service = ServerApiService(dio); |
| 550 | |
| 551 | await service.unloadModel(ollamaServer, 'llama3.2:latest'); |
| 552 | |
| 553 | final recorder = |
| 554 | dio.interceptors.firstWhere((i) => i is RequestRecordingInterceptor) |
| 555 | as RequestRecordingInterceptor; |
| 556 | expect(recorder.requests, hasLength(1)); |
| 557 | expect(recorder.requests.first.path, endsWith('/api/generate')); |
| 558 | expect(recorder.requests.first.data, containsPair('keep_alive', 0)); |
| 559 | expect( |
| 560 | recorder.requests.first.data, |
| 561 | containsPair('model', 'llama3.2:latest'), |
| 562 | ); |
| 563 | expect(recorder.requests.first.data, containsPair('prompt', '')); |
| 564 | }, |
| 565 | ); |
| 566 | |
| 567 | test( |
| 568 | 'fetchModels enriches Ollama models with vision capability from /api/show', |
| 569 | () async { |
| 570 | final ollamaServer = Server( |
| 571 | id: 'test-ollama', |
| 572 | name: 'Test Ollama', |
| 573 | type: ServerType.ollama, |
| 574 | host: 'localhost', |
| 575 | port: 11434, |
| 576 | createdAt: DateTime.now(), |
| 577 | lastConnectedAt: DateTime.now(), |
| 578 | ); |
| 579 | |
| 580 | final interceptor = RoutingInterceptor({ |
| 581 | 'GET /api/tags': { |
| 582 | 'models': [ |
| 583 | { |
| 584 | 'name': 'qwen3-vl:2b', |
| 585 | 'size': 1800000000, |
| 586 | 'details': { |
| 587 | 'parameter_size': '2.1B', |
| 588 | 'quantization_level': 'Q4_K_M', |
| 589 | 'family': 'qwen3', |
| 590 | }, |
| 591 | }, |
| 592 | { |
| 593 | 'name': 'llama3.2:3b', |
| 594 | 'size': 2000000000, |
| 595 | 'details': { |
| 596 | 'parameter_size': '3B', |
| 597 | 'quantization_level': 'Q4_K_M', |
| 598 | 'family': 'llama', |
| 599 | }, |
| 600 | }, |
| 601 | ], |
| 602 | }, |
| 603 | 'POST /api/show:qwen3-vl:2b': { |
| 604 | 'capabilities': ['completion', 'vision'], |
| 605 | }, |
| 606 | 'POST /api/show:llama3.2:3b': { |
| 607 | 'capabilities': ['completion'], |
| 608 | }, |
| 609 | }); |
| 610 | |
| 611 | final service = ServerApiService(Dio()..interceptors.add(interceptor)); |
| 612 | final models = await service.fetchModels(ollamaServer); |
| 613 | |
| 614 | expect(models, hasLength(2)); |
| 615 | final byId = {for (final m in models) m.id: m}; |
| 616 | expect(byId['qwen3-vl:2b']?.supportsVision, isTrue); |
| 617 | expect(byId['llama3.2:3b']?.supportsVision, isFalse); |
| 618 | }, |
| 619 | ); |
| 620 | |
| 621 | test('fetchModels sends canonical model key to Ollama /api/show', () async { |
| 622 | final ollamaServer = Server( |
| 623 | id: 'test-ollama', |
| 624 | name: 'Test Ollama', |
| 625 | type: ServerType.ollama, |
| 626 | host: 'localhost', |
| 627 | port: 11434, |
| 628 | createdAt: DateTime.now(), |
| 629 | lastConnectedAt: DateTime.now(), |
| 630 | ); |
| 631 | |
| 632 | final showBodies = <dynamic>[]; |
| 633 | final dio = Dio() |
| 634 | ..interceptors.add( |
| 635 | InterceptorsWrapper( |
| 636 | onRequest: (options, handler) { |
| 637 | if (options.path.endsWith('/api/tags')) { |
| 638 | handler.resolve( |
| 639 | Response( |
| 640 | requestOptions: options, |
| 641 | statusCode: 200, |
| 642 | data: { |
| 643 | 'models': [ |
| 644 | { |
| 645 | 'name': 'qwen3.5:9b', |
| 646 | 'size': 6600000000, |
| 647 | 'details': { |
| 648 | 'parameter_size': '9.7B', |
| 649 | 'quantization_level': 'Q4_K_M', |
| 650 | 'family': 'qwen35', |
| 651 | }, |
| 652 | }, |
| 653 | ], |
| 654 | }, |
| 655 | ), |
| 656 | ); |
| 657 | return; |
| 658 | } |
| 659 | if (options.path.endsWith('/api/show')) { |
| 660 | showBodies.add(options.data); |
| 661 | handler.resolve( |
| 662 | Response( |
| 663 | requestOptions: options, |
| 664 | statusCode: 200, |
| 665 | data: { |
| 666 | 'capabilities': ['completion', 'thinking'], |
| 667 | }, |
| 668 | ), |
| 669 | ); |
| 670 | return; |
| 671 | } |
| 672 | handler.next(options); |
| 673 | }, |
| 674 | ), |
| 675 | ); |
| 676 | |
| 677 | final service = ServerApiService(dio); |
| 678 | final models = await service.fetchModels(ollamaServer); |
| 679 | |
| 680 | expect(models, hasLength(1)); |
| 681 | expect(showBodies, hasLength(1)); |
| 682 | final showBody = showBodies.single; |
| 683 | expect(showBody, isA<Map>()); |
| 684 | expect((showBody as Map).containsKey('model'), isTrue); |
| 685 | expect(showBody['model'], 'qwen3.5:9b'); |
| 686 | expect(showBody.containsKey('name'), isFalse); |
| 687 | }); |
| 688 | |
| 689 | test( |
| 690 | 'fetchModels marks Ollama thinking capability with binary off/on', |
| 691 | () async { |
| 692 | final ollamaServer = Server( |
| 693 | id: 'test-ollama', |
| 694 | name: 'Test Ollama', |
| 695 | type: ServerType.ollama, |
| 696 | host: 'localhost', |
| 697 | port: 11434, |
| 698 | createdAt: DateTime.now(), |
| 699 | lastConnectedAt: DateTime.now(), |
| 700 | ); |
| 701 | |
| 702 | final interceptor = RoutingInterceptor({ |
| 703 | 'GET /api/tags': { |
| 704 | 'models': [ |
| 705 | { |
| 706 | 'name': 'qwen3.5:9b', |
| 707 | 'size': 6600000000, |
| 708 | 'details': { |
| 709 | 'parameter_size': '9.7B', |
| 710 | 'quantization_level': 'Q4_K_M', |
| 711 | 'family': 'qwen35', |
| 712 | }, |
| 713 | }, |
| 714 | { |
| 715 | 'name': 'llama3.2:3b', |
| 716 | 'size': 2000000000, |
| 717 | 'details': { |
| 718 | 'parameter_size': '3B', |
| 719 | 'quantization_level': 'Q4_K_M', |
| 720 | 'family': 'llama', |
| 721 | }, |
| 722 | }, |
| 723 | ], |
| 724 | }, |
| 725 | 'POST /api/show:qwen3.5:9b': { |
| 726 | 'capabilities': ['completion', 'tools', 'thinking'], |
| 727 | 'details': { |
| 728 | 'family': 'qwen35', |
| 729 | 'families': ['qwen35'], |
| 730 | }, |
| 731 | }, |
| 732 | 'POST /api/show:llama3.2:3b': { |
| 733 | 'capabilities': ['completion'], |
| 734 | }, |
| 735 | }); |
| 736 | |
| 737 | final service = ServerApiService(Dio()..interceptors.add(interceptor)); |
| 738 | final models = await service.fetchModels(ollamaServer); |
| 739 | |
| 740 | expect(models, hasLength(2)); |
| 741 | final byId = {for (final m in models) m.id: m}; |
| 742 | final thinker = byId['qwen3.5:9b']!; |
| 743 | expect(thinker.supportsReasoning, isTrue); |
| 744 | expect(thinker.supportedReasoningEfforts, ['off', 'on']); |
| 745 | expect(thinker.reasoningMandatory, isFalse); |
| 746 | expect(thinker.supportsToolUse, isTrue); |
| 747 | |
| 748 | final plain = byId['llama3.2:3b']!; |
| 749 | expect(plain.supportsReasoning, isFalse); |
| 750 | expect(plain.supportedReasoningEfforts, isNull); |
| 751 | expect(plain.reasoningMandatory, isFalse); |
| 752 | }, |
| 753 | ); |
| 754 | |
| 755 | test( |
| 756 | 'fetchModels treats Ollama gptoss family as mandatory granular', |
| 757 | () async { |
| 758 | final ollamaServer = Server( |
| 759 | id: 'test-ollama', |
| 760 | name: 'Test Ollama', |
| 761 | type: ServerType.ollama, |
| 762 | host: 'localhost', |
| 763 | port: 11434, |
| 764 | createdAt: DateTime.now(), |
| 765 | lastConnectedAt: DateTime.now(), |
| 766 | ); |
| 767 | |
| 768 | final interceptor = RoutingInterceptor({ |
| 769 | 'GET /api/tags': { |
| 770 | 'models': [ |
| 771 | { |
| 772 | 'name': 'gpt-oss:20b', |
| 773 | 'size': 13700000000, |
| 774 | 'details': { |
| 775 | 'parameter_size': '20.9B', |
| 776 | 'quantization_level': 'MXFP4', |
| 777 | 'family': 'gptoss', |
| 778 | }, |
| 779 | }, |
| 780 | ], |
| 781 | }, |
| 782 | 'POST /api/show:gpt-oss:20b': { |
| 783 | 'capabilities': ['completion'], |
| 784 | 'details': { |
| 785 | 'family': 'gptoss', |
| 786 | 'families': ['gptoss'], |
| 787 | }, |
| 788 | }, |
| 789 | }); |
| 790 | |
| 791 | final service = ServerApiService(Dio()..interceptors.add(interceptor)); |
| 792 | final models = await service.fetchModels(ollamaServer); |
| 793 | |
| 794 | expect(models, hasLength(1)); |
| 795 | final model = models.single; |
| 796 | expect(model.supportsReasoning, isTrue); |
| 797 | expect(model.supportedReasoningEfforts, ['low', 'medium', 'high']); |
| 798 | expect(model.reasoningMandatory, isTrue); |
| 799 | }, |
| 800 | ); |
| 801 | |
| 802 | test( |
| 803 | 'fetchModels treats the Ollama LFM2 thinking renderer as mandatory', |
| 804 | () async { |
| 805 | final ollamaServer = Server( |
| 806 | id: 'test-ollama', |
| 807 | name: 'Test Ollama', |
| 808 | type: ServerType.ollama, |
| 809 | host: 'localhost', |
| 810 | port: 11434, |
| 811 | createdAt: DateTime.now(), |
| 812 | lastConnectedAt: DateTime.now(), |
| 813 | ); |
| 814 | |
| 815 | final interceptor = RoutingInterceptor({ |
| 816 | 'GET /api/tags': { |
| 817 | 'models': [ |
| 818 | { |
| 819 | 'name': 'lfm2.5-thinking:latest', |
| 820 | 'size': 731163903, |
| 821 | 'details': { |
| 822 | 'parameter_size': '1.2B', |
| 823 | 'quantization_level': 'Q4_K_M', |
| 824 | 'family': 'lfm2', |
| 825 | }, |
| 826 | }, |
| 827 | ], |
| 828 | }, |
| 829 | 'POST /api/show:lfm2.5-thinking:latest': { |
| 830 | 'capabilities': ['completion', 'tools', 'thinking'], |
| 831 | 'details': { |
| 832 | 'family': 'lfm2', |
| 833 | 'families': ['lfm2'], |
| 834 | }, |
| 835 | 'modelfile': |
| 836 | 'TEMPLATE {{ .Prompt }}\n' |
| 837 | 'RENDERER lfm2-thinking\n' |
| 838 | 'PARSER lfm2-thinking', |
| 839 | }, |
| 840 | }); |
| 841 | |
| 842 | final service = ServerApiService(Dio()..interceptors.add(interceptor)); |
| 843 | final model = (await service.fetchModels(ollamaServer)).single; |
| 844 | |
| 845 | expect(model.supportsReasoning, isTrue); |
| 846 | expect(model.supportedReasoningEfforts, ['on']); |
| 847 | expect(model.reasoningMandatory, isTrue); |
| 848 | expect(model.supportsToolUse, isTrue); |
| 849 | }, |
| 850 | ); |
| 851 | |
| 852 | test( |
| 853 | 'fetchModels recognizes an older LFM2 thinking response by model id', |
| 854 | () async { |
| 855 | final ollamaServer = Server( |
| 856 | id: 'test-ollama', |
| 857 | name: 'Test Ollama', |
| 858 | type: ServerType.ollama, |
| 859 | host: 'localhost', |
| 860 | port: 11434, |
| 861 | createdAt: DateTime.now(), |
| 862 | lastConnectedAt: DateTime.now(), |
| 863 | ); |
| 864 | final interceptor = RoutingInterceptor({ |
| 865 | 'GET /api/tags': { |
| 866 | 'models': [ |
| 867 | { |
| 868 | 'name': 'lfm2.5-thinking:latest', |
| 869 | 'details': {'family': 'lfm2'}, |
| 870 | }, |
| 871 | ], |
| 872 | }, |
| 873 | 'POST /api/show:lfm2.5-thinking:latest': { |
| 874 | 'capabilities': ['completion'], |
| 875 | 'details': {'family': 'lfm2'}, |
| 876 | }, |
| 877 | }); |
| 878 | |
| 879 | final service = ServerApiService(Dio()..interceptors.add(interceptor)); |
| 880 | final model = (await service.fetchModels(ollamaServer)).single; |
| 881 | |
| 882 | expect(model.supportedReasoningEfforts, ['on']); |
| 883 | expect(model.reasoningMandatory, isTrue); |
| 884 | }, |
| 885 | ); |
| 886 | |
| 887 | test( |
| 888 | 'fetchModels handles absent or malformed Ollama capabilities', |
| 889 | () async { |
| 890 | final ollamaServer = Server( |
| 891 | id: 'test-ollama', |
| 892 | name: 'Test Ollama', |
| 893 | type: ServerType.ollama, |
| 894 | host: 'localhost', |
| 895 | port: 11434, |
| 896 | createdAt: DateTime.now(), |
| 897 | lastConnectedAt: DateTime.now(), |
| 898 | ); |
| 899 | |
| 900 | final interceptor = RoutingInterceptor({ |
| 901 | 'GET /api/tags': { |
| 902 | 'models': [ |
| 903 | {'name': 'missing-caps:latest', 'size': 1000, 'details': {}}, |
| 904 | {'name': 'malformed-caps:latest', 'size': 1000, 'details': {}}, |
| 905 | {'name': 'failing-caps:latest', 'size': 1000, 'details': {}}, |
| 906 | ], |
| 907 | }, |
| 908 | 'POST /api/show:missing-caps:latest': {'details': {}}, |
| 909 | 'POST /api/show:malformed-caps:latest': {'capabilities': 'thinking'}, |
| 910 | // No mock for failing-caps:latest — enrichment must fall back to the |
| 911 | // base model instead of failing the whole fetch. |
| 912 | }); |
| 913 | |
| 914 | final service = ServerApiService(Dio()..interceptors.add(interceptor)); |
| 915 | final models = await service.fetchModels(ollamaServer); |
| 916 | |
| 917 | expect(models, hasLength(3)); |
| 918 | for (final model in models) { |
| 919 | expect(model.supportsReasoning, isFalse); |
| 920 | expect(model.supportedReasoningEfforts, isNull); |
| 921 | expect(model.reasoningMandatory, isFalse); |
| 922 | } |
| 923 | }, |
| 924 | ); |
| 925 | |
| 926 | test('unloadModel for Ollama surfaces failures', () async { |
| 927 | final ollamaServer = Server( |
| 928 | id: 'test-ollama', |
| 929 | name: 'Test Ollama', |
| 930 | type: ServerType.ollama, |
| 931 | host: 'localhost', |
| 932 | port: 11434, |
| 933 | createdAt: DateTime.now(), |
| 934 | lastConnectedAt: DateTime.now(), |
| 935 | ); |
| 936 | |
| 937 | final dio = Dio()..interceptors.add(FailingInterceptor()); |
| 938 | final service = ServerApiService(dio); |
| 939 | |
| 940 | expect( |
| 941 | () => service.unloadModel(ollamaServer, 'llama3.2:latest'), |
| 942 | throwsA(isA<Exception>()), |
| 943 | ); |
| 944 | }); |
| 945 | }); |
| 946 | |
| 947 | group('ServerApiService - OpenRouter model parsing', () { |
| 948 | late Server openRouterServer; |
| 949 | |
| 950 | setUp(() { |
| 951 | openRouterServer = Server( |
| 952 | id: 'test-openrouter', |
| 953 | name: 'Test OpenRouter', |
| 954 | type: ServerType.openRouter, |
| 955 | host: 'openrouter.ai', |
| 956 | port: 443, |
| 957 | apiKey: 'test-key', |
| 958 | createdAt: DateTime.now(), |
| 959 | lastConnectedAt: DateTime.now(), |
| 960 | ); |
| 961 | }); |
| 962 | |
| 963 | test( |
| 964 | 'parses capabilities and per-1M pricing for OpenRouter models', |
| 965 | () async { |
| 966 | final data = { |
| 967 | 'data': [ |
| 968 | { |
| 969 | 'id': 'openai/gpt-4o', |
| 970 | 'name': 'OpenAI: GPT-4o', |
| 971 | 'created': 1715558400, |
| 972 | 'context_length': 128000, |
| 973 | 'architecture': { |
| 974 | 'modality': 'text+image+file->text', |
| 975 | 'input_modalities': ['text', 'image', 'file'], |
| 976 | 'output_modalities': ['text'], |
| 977 | 'tokenizer': 'GPT', |
| 978 | 'instruct_type': null, |
| 979 | }, |
| 980 | 'pricing': {'prompt': '0.0000025', 'completion': '0.00001'}, |
| 981 | }, |
| 982 | { |
| 983 | 'id': 'deepseek/deepseek-reasoner', |
| 984 | 'name': 'DeepSeek: R1', |
| 985 | 'context_length': 163840, |
| 986 | 'architecture': { |
| 987 | 'modality': 'text->text', |
| 988 | 'input_modalities': ['text'], |
| 989 | 'output_modalities': ['text'], |
| 990 | 'tokenizer': 'DeepSeek', |
| 991 | }, |
| 992 | 'reasoning': { |
| 993 | 'mandatory': false, |
| 994 | 'default_effort': 'high', |
| 995 | 'supported_efforts': ['max', 'high', 'low'], |
| 996 | }, |
| 997 | 'supported_parameters': ['tools', 'reasoning', 'temperature'], |
| 998 | }, |
| 999 | { |
| 1000 | 'id': 'meta-llama/llama-3.1-8b-instruct:free', |
| 1001 | 'name': 'Meta: Llama 3.1 8B (Free)', |
| 1002 | 'context_length': 131072, |
| 1003 | 'architecture': { |
| 1004 | 'modality': 'text->text', |
| 1005 | 'input_modalities': ['text'], |
| 1006 | 'output_modalities': ['text'], |
| 1007 | 'tokenizer': 'Llama3', |
| 1008 | }, |
| 1009 | 'pricing': {'prompt': '0', 'completion': '0'}, |
| 1010 | }, |
| 1011 | ], |
| 1012 | }; |
| 1013 | |
| 1014 | final service = ServerApiService( |
| 1015 | Dio()..interceptors.add(TestInterceptor(data)), |
| 1016 | ); |
| 1017 | final models = await service.fetchModels(openRouterServer); |
| 1018 | |
| 1019 | expect(models, hasLength(3)); |
| 1020 | final byId = {for (final m in models) m.id: m}; |
| 1021 | |
| 1022 | final gpt4o = byId['openai/gpt-4o']!; |
| 1023 | expect(gpt4o.supportsVision, isTrue); |
| 1024 | expect(gpt4o.inputPricePerMillion, closeTo(2.5, 0.0001)); |
| 1025 | expect(gpt4o.outputPricePerMillion, closeTo(10.0, 0.0001)); |
| 1026 | expect(gpt4o.pricingLabel, '\$2.50/\$10.00'); |
| 1027 | |
| 1028 | final reasoner = byId['deepseek/deepseek-reasoner']!; |
| 1029 | expect(reasoner.supportsReasoning, isTrue); |
| 1030 | expect(reasoner.supportsToolUse, isTrue); |
| 1031 | expect(reasoner.supportsVision, isFalse); |
| 1032 | expect(reasoner.reasoningMandatory, isFalse); |
| 1033 | expect(reasoner.defaultReasoningEffort, 'high'); |
| 1034 | expect(reasoner.supportedReasoningEfforts, ['max', 'high', 'low']); |
| 1035 | |
| 1036 | final free = byId['meta-llama/llama-3.1-8b-instruct:free']!; |
| 1037 | expect(free.isPricingFree, isTrue); |
| 1038 | expect(free.pricingLabel, 'Free'); |
| 1039 | }, |
| 1040 | ); |
| 1041 | |
| 1042 | test( |
| 1043 | 'detects vision via modality when input_modalities is absent', |
| 1044 | () async { |
| 1045 | final data = { |
| 1046 | 'data': [ |
| 1047 | { |
| 1048 | 'id': 'google/gemini-pro', |
| 1049 | 'name': 'Google: Gemini Pro', |
| 1050 | 'context_length': 32768, |
| 1051 | 'architecture': { |
| 1052 | 'modality': 'text+image->text', |
| 1053 | 'tokenizer': 'Gemini', |
| 1054 | }, |
| 1055 | }, |
| 1056 | ], |
| 1057 | }; |
| 1058 | |
| 1059 | final service = ServerApiService( |
| 1060 | Dio()..interceptors.add(TestInterceptor(data)), |
| 1061 | ); |
| 1062 | final models = await service.fetchModels(openRouterServer); |
| 1063 | |
| 1064 | expect(models, hasLength(1)); |
| 1065 | expect(models.single.supportsVision, isTrue); |
| 1066 | expect(models.single.supportsReasoning, isFalse); |
| 1067 | expect(models.single.inputPricePerMillion, isNull); |
| 1068 | expect(models.single.pricingLabel, isNull); |
| 1069 | }, |
| 1070 | ); |
| 1071 | |
| 1072 | test('parses mandatory reasoning with a single supported effort', () async { |
| 1073 | final data = { |
| 1074 | 'data': [ |
| 1075 | { |
| 1076 | 'id': 'openai/gpt-5-pro', |
| 1077 | 'name': 'OpenAI: GPT-5 Pro', |
| 1078 | 'context_length': 200000, |
| 1079 | 'architecture': {'modality': 'text->text', 'tokenizer': 'GPT'}, |
| 1080 | 'reasoning': { |
| 1081 | 'mandatory': true, |
| 1082 | 'default_effort': 'high', |
| 1083 | 'supported_efforts': ['high'], |
| 1084 | }, |
| 1085 | }, |
| 1086 | ], |
| 1087 | }; |
| 1088 | |
| 1089 | final service = ServerApiService( |
| 1090 | Dio()..interceptors.add(TestInterceptor(data)), |
| 1091 | ); |
| 1092 | final models = await service.fetchModels(openRouterServer); |
| 1093 | |
| 1094 | expect(models, hasLength(1)); |
| 1095 | final model = models.single; |
| 1096 | expect(model.supportsReasoning, isTrue); |
| 1097 | expect(model.reasoningMandatory, isTrue); |
| 1098 | expect(model.supportedReasoningEfforts, ['high']); |
| 1099 | expect(model.defaultReasoningEffort, 'high'); |
| 1100 | }); |
| 1101 | |
| 1102 | test('treats non-finite pricing values as unknown', () async { |
| 1103 | final data = { |
| 1104 | 'data': [ |
| 1105 | { |
| 1106 | 'id': 'weird/model', |
| 1107 | 'name': 'Weird Model', |
| 1108 | 'context_length': 8192, |
| 1109 | 'architecture': {'modality': 'text->text', 'tokenizer': 'X'}, |
| 1110 | 'pricing': {'prompt': '1e309', 'completion': 'Infinity'}, |
| 1111 | }, |
| 1112 | ], |
| 1113 | }; |
| 1114 | |
| 1115 | final service = ServerApiService( |
| 1116 | Dio()..interceptors.add(TestInterceptor(data)), |
| 1117 | ); |
| 1118 | final models = await service.fetchModels(openRouterServer); |
| 1119 | |
| 1120 | expect(models, hasLength(1)); |
| 1121 | expect(models.single.inputPricePerMillion, isNull); |
| 1122 | expect(models.single.outputPricePerMillion, isNull); |
| 1123 | expect(models.single.pricingLabel, isNull); |
| 1124 | }); |
| 1125 | }); |
| 1126 | |
| 1127 | group('ServerApiService - Requesty model parsing', () { |
| 1128 | late Server requestyServer; |
| 1129 | |
| 1130 | setUp(() { |
| 1131 | requestyServer = Server( |
| 1132 | id: 'test-requesty', |
| 1133 | name: 'Test Requesty', |
| 1134 | type: ServerType.requesty, |
| 1135 | host: '', |
| 1136 | port: 443, |
| 1137 | apiKey: 'test-key', |
| 1138 | createdAt: DateTime.now(), |
| 1139 | lastConnectedAt: DateTime.now(), |
| 1140 | ); |
| 1141 | }); |
| 1142 | |
| 1143 | test('uses the Requesty router endpoints', () { |
| 1144 | expect(requestyServer.baseUrl, 'https://router.requesty.ai/v1'); |
| 1145 | expect( |
| 1146 | requestyServer.chatEndpoint, |
| 1147 | 'https://router.requesty.ai/v1/chat/completions', |
| 1148 | ); |
| 1149 | expect( |
| 1150 | requestyServer.modelsEndpoint, |
| 1151 | 'https://router.requesty.ai/v1/models', |
| 1152 | ); |
| 1153 | expect(requestyServer.displayAddress, 'router.requesty.ai'); |
| 1154 | }); |
| 1155 | |
| 1156 | test( |
| 1157 | 'lists managed policies first, then the catalog without duplicates', |
| 1158 | () async { |
| 1159 | final service = ServerApiService( |
| 1160 | Dio() |
| 1161 | ..interceptors.add( |
| 1162 | RoutingInterceptor({ |
| 1163 | 'GET https://router.requesty.ai/v1/models': { |
| 1164 | 'object': 'list', |
| 1165 | 'data': [ |
| 1166 | { |
| 1167 | 'id': 'openai/gpt-4o-mini', |
| 1168 | 'api': 'chat', |
| 1169 | 'context_window': 128000, |
| 1170 | 'input_price': 1.5e-07, |
| 1171 | 'output_price': 6e-07, |
| 1172 | 'supports_vision': true, |
| 1173 | 'supports_tool_calling': true, |
| 1174 | 'supports_reasoning': false, |
| 1175 | }, |
| 1176 | { |
| 1177 | 'id': 'claude-sonnet-4-5', |
| 1178 | 'api': 'chat', |
| 1179 | 'context_window': 200000, |
| 1180 | }, |
| 1181 | {'id': 'openai/text-embedding-3-small', 'api': 'embedding'}, |
| 1182 | ], |
| 1183 | }, |
| 1184 | 'GET https://router.requesty.ai/v1/models/managed': { |
| 1185 | 'object': 'list', |
| 1186 | 'data': [ |
| 1187 | { |
| 1188 | 'id': 'claude-sonnet-4-5', |
| 1189 | 'name': 'Claude Sonnet 4.5', |
| 1190 | 'api': 'chat', |
| 1191 | 'context_window': 1000000, |
| 1192 | 'input_price': 3e-06, |
| 1193 | 'output_price': 1.5e-05, |
| 1194 | 'supports_reasoning': true, |
| 1195 | 'supports_tool_calling': true, |
| 1196 | }, |
| 1197 | ], |
| 1198 | }, |
| 1199 | }), |
| 1200 | ), |
| 1201 | ); |
| 1202 | final models = await service.fetchModels(requestyServer); |
| 1203 | |
| 1204 | expect(models.map((m) => m.id), [ |
| 1205 | 'claude-sonnet-4-5', |
| 1206 | 'openai/gpt-4o-mini', |
| 1207 | ]); |
| 1208 | |
| 1209 | final managed = models.first; |
| 1210 | expect(managed.name, 'Claude Sonnet 4.5'); |
| 1211 | expect(managed.contextLength, 1000000); |
| 1212 | expect(managed.supportsReasoning, isTrue); |
| 1213 | expect(managed.supportsToolUse, isTrue); |
| 1214 | expect(managed.inputPricePerMillion, closeTo(3.0, 0.0001)); |
| 1215 | expect(managed.outputPricePerMillion, closeTo(15.0, 0.0001)); |
| 1216 | |
| 1217 | final mini = models.last; |
| 1218 | expect(mini.name, 'openai/gpt-4o-mini'); |
| 1219 | expect(mini.serverType, ServerType.requesty); |
| 1220 | expect(mini.contextLength, 128000); |
| 1221 | expect(mini.supportsVision, isTrue); |
| 1222 | expect(mini.supportsReasoning, isFalse); |
| 1223 | expect(mini.pricingLabel, '\$0.15/\$0.60'); |
| 1224 | }, |
| 1225 | ); |
| 1226 | |
| 1227 | test('falls back to the catalog when managed models fail', () async { |
| 1228 | final service = ServerApiService( |
| 1229 | Dio() |
| 1230 | ..interceptors.add( |
| 1231 | RoutingInterceptor({ |
| 1232 | 'GET https://router.requesty.ai/v1/models': { |
| 1233 | 'object': 'list', |
| 1234 | 'data': [ |
| 1235 | {'id': 'openai/gpt-4o-mini', 'api': 'chat'}, |
| 1236 | ], |
| 1237 | }, |
| 1238 | }), |
| 1239 | ), |
| 1240 | ); |
| 1241 | final models = await service.fetchModels(requestyServer); |
| 1242 | |
| 1243 | expect(models.map((m) => m.id), ['openai/gpt-4o-mini']); |
| 1244 | }); |
| 1245 | }); |
| 1246 | |
| 1247 | group('ServerApiService - LM Studio load model behavior', () { |
| 1248 | late Server lmStudioServer; |
| 1249 | |
| 1250 | setUp(() { |
| 1251 | lmStudioServer = Server( |
| 1252 | id: 'test-lmstudio', |
| 1253 | name: 'Test LM Studio', |
| 1254 | type: ServerType.lmStudio, |
| 1255 | host: 'localhost', |
| 1256 | port: 1234, |
| 1257 | createdAt: DateTime.now(), |
| 1258 | lastConnectedAt: DateTime.now(), |
| 1259 | ); |
| 1260 | }); |
| 1261 | |
| 1262 | test( |
| 1263 | 'loadModelWithInstanceId succeeds with LM Studio response having "type": "llm"', |
| 1264 | () async { |
| 1265 | final successData = { |
| 1266 | 'type': 'llm', |
| 1267 | 'instance_id': 'liquid/lfm2-1.2b', |
| 1268 | 'load_config': {'context_length': 4096}, |
| 1269 | }; |
| 1270 | |
| 1271 | final dio = Dio()..interceptors.add(TestInterceptor(successData)); |
| 1272 | final service = ServerApiService(dio); |
| 1273 | |
| 1274 | final instanceId = await service.loadModelWithInstanceId( |
| 1275 | lmStudioServer, |
| 1276 | 'liquid/lfm2-1.2b', |
| 1277 | contextLength: 4096, |
| 1278 | ); |
| 1279 | |
| 1280 | expect(instanceId, 'liquid/lfm2-1.2b'); |
| 1281 | }, |
| 1282 | ); |
| 1283 | |
| 1284 | test( |
| 1285 | 'loadModelWithInstanceId throws when response contains error object', |
| 1286 | () async { |
| 1287 | final errorData = { |
| 1288 | 'error': { |
| 1289 | 'type': 'invalid_request_error', |
| 1290 | 'message': 'Context length exceeds model limit', |
| 1291 | }, |
| 1292 | }; |
| 1293 | |
| 1294 | final dio = Dio()..interceptors.add(TestInterceptor(errorData)); |
| 1295 | final service = ServerApiService(dio); |
| 1296 | |
| 1297 | expect( |
| 1298 | () => service.loadModelWithInstanceId( |
| 1299 | lmStudioServer, |
| 1300 | 'liquid/lfm2-1.2b', |
| 1301 | contextLength: 999999, |
| 1302 | ), |
| 1303 | throwsA( |
| 1304 | predicate( |
| 1305 | (e) => |
| 1306 | e is Exception && |
| 1307 | e.toString().contains( |
| 1308 | 'invalid_request_error: Context length exceeds model limit', |
| 1309 | ), |
| 1310 | ), |
| 1311 | ), |
| 1312 | ); |
| 1313 | }, |
| 1314 | ); |
| 1315 | |
| 1316 | test( |
| 1317 | 'loadModelWithInstanceId throws when response contains status error', |
| 1318 | () async { |
| 1319 | final errorData = { |
| 1320 | 'status': 'error', |
| 1321 | 'message': 'Failed to allocate memory', |
| 1322 | }; |
| 1323 | |
| 1324 | final dio = Dio()..interceptors.add(TestInterceptor(errorData)); |
| 1325 | final service = ServerApiService(dio); |
| 1326 | |
| 1327 | expect( |
| 1328 | () => service.loadModelWithInstanceId( |
| 1329 | lmStudioServer, |
| 1330 | 'liquid/lfm2-1.2b', |
| 1331 | ), |
| 1332 | throwsA( |
| 1333 | predicate( |
| 1334 | (e) => |
| 1335 | e is Exception && |
| 1336 | e.toString().contains('Failed to allocate memory'), |
| 1337 | ), |
| 1338 | ), |
| 1339 | ); |
| 1340 | }, |
| 1341 | ); |
| 1342 | }); |
| 1343 | } |