1#![forbid(unsafe_code)]
8#![warn(missing_docs)]
9
10pub mod admission;
12pub mod artifact;
14pub mod attention;
16pub mod automatic;
18pub mod backend;
20pub mod cache;
22pub mod capability;
24pub mod capture;
26pub mod checkpoint;
28pub mod consensus;
30pub mod discovery;
32pub use discovery::*;
33pub mod execution;
35pub mod execution_control;
37pub mod generation;
39pub mod input;
41pub mod inspection;
43pub mod intervention;
45pub mod media;
47pub mod observation;
49pub mod realtime;
51pub mod residency;
53pub mod scheduler;
55pub mod session_authority;
57pub mod speculative;
59pub mod topology;
61
62pub use admission::{
63 admit_preparation, ArchitecturePreparationCapabilities, PreparationAdmission,
64 PreparationAdmissionError, PreparationAdmissionRequest, PreparationMechanismCapabilities,
65};
66pub use artifact::{
67 gguf_u32_metadata_values, inspect_artifact, plan_model_preparation, resolve_gguf_companions,
68 validate_preparation_policy, ArtifactFormat, ArtifactInspection, GgufCompanionEncoding,
69 GgufCompanionRequirement, GgufCompanionRole, LoadingProtocol, MaterializationRoute,
70 ModelArtifact, ModelConfiguration, ModelConfigurationResolver, ModelPreparationPlan,
71 PreparationPolicy, QuantizationRequest, ResidencyRequest, ResolvedModelConfiguration,
72 ValidatedGguf, ValidatedGgufCompanion,
73};
74pub use attention::{AttentionPolicy, LayerSchedule, LayerScheduleError};
75pub use automatic::{
76 realize_execution_plan_drafting, realize_execution_plan_target, select_execution_plan_drafting,
77 select_execution_plan_target, speculative_decoding_telemetry, AllocatorTelemetry,
78 AutomaticPlanRequest, AutomaticPlanner, AutomaticPlannerPolicy, AutomaticPlanningBackend,
79 AutomaticPlanningError, BoundedResidencyRequirement, CandidateAdmission, DurationSeconds,
80 ExecutionPlanBackendFactory, ExecutionPlanReport, ExecutionPlanTarget,
81 ExecutionPlanTargetLoadError, ExecutionPlanTargetSelection, ExecutionTelemetry,
82 ExpertCacheTelemetry, ExternalDraftArtifact, HardwareBackendProfile, HardwareDeviceProfile,
83 HardwareMemorySemantics, HardwareProfile, ModelResourceProfile, ObservationKind, Observed,
84 PlanExplanation, PlanExplanationEntry, PlanExplanationLevel, PreparedExecutionPlanTarget,
85 RealizedDrafting, ResidencyTelemetry, RetainedAutomaticPlan, SelectedExecutionPlanDrafting,
86 SelectedExecutionPlanTarget, SpeculativeDecodingTelemetry, TimingTelemetry,
87 TokenizerCompatibilityError, TokenizerCompatibilityProof, TransferTelemetry,
88 AUTOMATIC_SCHEMA_VERSION,
89};
90pub use backend::{
91 load_model, prepare_inspected_model, BackendDescriptor, BackendError, BackendProvider,
92 BackendSession, BoundedCompletion, BoundedCompletionOutcome, BoundedCompletionWait,
93 BoundedCompletionWaitError, BoundedSubmissionOutcome, CollectiveGroupDescriptor,
94 CollectiveGroupId, CollectiveScope, Completion, CompletionCancellationMode,
95 ControlledTextGeneration, ControlledTextGenerationError, ControlledToken, DeviceCapabilities,
96 DeviceDescriptor, DistributedBackend, DistributedCapabilities, DistributedCommitEpoch,
97 DistributedCommitOutcome, DistributedCommitPhase, DistributedSession,
98 DistributedSessionDescriptor, InspectableBackendSession, ModelCapabilityBackend,
99 ModelLoadError, ModelLoadingBackend, ModelRuntime, MultimodalPreparationBackend,
100 MultimodalPreparationFailure, PendingTextInput, PreparedModel, SelectedModelPreparation,
101 SessionCapabilities, SessionCapabilityError, SpeculativeTokenFilterController, Submission,
102 TextContinuationBoundary, TextContinuationError, TextContinuationIdentity, TextDriverIdentity,
103 TextGeneration, TextGenerationBackend, TextGenerationConfig, TextGenerationContinuation,
104 TextGenerationDriver, TextSamplingStrategy, TokenFilter, TokenFilterController,
105 TokenFilterError, TokenOutput, ValueDescriptor,
106};
107pub use capability::{
108 apply_admission_policy, estimate_runtime_state, Admission, AdmissionRejection,
109 AdmissionRequest, AdmissionResult, AvailableMemory, CacheStateStrategy, CapabilityError,
110 EstimationCompleteness, InputModalities, InputTokenCount, ModelCapabilities,
111 PhysicalMemorySemantics, RuntimeStateEstimate, SlidingWindowLayerCount, StateMemoryAssumptions,
112 StateMemoryLayout, StaticMemoryReport,
113};
114pub use execution::{
115 BackendId, DevicePlan, DraftPlacementPlan, DraftingPlan, ExecutionPlan, ExecutionPlanError,
116 ExpertCachePlan, ResidencyPlan, WeightTransformationPlan, DEFAULT_MAX_CACHED_SHARDS,
117 EXECUTION_PLAN_SCHEMA_VERSION,
118};
119pub use generation::{
120 resolve_generation_config, resolve_optimistic_reuse, CheckpointGenerationConfig, FinishReason,
121 GenerationCancellationToken, GenerationConfigOverrides, GenerationError, GenerationSequence,
122 OptimisticReuseDecision, ResolvedGenerationConfig, SemanticEvent,
123 SpeculativeCancellationDisposition, SpeculativeCommitPlan, SpeculativeConfig,
124 SpeculativeRequestId, SpeculativeRequestLifecycle, SpeculativeRequestStatus, SpeculativeRound,
125 SpeculativeSchedulerOptions, SpeculativeTail, TokenCommit, TokenTerminalSignals,
126};
127pub use input::{
128 InputExtent, InputMetadataKey, InputModality, InputPartDescriptor, InputPayloadKind,
129 InputTensorIdentity, PreparedInputError, PreparedInputIdentity,
130};
131pub use inspection::{
132 assemble_portable_model_inspection, finalize_realized_model_inspection,
133 media_feature_readiness, record_media_projector_inspection, record_processor_inspection,
134 reject_portable_artifact_inspection, ArtifactModality, ArtifactTensorEncoding, InspectionIssue,
135 InspectionIssueCode, InspectionReadiness, InspectionRequirement, InspectionSeverity,
136 MediaFeatureAvailability, MediaProjectorRequirement, ModelInspectionReport,
137 RealizedInspectionOutcomes,
138};
139pub use media::{
140 Audio, Media, MediaBinding, MediaRequestError, MultimodalRequest, MultimodalSegment, RgbImage,
141 TokenizedMultimodalRequest, TokenizedMultimodalSegment, Video, VideoSampling,
142};
143pub use observation::{
144 InspectedOutput, ObservationError, ObservationRequest, ObservationSelector, ObservationSet,
145 ObservationValue, TensorObservation, TensorObservationData,
146 AUDIO_PROJECTOR_OUTPUT_OBSERVATION_PATH, MODALITY_MERGE_OUTPUT_OBSERVATION_PATH,
147 MODEL_LOGITS_OBSERVATION_PATH, PROCESSOR_OUTPUT_OBSERVATION_PATH,
148 VISION_PROJECTOR_OUTPUT_OBSERVATION_PATH,
149};
150pub use realtime::{
151 RealtimeConfigError, RealtimeDecisionDiagnostics, RealtimeError, RealtimeForcedSource,
152 RealtimeFrameConvention, RealtimeFrameForcing, RealtimeFrameScheduleState, RealtimeFrameSlot,
153 RealtimeFrameTransition, RealtimeInputDescriptorError, RealtimeInputFrame, RealtimeOutputFrame,
154 RealtimeSampling, RealtimeScheduleError, RealtimeSlotCoordinate, RealtimeSlotOccupancy,
155 RealtimeSpeechConfig, RealtimeTargetDecision, RealtimeTargetSource, RealtimeTemporalSource,
156 MAX_REALTIME_FRAME_DELAY,
157};
158pub use residency::{
159 BackgroundPrefetchReport, PrefetchAdmission, PrefetchCompletion, PrefetchDemandObservation,
160 PrefetchDemandResolution, PrefetchExecutionState, PrefetchStateError, PrefetchWork,
161};
162pub use session_authority::{
163 SessionAdmission, SessionAdmissionError, SessionAuthority, SessionAuthorityError,
164 SubmissionLease,
165};
166pub use speculative::{
167 cancel_pending_verification, decide_speculative_proposal, propose_block,
168 resolve_commit_and_publish, resolve_optimistic_branch, resolve_round,
169 speculative_acceptance_probability, submit_verification_transaction,
170 CompletedSpeculativeRequest, CompletedSpeculativeSchedule, PendingSpeculativeVerification,
171 PreparedSpeculativeLane, ProposalDecision, PublishedSpeculativeResult,
172 PublishedSpeculativeVerification, ResolvedSpeculativeRound, SamplingPlacement,
173 SpeculativeAction, SpeculativeCallbackPublisher, SpeculativeCandidate, SpeculativeCapability,
174 SpeculativeCommit, SpeculativeConstraint, SpeculativeContinuation, SpeculativeDraft,
175 SpeculativeDraftBlock, SpeculativeDraftRandomPosition, SpeculativeDraftSource,
176 SpeculativeDriverError, SpeculativeExecutionTopology, SpeculativeExecutor,
177 SpeculativeGenerationBackend, SpeculativeGenerationBatchOutput,
178 SpeculativeGenerationBatchRequest, SpeculativeGenerationLane, SpeculativeGenerationOutput,
179 SpeculativeGenerationVisitor, SpeculativeLifecycleObserver, SpeculativeLifecycleStage,
180 SpeculativeOptimisticBranch, SpeculativeOutputError, SpeculativeOutputRuntime,
181 SpeculativePrefill, SpeculativeProposal, SpeculativePublicationStatus, SpeculativePublisher,
182 SpeculativeRandomness, SpeculativeRequest, SpeculativeRequestTable, SpeculativeSampling,
183 SpeculativeSchedule, SpeculativeSchedulerStats, SpeculativeSemanticConstraint,
184 SpeculativeSemanticState, SpeculativeStats, SpeculativeTelemetry,
185};
186pub use topology::{
187 balanced_contiguous_range, ParallelAxis, ParallelCoordinates, ParallelRankTopology,
188 ParallelTopology, SubgroupMembership, TopologyError, TopologyPreflightReport,
189};