@github/copilot-sdk 1.0.18-preview.3 → 1.0.18-preview.4

package/dist/generated/session-events.d.ts

@@ -9,7 +9,7 @@ export type JsonValue = null | boolean | number | string | JsonValue[] | {
 /**
  * Union of all session event variants emitted by the Copilot CLI runtime.
  */
-export type SessionEvent = StartEvent | ResumeEvent | RemoteSteerableChangedEvent | ErrorEvent | IdleEvent | TitleChangedEvent | ScheduleCreatedEvent | ScheduleCancelledEvent | ScheduleRearmedEvent | AutopilotObjectiveChangedEvent | InfoEvent | IndexedSearchEvent | WarningEvent | ModelChangeEvent | ModelDeselectedEvent | AutoTierRecommendationEvent | AutoTierSwitchFailedEvent | ModeChangedEvent | ModeNoticeDeliveredEvent | SessionLimitsChangedEvent | PermissionsChangedEvent | PlanChangedEvent | TodosChangedEvent | WorkspaceFileChangedEvent | HandoffEvent | TruncationEvent | SnapshotRewindEvent | ShutdownEvent | UsageCheckpointEvent | ContextChangedEvent | UsageInfoEvent | ContextClearedEvent | CompactionStartEvent | CompactionCompleteEvent | TaskCompleteEvent | CompletionReceiptEvent | FusionRouteStartedEvent | FusionRouteFailedEvent | FusionResolvedEvent | FusionCompletedEvent | PermissionRecoveryEvent | UserMessageEvent | PendingMessagesModifiedEvent | AssistantTurnStartEvent | AssistantIntentEvent | AssistantFusionPhaseStartedEvent | AssistantFusionPhaseActivityEvent | AssistantFusionPhaseCompletedEvent | AssistantFusionPhaseFailedEvent | AssistantServerToolProgressEvent | AssistantReasoningEvent | AssistantReasoningDeltaEvent | AssistantToolCallDeltaEvent | AssistantStreamingDeltaEvent | AssistantMessageEvent | AssistantMessageStartEvent | AssistantMessageDeltaEvent | AssistantTurnEndEvent | AssistantIdleEvent | AssistantUsageEvent | ModelCallFailureEvent | ModelCallFinishedEvent | AbortEvent | ToolUserRequestedEvent | ToolExecutionStartEvent | ToolExecutionPartialResultEvent | ToolShellOutputEvent | ToolExecutionProgressEvent | ToolExecutionCompleteEvent | ToolSearchActivatedEvent | SkillInvokedEvent | SubagentStartedEvent | SubagentConfiguredEvent | SubagentCompletedEvent | SubagentFailedEvent | SubagentSelectedEvent | SubagentDeselectedEvent | HookStartEvent | HookEndEvent | HookProgressEvent | BinaryAssetEvent | SystemMessageEvent | SystemNotificationEvent | PermissionRequestedEvent | PermissionCompletedEvent | PermissionCarriedForwardEvent | PermissionMessageAuthorizationEvent | PermissionMessageAuthorizationReadEvent | PermissionMessageAuthorizationDegradedEvent | PermissionAssentDetectedEvent | PermissionContextualAuthorizationEvent | UserInputRequestedEvent | UserInputCompletedEvent | ElicitationRequestedEvent | ElicitationCompletedEvent | SamplingRequestedEvent | SamplingCompletedEvent | McpOauthRequiredEvent | McpOauthCompletedEvent | McpHeadersRefreshRequiredEvent | McpHeadersRefreshCompletedEvent | CustomNotificationEvent | UIEphemeralQueryEvent | ExternalToolRequestedEvent | ExternalToolCompletedEvent | CommandQueuedEvent | CommandExecuteEvent | CommandCompletedEvent | AutoModeSwitchRequestedEvent | AutoModeSwitchCompletedEvent | SessionLimitsExhaustedRequestedEvent | SessionLimitsExhaustedCompletedEvent | AutoModeResolvedEvent | ManagedSettingsResolvedEvent | ManagedSettingsEnforcedEvent | CommandsChangedEvent | CapabilitiesChangedEvent | ExitPlanModeRequestedEvent | ExitPlanModeCompletedEvent | HumanResponseRecordedEvent | ToolsUpdatedEvent | BackgroundTasksChangedEvent | WorkflowRunUpdatedEvent | WorkflowRunStartedEvent | WorkflowRunSettledEvent | SkillsLoadedEvent | CustomAgentsUpdatedEvent | McpServersLoadedEvent | McpServerStatusChangedEvent | McpServerRemovedEvent | McpServerNeedsReconnectEvent | McpToolsListChangedEvent | McpResourcesListChangedEvent | McpPromptsListChangedEvent | ExtensionsLoadedEvent | CanvasOpenedEvent | CanvasRegistryChangedEvent | CanvasClosedEvent | CanvasUnavailableEvent | CanvasRecordedEvent | CanvasRemovedEvent | ExtensionsAttachmentsPushedEvent | McpAppToolCallCompleteEvent;
+export type SessionEvent = StartEvent | ResumeEvent | RemoteSteerableChangedEvent | ErrorEvent | IdleEvent | TitleChangedEvent | ScheduleCreatedEvent | ScheduleCancelledEvent | ScheduleRearmedEvent | AutopilotObjectiveChangedEvent | InfoEvent | IndexedSearchEvent | WarningEvent | ModelChangeEvent | ModelDeselectedEvent | AutoTierRecommendationEvent | AutoTierSwitchFailedEvent | ModeChangedEvent | ModeNoticeDeliveredEvent | SessionLimitsChangedEvent | PermissionsChangedEvent | PlanChangedEvent | TodosChangedEvent | WorkspaceFileChangedEvent | HandoffEvent | TruncationEvent | SnapshotRewindEvent | ShutdownEvent | UsageCheckpointEvent | ProviderQuotaObservationEvent | ContextChangedEvent | UsageInfoEvent | ContextClearedEvent | CompactionStartEvent | CompactionCompleteEvent | TaskCompleteEvent | CompletionReceiptEvent | FusionRouteStartedEvent | FusionRouteFailedEvent | FusionResolvedEvent | FusionCompletedEvent | PermissionRecoveryEvent | UserMessageEvent | PendingMessagesModifiedEvent | AssistantTurnStartEvent | AssistantIntentEvent | AssistantFusionPhaseStartedEvent | AssistantFusionPhaseActivityEvent | AssistantFusionPhaseCompletedEvent | AssistantFusionPhaseFailedEvent | AssistantServerToolProgressEvent | AssistantReasoningEvent | AssistantReasoningDeltaEvent | AssistantToolCallDeltaEvent | AssistantStreamingDeltaEvent | AssistantMessageEvent | AssistantMessageStartEvent | AssistantMessageDeltaEvent | AssistantTurnEndEvent | AssistantIdleEvent | AssistantUsageEvent | ModelCallFailureEvent | ModelCallFinishedEvent | AbortEvent | ToolUserRequestedEvent | ToolExecutionStartEvent | ToolExecutionPartialResultEvent | ToolShellOutputEvent | ToolExecutionProgressEvent | ToolExecutionCompleteEvent | ToolSearchActivatedEvent | SkillInvokedEvent | SubagentStartedEvent | SubagentConfiguredEvent | SubagentCompletedEvent | SubagentFailedEvent | SubagentSelectedEvent | SubagentDeselectedEvent | HookStartEvent | HookEndEvent | HookProgressEvent | BinaryAssetEvent | SystemMessageEvent | SystemNotificationEvent | PermissionRequestedEvent | PermissionCompletedEvent | PermissionCarriedForwardEvent | PermissionMessageAuthorizationEvent | PermissionMessageAuthorizationReadEvent | PermissionMessageAuthorizationDegradedEvent | PermissionAssentDetectedEvent | PermissionContextualAuthorizationEvent | UserInputRequestedEvent | UserInputCompletedEvent | ElicitationRequestedEvent | ElicitationCompletedEvent | SamplingRequestedEvent | SamplingCompletedEvent | McpOauthRequiredEvent | McpOauthCompletedEvent | McpHeadersRefreshRequiredEvent | McpHeadersRefreshCompletedEvent | CustomNotificationEvent | UIEphemeralQueryEvent | ExternalToolRequestedEvent | ExternalToolCompletedEvent | CommandQueuedEvent | CommandExecuteEvent | CommandCompletedEvent | AutoModeSwitchRequestedEvent | AutoModeSwitchCompletedEvent | SessionLimitsExhaustedRequestedEvent | SessionLimitsExhaustedCompletedEvent | AutoModeResolvedEvent | ManagedSettingsResolvedEvent | ManagedSettingsEnforcedEvent | CommandsChangedEvent | CapabilitiesChangedEvent | ExitPlanModeRequestedEvent | ExitPlanModeCompletedEvent | HumanResponseRecordedEvent | ToolsUpdatedEvent | BackgroundTasksChangedEvent | WorkflowRunUpdatedEvent | WorkflowRunStartedEvent | WorkflowRunSettledEvent | SkillsLoadedEvent | CustomAgentsUpdatedEvent | McpServersLoadedEvent | McpServerStatusChangedEvent | McpServerRemovedEvent | McpServerNeedsReconnectEvent | McpToolsListChangedEvent | McpResourcesListChangedEvent | McpPromptsListChangedEvent | ExtensionsLoadedEvent | CanvasOpenedEvent | CanvasRegistryChangedEvent | CanvasClosedEvent | CanvasUnavailableEvent | CanvasRecordedEvent | CanvasRemovedEvent | ExtensionsAttachmentsPushedEvent | McpAppToolCallCompleteEvent;
 /**
  * Extensible routing preference for the virtual `auto` model. New identifiers must be advertised and enabled by the provider. `fast` is an integrator-only latency preset.
  */
@@ -366,6 +366,25 @@ export type HandoffSourceType =
 "remote"
 /** The handoff originated from a local session. */
  | "local";
+/**
+ * Whether the accumulated numeric AI-credit subtotal covers the observed calls.
+ */
+export type AiCreditsStatus = 
+/** Every observed call supplied an AI-credit amount, including an explicit zero. */
+"complete"
+/** Some calls supplied amounts and some did not; the numeric value is only a reported subtotal. */
+ | "partial"
+/** No AI-credit amount was reported for the observed calls. Numeric zero is not a zero-cost claim. */
+ | "unavailable";
+/**
+ * The neutral kind of a model provider — the model analog of `AccountKind`. A model provider is the live, entitled source a model came from; central code never branches on this beyond a single dispatch.
+ */
+/** @experimental */
+export type ModelProviderKind = 
+/** GitHub Copilot / CAPI models, spawned by a github-resolving account that holds a Copilot seat. */
+"copilot"
+/** Microsoft 365 Copilot (Loki) inference models, spawned by a resolvable Entra-derived Loki account. */
+ | "loki";
 /**
  * Whether the session ended normally ("routine") or due to a crash/fatal error ("error")
  */
@@ -374,6 +393,78 @@ export type ShutdownType =
 "routine"
 /** The session ended because of a crash or fatal error. */
  | "error";
+/**
+ * Access policy reported by the quota service.
+ */
+export type ProviderQuotaAccessState = 
+/** Access is allowed. */
+"allowed"
+/** Access is denied. */
+ | "denied"
+/** Quota is not required for this access. */
+ | "not_required"
+/** The service did not supply a recognized access state. */
+ | "unknown"
+/** Client-only compatibility placeholder when acquisitionStatus is unavailable or failed; not an observed service verdict. */
+ | "unavailable";
+/**
+ * Acquisition state, independent of the service's business access/capacity verdict.
+ */
+export type ProviderQuotaAcquisitionStatus = 
+/** A service quota observation was received. */
+"succeeded"
+/** No service snapshot is available for this caller or surface. */
+ | "unavailable"
+/** Acquisition failed; inspect httpStatus/acquisitionError without inferring a business verdict. */
+ | "failed";
+/**
+ * Capacity is independent of whether a numeric balance was supplied.
+ */
+export type ProviderQuotaCapacityState = 
+/** Capacity is available. */
+"available"
+/** The service explicitly reports exhaustion. */
+ | "exhausted"
+/** The reported entitlement is unlimited. */
+ | "unlimited"
+/** No billable quota is required. */
+ | "not_required"
+/** The service reports that capacity is not applicable. */
+ | "not_applicable"
+/** No recognized capacity state was supplied. */
+ | "unknown"
+/** Client-only compatibility placeholder when acquisitionStatus is unavailable or failed; not an observed service verdict. */
+ | "unavailable";
+/**
+ * Timing and purpose of a provider quota observation.
+ */
+export type ProviderQuotaObservationKind = 
+/** An explicit account read; consumption may still be subject to service aggregation delay. */
+"account_snapshot"
+/** A verdict attached before an inference response. May be cached; never a post-charge balance. */
+ | "admission_state";
+/**
+ * Interpretation permitted for independently reported quantities.
+ */
+export type ProviderQuotaQuantityKind = 
+/** Quantities form an authoritative available-of-entitled budget suitable for percentage presentation. */
+"authoritative_budget"
+/** Raw advisory quantities; do not infer permission, usage percentage, reset cadence, or post-charge balance. */
+ | "advisory_balance"
+/** No quantity measurement, as with admission-only response headers. */
+ | "none";
+/**
+ * Units explicitly reported by a quota provider.
+ */
+export type ProviderQuotaUnit = 
+/** AI credits, not tokens or currency. */
+"ai_credits"
+/** Requests charged against an entitlement. */
+ | "requests"
+/** Model tokens. */
+ | "tokens"
+/** The observation does not establish a quantity unit. */
+ | "unknown";
 /**
  * What initiated a conversation compaction
  */
@@ -1487,6 +1578,10 @@ export interface StartData {
      * Identifier of the software producing the events (e.g., "copilot-agent")
      */
     producer: string;
+    /**
+     * Provider of selectedModel at creation time, when explicitly selected.
+     */
+    providerId?: string;
     /**
      * Reasoning effort level used for model calls, if applicable (e.g. "none", "low", "medium", "high", "xhigh", "max")
      */
@@ -1655,6 +1750,10 @@ export interface ResumeData {
      * On-disk byte size of the session's persisted events.jsonl file at resume time; omitted when the file does not exist or cannot be stat'd
      */
     eventsFileSizeBytes?: number;
+    /**
+     * Provider of selectedModel at resume time, when explicitly selected.
+     */
+    providerId?: string;
     /**
      * Reasoning effort level used for model calls, if applicable (e.g. "none", "low", "medium", "high", "xhigh", "max")
      */
@@ -2267,12 +2366,20 @@ export interface ModelChangeData {
      * Model that was previously selected, if any
      */
     previousModel?: string;
+    /**
+     * Provider of previousModel, when known. A provider-only change is a model selection change even when the model identifiers are equal.
+     */
+    previousProviderId?: string;
     /**
      * Reasoning effort level before the model change, if applicable
      */
     previousReasoningEffort?: string;
     previousReasoningSummary?: ReasoningSummary;
     previousVerbosity?: Verbosity;
+    /**
+     * Provider selected for newModel. Omitted for legacy or unattributed selections; never inferred from a later selection.
+     */
+    providerId?: string;
     /**
      * Reasoning effort level after the model change, if applicable
      */
@@ -2945,6 +3052,10 @@ export interface ShutdownData {
     modelMetrics: {
         [k: string]: ShutdownModelMetric | undefined;
     };
+    /**
+     * Provider/model accounting preserved for replay without consulting current provider selection.
+     */
+    providerModelMetrics?: UsageMetricsProviderModelMetric[];
     /**
      * Unix timestamp (milliseconds) when the session started
      */
@@ -2974,6 +3085,233 @@ export interface ShutdownData {
      * @experimental
      */
     totalNanoAiu?: number;
+    /**
+     * Per-source accounting sequence watermarks subsuming preceding usage receipts.
+     */
+    usageAccountingWatermarks?: {
+        [k: string]: number | undefined;
+    };
+}
+/**
+ * Accumulated session usage metrics, including premium request cost, token counts, model breakdown, and code-change totals.
+ */
+export interface UsageGetMetricsResult {
+    /**
+     * Per-agent usage metrics, keyed by agent instance identifier. The main conversation uses the stable key `main`.
+     */
+    agentMetrics?: {
+        [k: string]: UsageMetricsAgentMetric | undefined;
+    };
+    aiCreditsStatus?: AiCreditsStatus;
+    codeChanges: UsageMetricsCodeChanges;
+    /**
+     * Currently active model identifier
+     */
+    currentModel?: string;
+    /**
+     * Input tokens from the most recent main-agent API call
+     */
+    lastCallInputTokens: number;
+    /**
+     * Output tokens from the most recent main-agent API call
+     */
+    lastCallOutputTokens: number;
+    /**
+     * Per-model token and request metrics, keyed by model identifier
+     */
+    modelMetrics: {
+        [k: string]: UsageMetricsModelMetric | undefined;
+    };
+    /**
+     * Provider-attributed model metrics, ordered by first observed provider and model.
+     * Null attribution preserves historical usage whose provider or model was not recorded.
+     */
+    providerModelMetrics?: UsageMetricsProviderModelMetric[];
+    /**
+     * ISO 8601 timestamp when the session started
+     */
+    sessionStartTime: string;
+    /**
+     * Session-wide per-token-type accumulated token counts
+     */
+    tokenDetails?: {
+        [k: string]: UsageMetricsTokenDetail | undefined;
+    };
+    /**
+     * Total time spent in model API calls (milliseconds)
+     */
+    totalApiDurationMs: number;
+    /**
+     * Session-wide accumulated nano-AI units cost
+     */
+    totalNanoAiu?: number;
+    /**
+     * Total user-initiated premium request cost across all models (may be fractional due to multipliers)
+     */
+    totalPremiumRequestCost: number;
+    /**
+     * Raw count of user-initiated API requests
+     */
+    totalUserRequests: number;
+}
+/**
+ * Usage attributed to one agent instance, including its identity, API duration, AI units, and per-model breakdown.
+ */
+export interface UsageMetricsAgentMetric {
+    /**
+     * Human-readable label for this subagent invocation, copied from the originating `subagent.started` event. For task-tool subagents this is the invocation's task description rather than the agent's configured display name, so group by `agentName` for stable per-agent labels.
+     */
+    agentDisplayName?: string;
+    /**
+     * Configured agent name, when this is a subagent
+     */
+    agentName?: string;
+    /**
+     * Per-model usage for this agent, keyed by model identifier
+     */
+    modelMetrics: {
+        [k: string]: UsageMetricsModelMetric | undefined;
+    };
+    /**
+     * Time spent in model API calls by this agent, in milliseconds
+     */
+    totalApiDurationMs: number;
+    /**
+     * Accumulated nano-AI units cost for this agent
+     */
+    totalNanoAiu: number;
+}
+/**
+ * Per-model usage metrics, including request counts/costs, token usage, nano-AI units, and per-token-type details.
+ */
+export interface UsageMetricsModelMetric {
+    aiCreditsStatus?: AiCreditsStatus;
+    /**
+     * Latest known prompt-cache expiration for this model. A timestamp in the past indicates that the observed cache has expired.
+     */
+    cacheExpiresAt?: string;
+    requests: UsageMetricsModelMetricRequests;
+    /**
+     * Token count details per type
+     */
+    tokenDetails?: {
+        [k: string]: UsageMetricsModelMetricTokenDetail | undefined;
+    };
+    /**
+     * Accumulated nano-AI units cost for this model
+     */
+    totalNanoAiu?: number;
+    usage: UsageMetricsModelMetricUsage;
+}
+/**
+ * Request count and cost metrics for this model
+ */
+export interface UsageMetricsModelMetricRequests {
+    /**
+     * User-initiated premium request cost (with multiplier applied)
+     */
+    cost: number;
+    /**
+     * Number of API requests made with this model
+     */
+    count: number;
+}
+/**
+ * Per-model token-detail entry containing the accumulated token count for one token type.
+ */
+export interface UsageMetricsModelMetricTokenDetail {
+    /**
+     * Accumulated token count for this token type
+     */
+    tokenCount: number;
+}
+/**
+ * Token usage metrics for this model
+ */
+export interface UsageMetricsModelMetricUsage {
+    /**
+     * Total tokens read from prompt cache
+     */
+    cacheReadTokens: number;
+    /**
+     * Total tokens written to prompt cache
+     */
+    cacheWriteTokens: number;
+    /**
+     * Total input tokens consumed
+     */
+    inputTokens: number;
+    /**
+     * Total output tokens produced
+     */
+    outputTokens: number;
+    /**
+     * Total output tokens used for reasoning
+     */
+    reasoningTokens?: number;
+}
+/**
+ * Aggregated code change metrics
+ */
+export interface UsageMetricsCodeChanges {
+    /**
+     * Distinct file paths modified during the session
+     */
+    filesModified: string[];
+    /**
+     * Number of distinct files modified
+     */
+    filesModifiedCount: number;
+    /**
+     * Total lines of code added
+     */
+    linesAdded: number;
+    /**
+     * Total lines of code removed
+     */
+    linesRemoved: number;
+}
+/**
+ * Usage for one recorded provider and model, without merging identical model IDs across providers.
+ */
+export interface UsageMetricsProviderModelMetric {
+    metrics: UsageMetricsModelMetric | undefined;
+    /**
+     * Model display name captured at call time, when known.
+     */
+    modelDisplayName?: string;
+    /**
+     * Model identity, or null for legacy aggregate-only usage.
+     */
+    modelId: string | null;
+    /**
+     * Provider identity and product label captured when the call was dispatched; null when unknown.
+     */
+    provider: ModelProviderRef | null;
+}
+/**
+ * A neutral reference to the model provider that produced a model: an opaque id, a human-readable label, and the provider kind. Carried on each enumerated Model so consumers can group by provider without reaching into a provider-shaped internal type.
+ */
+/** @experimental */
+export interface ModelProviderRef {
+    /**
+     * Opaque, stable id of the provider that produced this model. Matches the enumerated `ModelProviderDescriptor.id`.
+     */
+    id: string;
+    kind: ModelProviderKind;
+    /**
+     * Human-readable provider label, owned by the runtime so every consumer renders identical text.
+     */
+    label: string;
+}
+/**
+ * Session-wide token-detail entry containing the accumulated token count for one token type.
+ */
+export interface UsageMetricsTokenDetail {
+    /**
+     * Accumulated token count for this token type
+     */
+    tokenCount: number;
 }
 /**
  * Usage attributed to one agent instance at session shutdown.
@@ -3132,10 +3470,165 @@ export interface UsageCheckpointEvent {
  * Durable session usage checkpoint for reconstructing aggregate accounting on resume
  */
 export interface UsageCheckpointData {
+    /**
+     * Provider/model accounting snapshot that subsumes preceding usage receipts.
+     */
+    providerModelMetrics?: UsageMetricsProviderModelMetric[];
     /**
      * Session-wide accumulated nano-AI units cost at checkpoint time
      */
     totalNanoAiu: number;
+    /**
+     * Highest accounted sequence per source session; no per-call deduplication history is retained.
+     */
+    usageAccountingWatermarks?: {
+        [k: string]: number | undefined;
+    };
+}
+/**
+ * Session event "session.quota_observation". A provider-owned quota observation, distinct from per-call usage and charge accounting.
+ */
+/** @experimental */
+export interface ProviderQuotaObservationEvent {
+    /**
+     * Sub-agent instance identifier. Absent for events from the root/main agent and session-level events.
+     */
+    agentId?: string;
+    data: ProviderQuotaObservationData;
+    /**
+     * Always true for events that are transient and not persisted to the session event log on disk.
+     */
+    ephemeral: true;
+    /**
+     * Unique event identifier (UUID v4), generated when the event is emitted
+     */
+    id: string;
+    /**
+     * ID of the chronologically preceding event in the session, forming a linked chain. Null for the first event.
+     */
+    parentId: string | null;
+    /**
+     * ISO 8601 timestamp when the event was created
+     */
+    timestamp: string;
+    /**
+     * Type discriminator. Always "session.quota_observation".
+     */
+    type: "session.quota_observation";
+}
+/**
+ * A provider-owned quota observation, distinct from per-call usage and charge accounting.
+ */
+/** @experimental */
+export interface ProviderQuotaObservationData {
+    observation: ProviderQuotaState;
+}
+/**
+ * An account quota reading. Absence of a quantity is unknown, never zero.
+ */
+/** @experimental */
+export interface ProviderQuotaState {
+    accessState: ProviderQuotaAccessState;
+    /**
+     * Service error code or client acquisition category, separate from the business-state reason.
+     */
+    acquisitionError?: string;
+    acquisitionStatus?: ProviderQuotaAcquisitionStatus;
+    /**
+     * Independently reported signed 64-bit available quantity. Zero does not override the service access/capacity verdict; omission and null are preserved.
+     */
+    availableQuantity?: number | null;
+    budgetMetadata?: ProviderQuotaBudgetMetadata;
+    capacityState: ProviderQuotaCapacityState;
+    /**
+     * Key for the backwards-compatible snapshots projection, when the authoritative budget supports that contract.
+     */
+    compatibilityKey?: string;
+    /**
+     * Independently reported signed 64-bit entitlement. -1 is an unlimited sentinel, not a capacity-state rewrite; omission and null are preserved.
+     */
+    entitledQuantity?: number | null;
+    /**
+     * Explicit service admission flag, when reported.
+     */
+    hasQuota?: boolean | null;
+    /**
+     * HTTP status from acquisition, when available.
+     */
+    httpStatus?: number;
+    observationKind?: ProviderQuotaObservationKind;
+    /**
+     * When the runtime observed this reading, not a charge timestamp or guarantee that consumption has settled.
+     */
+    observedAt?: string;
+    provider: ModelProviderRef;
+    quantityKind?: ProviderQuotaQuantityKind;
+    /**
+     * Service-owned quota identifier within this provider.
+     */
+    quotaId: string;
+    /**
+     * Service-reported explanation for the state.
+     */
+    reason?: string | null;
+    /**
+     * Service name owning this reading.
+     */
+    service?: string | null;
+    /**
+     * Service-reported quota source.
+     */
+    source?: string | null;
+    unit: ProviderQuotaUnit;
+}
+/**
+ * Authoritative budget measurements and policy metadata, independent of provider.
+ */
+export interface ProviderQuotaBudgetMetadata {
+    /**
+     * Exact measured consumption in the budget's unit.
+     */
+    consumed: number;
+    /**
+     * Exact budget entitlement, retaining incumbent fractional-unit compatibility.
+     */
+    entitlement: number;
+    /**
+     * Usage beyond entitlement, in the budget's unit.
+     */
+    overage: number;
+    /**
+     * Whether additional usage is allowed when the budget is exhausted.
+     */
+    overageAllowedWhenExhausted: boolean;
+    /**
+     * Optional additional-usage budget cap.
+     */
+    overageLimit?: number;
+    /**
+     * Service-reported remaining percentage for authoritative budget presentation.
+     */
+    remainingPercentage: number;
+    /**
+     * Reset instant in epoch milliseconds when this budget actually defines a window.
+     */
+    resetAtEpochMs?: number;
+    /**
+     * Whether the reset instant is an estimate.
+     */
+    resetEstimated?: boolean;
+    /**
+     * Whether this budget uses token-based billing.
+     */
+    tokenBasedBilling?: boolean;
+    /**
+     * Whether the budget has unlimited entitlement.
+     */
+    unlimited: boolean;
+    /**
+     * Whether service policy allows continued usage after exhaustion.
+     */
+    usageAllowedWhenExhausted: boolean;
 }
 /**
  * Session event "session.context_changed". Updated working directory and git context after the change
@@ -3451,6 +3944,7 @@ export interface CompactionCompleteData {
  * Token usage breakdown for the compaction LLM call (aligned with assistant.usage format)
  */
 export interface CompactionCompleteCompactionTokensUsed {
+    aiCreditsStatus?: AiCreditsStatus;
     /**
      * Cached input tokens reused in the compaction LLM call
      */
@@ -3471,10 +3965,15 @@ export interface CompactionCompleteCompactionTokensUsed {
      * Model identifier used for the compaction LLM call
      */
     model?: string;
+    /**
+     * Model display name captured for the compaction call.
+     */
+    modelDisplayName?: string;
     /**
      * Output tokens produced by the compaction LLM call
      */
     outputTokens?: number;
+    provider?: ModelProviderRef;
 }
 /**
  * Token usage detail for a single billing category
@@ -5451,6 +5950,10 @@ export interface AssistantMessageData {
      * Generation phase for phased-output models (e.g., thinking vs. response phases)
      */
     phase?: string;
+    /**
+     * Opaque origin provider identity captured at dispatch. Only an exact match with the target provider permits replay of reasoningOpaque, encryptedContent, reasoningBlocks, serverTools, and provider response/item identifiers. Matching model ids or wire protocols is insufficient. Omitted for legacy or unattributed messages; the runtime excludes unproven opaque state when projecting to a known provider. Text content and client tool requests/results remain portable semantic history.
+     */
+    providerId?: string;
     reasoningBlocks?: AssistantMessageReasoningBlocks;
     /**
      * Opaque/encrypted extended thinking data from Anthropic models. Session-bound and stripped on resume.
@@ -5974,6 +6477,8 @@ export interface AssistantUsageData {
      * Number of accepted speculative prediction tokens
      */
     acceptedPredictionTokens?: number;
+    accounting?: UsageAccountingIdentity;
+    aiCreditsStatus?: AiCreditsStatus;
     /**
      * Completion ID from the model provider (e.g., chatcmpl-abc123)
      */
@@ -6056,6 +6561,10 @@ export interface AssistantUsageData {
      * Model identifier used for this API call
      */
     model: string;
+    /**
+     * Model display name captured for this call, not derived from the current selection.
+     */
+    modelDisplayName?: string;
     /**
      * Fixed-set provider family serving the bring-your-own-key model (for example "openai", "anthropic", "azure_openai", "ollama", "llama_cpp", or "other"). Never the caller-supplied provider name. Absent for Copilot-served models.
      */
@@ -6073,6 +6582,7 @@ export interface AssistantUsageData {
      * Parent tool call ID when this usage originates from a sub-agent
      */
     parentToolCallId?: string;
+    provider?: ModelProviderRef;
     /**
      * GitHub request tracing ID (x-github-request-id header) for server-side log correlation
      */
@@ -6113,6 +6623,23 @@ export interface AssistantUsageData {
     websocketFallbackAfterMs?: number;
     websocketFallbackReason?: ModelCallWebSocketFallbackReason;
 }
+/**
+ * Ordered accounting identity assigned under the source session's emission lock.
+ */
+export interface UsageAccountingIdentity {
+    /**
+     * Monotonically increasing sequence within the source session.
+     */
+    sequence: number;
+    /**
+     * Session that assigned this accounting sequence.
+     */
+    sourceSessionId: string;
+    /**
+     * Existing API call identifier, or a runtime-generated identity when none was supplied.
+     */
+    usageId: string;
+}
 /**
  * Per-request cost and usage data from the CAPI copilot_usage response field
  */

package/dist/session.d.ts

@@ -69,6 +69,7 @@ export declare class CopilotSession {
     private _capabilities;
     private openCanvasInstances;
     private disconnected;
+    private readonly pendingResponseWaits;
     private readonly pendingStructuredWaits;
     private disconnecting;
     private onDisconnected?;
@@ -173,6 +174,7 @@ export declare class CopilotSession {
     sendAndWait<TResult>(options: MessageOptions | string, responseSchema: ResponseSchema<TResult>, timeout?: number): Promise<TResult>;
     sendAndWait(options: MessageOptions, timeout?: number): Promise<AssistantMessageEvent | undefined>;
     private sendAndWaitForStructuredMessage;
+    private assertConnected;
     /**
      * Subscribes to events from this session.
      *
@@ -399,6 +401,7 @@ export declare class CopilotSession {
      * replace tools while the session is idle.
      *
      * @param tools - The complete set of tools this client supplies
+     * @throws Error if the session disconnects before the replacement is applied
      *
      * @experimental Wraps the experimental `session.tools.set` RPC and may change
      * or be removed in a future release.