@ondeinference / onde
Code Commits Issues Pulls Cloud Agent CI/CD
copilot/assess-inference-engine-complexity
copilot/assess-inference-engine-complexity copilot/fix-onde-npm-sdk-v1-1-2-release development feature/diffusion-vision main
dart 439 lines 13.8 KB
Raw
1 // This file is automatically generated, so please do not edit it.
2 // @generated by `flutter_rust_bridge`@ 2.12.0.
3
4 // ignore_for_file: type=lint, unused_import, unnecessary_import
5
6 // ignore_for_file: invalid_use_of_internal_member
7
8 import 'frb_generated.dart';
9 import 'package:flutter_rust_bridge/flutter_rust_bridge_for_generated.dart';
10 import 'package:freezed_annotation/freezed_annotation.dart' hide protected;
11 part 'api.freezed.dart';
12
13 // These function are ignored because they are on traits that is not defined in current crate (put an empty `#[frb]` on it to unignore): `assert_receiver_is_total_eq`, `assert_receiver_is_total_eq`, `clone`, `clone`, `clone`, `clone`, `clone`, `clone`, `clone`, `clone`, `eq`, `eq`, `fmt`, `fmt`, `fmt`, `fmt`, `fmt`, `fmt`, `fmt`, `fmt`, `fmt`, `fmt`, `from`, `from`, `from`, `from`, `from`, `from`, `from`, `from`, `from`, `from`, `from`, `from`, `from`
14
15 /// Return the platform-appropriate default `GgufModelConfig`.
16 ///
17 /// Selects the 1.5B model on iOS / tvOS / Android and the 3B model on
18 /// macOS / Windows / Linux.
19 GgufModelConfig defaultModelConfig() =>
20 RustLib.instance.api.crateApiDefaultModelConfig();
21
22 /// `GgufModelConfig` for Qwen 2.5 1.5B Instruct Q4_K_M (~941 MB).
23 GgufModelConfig qwen2515BConfig() =>
24 RustLib.instance.api.crateApiQwen2515BConfig();
25
26 /// `GgufModelConfig` for Qwen 2.5 3B Instruct Q4_K_M (~1.93 GB).
27 GgufModelConfig qwen253BConfig() =>
28 RustLib.instance.api.crateApiQwen253BConfig();
29
30 /// `GgufModelConfig` for Qwen 2.5 Coder 1.5B Instruct Q4_K_M (~941 MB).
31 GgufModelConfig qwen25Coder15BConfig() =>
32 RustLib.instance.api.crateApiQwen25Coder15BConfig();
33
34 /// `GgufModelConfig` for Qwen 2.5 Coder 3B Instruct Q4_K_M (~1.93 GB).
35 GgufModelConfig qwen25Coder3BConfig() =>
36 RustLib.instance.api.crateApiQwen25Coder3BConfig();
37
38 /// Default sampling config: `temperature=0.7`, `top_p=0.95`, `max_tokens=512`.
39 SamplingConfig defaultSamplingConfig() =>
40 RustLib.instance.api.crateApiDefaultSamplingConfig();
41
42 /// Deterministic sampling config: `temperature=0.0`, greedy, `max_tokens=512`.
43 SamplingConfig deterministicSamplingConfig() =>
44 RustLib.instance.api.crateApiDeterministicSamplingConfig();
45
46 /// Mobile sampling config: `temperature=0.7`, `top_p=0.95`, `max_tokens=128`.
47 SamplingConfig mobileSamplingConfig() =>
48 RustLib.instance.api.crateApiMobileSamplingConfig();
49
50 /// Seed the HuggingFace cache environment for sandboxed platforms.
51 ///
52 /// On iOS (and Android), the default `~/.cache/huggingface/hub` path is outside
53 /// the app sandbox and writes fail with `Operation not permitted (os error 1)`.
54 ///
55 /// Call this **once** at startup — before any `load_gguf_model` or
56 /// `download_model` — with the app's writable data directory.
57 ///
58 /// # Dart usage
59 ///
60 /// ```dart
61 /// import 'package:path_provider/path_provider.dart';
62 ///
63 /// final dir = await getApplicationSupportDirectory();
64 /// configureCacheDir(appDataDir: dir.path);
65 /// ```
66 ///
67 /// Internally this:
68 /// 1. Creates `<app_data_dir>/huggingface/hub/` if it doesn't exist.
69 /// 2. Sets the `HF_HOME` and `HF_HUB_CACHE` environment variables.
70 /// 3. Seeds `mistralrs_core::GLOBAL_HF_CACHE` so the model loader never
71 /// falls back to `Cache::default()`.
72 void configureCacheDir({required String appDataDir}) =>
73 RustLib.instance.api.crateApiConfigureCacheDir(appDataDir: appDataDir);
74
75 // Rust type: RustOpaqueMoi<flutter_rust_bridge::for_generated::RustAutoOpaqueInner<OndeChatEngine>>
76 abstract class OndeChatEngine implements RustOpaqueInterface {
77 /// Clear the conversation history while keeping the model loaded.
78 ///
79 /// Returns the number of turns that were removed.
80 Future<BigInt> clearHistory();
81
82 /// Clear the current system prompt.
83 Future<void> clearSystemPrompt();
84
85 /// One-shot generation from an explicit message list.
86 ///
87 /// Does **not** modify the conversation history.
88 Future<InferenceResult> generate({
89 required List<ChatMessage> messages,
90 SamplingConfig? sampling,
91 });
92
93 /// Returns a clone of the full conversation history.
94 Future<List<ChatMessage>> history();
95
96 /// Returns a point-in-time snapshot of the engine's state.
97 Future<EngineInfo> info();
98
99 /// Returns `true` if a model is currently loaded and ready.
100 Future<bool> isLoaded();
101
102 /// Load the model assigned to this app via the Onde dashboard.
103 ///
104 /// The [appId] and [appSecret] identify your app on
105 /// [ondeinference.com](https://ondeinference.com). If no model is
106 /// assigned, the platform default is loaded instead.
107 ///
108 /// Returns the wall-clock seconds taken to load the model.
109 Future<double> loadAssignedModel({
110 required String appId,
111 required String appSecret,
112 String? systemPrompt,
113 SamplingConfig? sampling,
114 });
115
116 /// Load a GGUF model into the engine.
117 ///
118 /// Returns the load duration in seconds on success.
119 Future<double> loadGgufModel({
120 required GgufModelConfig config,
121 String? systemPrompt,
122 SamplingConfig? sampling,
123 });
124
125 /// Create a new engine with no model loaded.
126 factory OndeChatEngine() => RustLib.instance.api.crateApiOndeChatEngineNew();
127
128 /// Append a message to history without running inference.
129 ///
130 /// Useful for restoring a saved conversation or injecting context.
131 Future<void> pushHistory({required ChatMessage message});
132
133 /// Send a user message and receive a complete assistant reply.
134 ///
135 /// Both the user message and assistant reply are appended to the
136 /// conversation history on success.
137 Future<InferenceResult> sendMessage({required String message});
138
139 /// Replace the sampling configuration used for all subsequent inference.
140 Future<void> setSampling({required SamplingConfig sampling});
141
142 /// Set or replace the system prompt.
143 ///
144 /// The prompt is prepended to every inference request and is **not**
145 /// stored in the conversation history.
146 Future<void> setSystemPrompt({required String prompt});
147
148 /// Stream a reply token-by-token.
149 ///
150 /// Each [`StreamChunk`] carries a `delta` string. The final chunk has
151 /// `done == true`. The `sink` is closed automatically when the stream
152 /// ends or if the receiver is dropped.
153 Stream<StreamChunk> streamMessage({required String message});
154
155 /// Unload the current model and release its memory.
156 ///
157 /// Returns the display name of the model that was unloaded, or `None`
158 /// if no model was loaded.
159 Future<String?> unloadModel();
160 }
161
162 /// A single message in a conversation.
163 class ChatMessage {
164 final ChatRole role;
165 final String content;
166
167 const ChatMessage({required this.role, required this.content});
168
169 @override
170 int get hashCode => role.hashCode ^ content.hashCode;
171
172 @override
173 bool operator ==(Object other) =>
174 identical(this, other) ||
175 other is ChatMessage &&
176 runtimeType == other.runtimeType &&
177 role == other.role &&
178 content == other.content;
179 }
180
181 /// Role of a participant in a chat conversation.
182 enum ChatRole { system, user, assistant }
183
184 /// A point-in-time snapshot of the engine's state.
185 class EngineInfo {
186 final EngineStatus status;
187
188 /// Display name of the currently loaded model, if any.
189 final String? modelName;
190
191 /// Approximate memory footprint of the loaded model, if any.
192 final String? approxMemory;
193
194 /// Number of conversation turns in the current history.
195 final BigInt historyLength;
196
197 const EngineInfo({
198 required this.status,
199 this.modelName,
200 this.approxMemory,
201 required this.historyLength,
202 });
203
204 @override
205 int get hashCode =>
206 status.hashCode ^
207 modelName.hashCode ^
208 approxMemory.hashCode ^
209 historyLength.hashCode;
210
211 @override
212 bool operator ==(Object other) =>
213 identical(this, other) ||
214 other is EngineInfo &&
215 runtimeType == other.runtimeType &&
216 status == other.status &&
217 modelName == other.modelName &&
218 approxMemory == other.approxMemory &&
219 historyLength == other.historyLength;
220 }
221
222 /// Lifecycle status of the inference engine.
223 enum EngineStatus { unloaded, loading, ready, generating, error }
224
225 /// Configuration for loading a pre-quantised GGUF model.
226 class GgufModelConfig {
227 /// HuggingFace repository ID, e.g. `"bartowski/Qwen2.5-1.5B-Instruct-GGUF"`.
228 final String modelId;
229
230 /// GGUF filename(s) within the repository.
231 final List<String> files;
232
233 /// Optional explicit tokenizer model ID (required on Android).
234 final String? tokModelId;
235
236 /// Human-friendly display name, e.g. `"Qwen 2.5 1.5B"`.
237 final String displayName;
238
239 /// Approximate memory footprint, e.g. `"~941 MB (GGUF Q4_K_M)"`.
240 final String approxMemory;
241
242 /// Optional chat template override (Jinja format).
243 final String? chatTemplate;
244
245 const GgufModelConfig({
246 required this.modelId,
247 required this.files,
248 this.tokModelId,
249 required this.displayName,
250 required this.approxMemory,
251 this.chatTemplate,
252 });
253
254 @override
255 int get hashCode =>
256 modelId.hashCode ^
257 files.hashCode ^
258 tokModelId.hashCode ^
259 displayName.hashCode ^
260 approxMemory.hashCode ^
261 chatTemplate.hashCode;
262
263 @override
264 bool operator ==(Object other) =>
265 identical(this, other) ||
266 other is GgufModelConfig &&
267 runtimeType == other.runtimeType &&
268 modelId == other.modelId &&
269 files == other.files &&
270 tokModelId == other.tokModelId &&
271 displayName == other.displayName &&
272 approxMemory == other.approxMemory &&
273 chatTemplate == other.chatTemplate;
274 }
275
276 /// The result of a completed inference request.
277 class InferenceResult {
278 /// The generated assistant reply text.
279 final String text;
280
281 /// Wall-clock inference duration in seconds.
282 final double durationSecs;
283
284 /// Human-readable duration string, e.g. `"4.5s"` or `"2m 3.1s"`.
285 final String durationDisplay;
286
287 /// Finish reason reported by the model, e.g. `"stop"` or `"length"`.
288 final String finishReason;
289
290 /// Tool calls requested by the model (empty if none).
291 final List<ToolCallInfo> toolCalls;
292
293 const InferenceResult({
294 required this.text,
295 required this.durationSecs,
296 required this.durationDisplay,
297 required this.finishReason,
298 required this.toolCalls,
299 });
300
301 @override
302 int get hashCode =>
303 text.hashCode ^
304 durationSecs.hashCode ^
305 durationDisplay.hashCode ^
306 finishReason.hashCode ^
307 toolCalls.hashCode;
308
309 @override
310 bool operator ==(Object other) =>
311 identical(this, other) ||
312 other is InferenceResult &&
313 runtimeType == other.runtimeType &&
314 text == other.text &&
315 durationSecs == other.durationSecs &&
316 durationDisplay == other.durationDisplay &&
317 finishReason == other.finishReason &&
318 toolCalls == other.toolCalls;
319 }
320
321 @freezed
322 sealed class OndeError with _$OndeError implements FrbException {
323 const OndeError._();
324
325 const factory OndeError.noModelLoaded() = OndeError_NoModelLoaded;
326 const factory OndeError.alreadyLoaded({required String modelName}) =
327 OndeError_AlreadyLoaded;
328 const factory OndeError.modelBuild({required String reason}) =
329 OndeError_ModelBuild;
330 const factory OndeError.inference({required String reason}) =
331 OndeError_Inference;
332 const factory OndeError.cancelled() = OndeError_Cancelled;
333 const factory OndeError.other({required String reason}) = OndeError_Other;
334 }
335
336 /// Sampling parameters for text generation. All fields are optional — `None`
337 /// means "use the engine default".
338 class SamplingConfig {
339 final double? temperature;
340 final double? topP;
341 final BigInt? topK;
342 final double? minP;
343 final BigInt? maxTokens;
344 final double? frequencyPenalty;
345 final double? presencePenalty;
346
347 const SamplingConfig({
348 this.temperature,
349 this.topP,
350 this.topK,
351 this.minP,
352 this.maxTokens,
353 this.frequencyPenalty,
354 this.presencePenalty,
355 });
356
357 @override
358 int get hashCode =>
359 temperature.hashCode ^
360 topP.hashCode ^
361 topK.hashCode ^
362 minP.hashCode ^
363 maxTokens.hashCode ^
364 frequencyPenalty.hashCode ^
365 presencePenalty.hashCode;
366
367 @override
368 bool operator ==(Object other) =>
369 identical(this, other) ||
370 other is SamplingConfig &&
371 runtimeType == other.runtimeType &&
372 temperature == other.temperature &&
373 topP == other.topP &&
374 topK == other.topK &&
375 minP == other.minP &&
376 maxTokens == other.maxTokens &&
377 frequencyPenalty == other.frequencyPenalty &&
378 presencePenalty == other.presencePenalty;
379 }
380
381 /// A single streaming token chunk emitted during inference.
382 class StreamChunk {
383 /// The token text delta (may be empty on the final chunk).
384 final String delta;
385
386 /// `true` when this is the last chunk in the stream.
387 final bool done;
388
389 /// Finish reason (set only on the final chunk).
390 final String? finishReason;
391
392 const StreamChunk({
393 required this.delta,
394 required this.done,
395 this.finishReason,
396 });
397
398 @override
399 int get hashCode => delta.hashCode ^ done.hashCode ^ finishReason.hashCode;
400
401 @override
402 bool operator ==(Object other) =>
403 identical(this, other) ||
404 other is StreamChunk &&
405 runtimeType == other.runtimeType &&
406 delta == other.delta &&
407 done == other.done &&
408 finishReason == other.finishReason;
409 }
410
411 /// Metadata about a tool call requested by the model.
412 class ToolCallInfo {
413 /// Unique identifier for this tool call.
414 final String id;
415
416 /// Name of the function to call.
417 final String functionName;
418
419 /// JSON-encoded arguments for the function.
420 final String arguments;
421
422 const ToolCallInfo({
423 required this.id,
424 required this.functionName,
425 required this.arguments,
426 });
427
428 @override
429 int get hashCode => id.hashCode ^ functionName.hashCode ^ arguments.hashCode;
430
431 @override
432 bool operator ==(Object other) =>
433 identical(this, other) ||
434 other is ToolCallInfo &&
435 runtimeType == other.runtimeType &&
436 id == other.id &&
437 functionName == other.functionName &&
438 arguments == other.arguments;
439 }