@canarycoders/ai 0.7.0 → 0.8.0

package/dist/index.d.cts

@@ -66,6 +66,18 @@ interface ApiEnvelope<T> {
     data: T;
     message?: string;
 }
+/**
+ * What a call cost you. Present on every task result (chat and media) and on
+ * the final chunk of a stream. Absent only on results from older gateways.
+ */
+interface CostInfo {
+    /** Customer price in USD for this call; 0 when it ran on your own provider key (BYOK). */
+    customerPrice: number;
+    currency: "USD";
+    billing: "metered" | "byok";
+    /** True when served from the completion cache. */
+    cached?: boolean;
+}
 
 type ConversationTemplateType = "interview" | "conversation";
 interface ConversationQuestion {
@@ -330,6 +342,8 @@ interface CompletionResult {
     toolCalls?: ToolCall[];
     citations?: Citation[];
     metadata?: Record<string, unknown>;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 /** Raw chunk shape emitted by the native `/queue/stream` endpoint. */
 interface StreamChunk {
@@ -338,6 +352,8 @@ interface StreamChunk {
     finishReason?: string;
     toolCallDeltas?: ToolCallDelta[];
     metadata?: Record<string, unknown>;
+    /** Final chunk only: cost of the call. */
+    cost?: CostInfo;
 }
 /** Normalized streaming event, identical across every wire protocol. */
 type ChatStreamEvent = {
@@ -366,6 +382,7 @@ type ChatStreamEvent = {
     type: "done";
     finishReason?: string;
     usage?: TokenUsage;
+    cost?: CostInfo;
     raw?: unknown;
 } | {
     type: "raw";
@@ -488,6 +505,8 @@ interface ImageGenerationResult {
     images: GeneratedImage[];
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface VideoGenerateParams {
     provider: "gemini" | "vertex" | "xai";
@@ -513,6 +532,8 @@ interface VideoGenerationResult {
     videos: GeneratedVideo[];
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface VideoUploadResult {
     fileId: string;
@@ -548,6 +569,8 @@ interface TTSResult {
     model: string;
     provider: string;
     characterCount?: number;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface TranscribeParams {
     provider: "elevenlabs" | "mlxaudio";
@@ -576,6 +599,8 @@ interface STTResult {
     words?: TranscriptWord[];
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface SoundEffectParams {
     text: string;
@@ -591,6 +616,8 @@ interface SoundEffectResult {
     mimeType: string;
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface MusicParams {
     prompt: string;
@@ -605,6 +632,8 @@ interface MusicResult {
     mimeType: string;
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface DialogueInput {
     text: string;
@@ -626,6 +655,8 @@ interface DialogueResult {
     mimeType: string;
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface EmbeddingParams {
     provider: "lmstudio";
@@ -649,6 +680,8 @@ interface EmbeddingResult {
         inputTokens: number;
         totalTokens: number;
     };
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 
 declare class AudioResource extends BaseResource {
@@ -909,12 +942,16 @@ interface ImageRecognitionResult {
     model: string;
     width?: number;
     height?: number;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface FaceDetectionResult {
     faces: Detection[];
     /** base64-encoded image with faces blurred (when `blur: true`) */
     blurredImage?: string;
     model: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface WebEntity {
     description?: string;
@@ -924,6 +961,8 @@ interface WebDetectionResult {
     entities: WebEntity[];
     pages?: unknown[];
     matchingImages?: unknown[];
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface AutoLabelParams {
     images: string[];
@@ -1033,6 +1072,8 @@ interface OcrResult {
         width: number;
         height: number;
     };
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface RedactParams {
     /** base64-encoded image bytes (max ~50MB) */
@@ -1082,6 +1123,8 @@ interface RedactResult {
         width: number;
         height: number;
     };
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface OcrLanguages {
     recognitionLevels: Record<string, string[]>;
@@ -1207,4 +1250,4 @@ interface StreamAdapterOptions extends SSEOptions {
 
 // @ts-ignore
 export = CanaryCodersAI;
-export { APIConnectionError, APIConnectionTimeoutError, APIError, type ApiEnvelope, type AudioOutputFormat, AuthenticationError, type AutoLabelParams, type AutoLabelResult, type AutoTrainParams, BadRequestError, type BoundingBox, BrokeredCredential, CanaryCodersAI, type CanaryCodersAIOptions, CanaryCodersAI as CanaryLLM, type CanaryCodersAIOptions as CanaryLLMOptions, type ChatProvider, type ChatStreamEvent, type ChatStreamOptions, type Citation, CompatTarget, type CompleteParams, type CompletionResult, ConflictError, type ContentPart, type ConversationQuestion, type ConversationSession, type ConversationSessionRecord, type ConversationTemplate, type ConversationTemplateParams, type ConversationTemplateType, type ConversationTemplateUpdate, type ConversationTool, type ConversationVoiceSettings, type CreateConversationSessionParams, CreateRealtimeSessionParams, type DetectParams, type Detection, type DialogueInput, type DialogueParams, type DialogueResult, type DocumentContent, type EmbeddingParams, type EmbeddingResult, type FaceDetectParams, type FaceDetectionResult, FinalizeRealtimeParams, type GeneratedImage, type GeneratedVideo, type ImageContent, type ImageGenerateParams, type ImageGenerationResult, type ImageRecognitionResult, InternalServerError, Job, type JobSnapshot, type JobStatus, type JobStreamOptions, type KeyInfo, type KeyValidation, type Message, type MessageContent, type MessageRole, type ModelCapability, type ModelInfo, type MusicParams, type MusicResult, NotFoundError, type OcrGranularity, type OcrLanguages, type OcrLine, type OcrRecognitionLevel, type OcrRecognizeParams, type OcrResult, type OcrWord, PermissionError, type PollOptions, type PortalPeriod, type ProviderType, RateLimitError, RealtimeSession, RealtimeSessionRecord, type RedactParams, type RedactPreset, type RedactResult, type RedactStyle, type RedactionDetection, type ResponseFormat, type STTResult, type SignedUrlParams, type SignedUrlResult, type SoundEffectParams, type SoundEffectResult, type SpeechParams, type StreamAdapterOptions, type StreamChunk, type StreamProtocol, type TTSResult, type TaskKind, type TextContent, type ThinkingMode, type TokenUsage, type ToolCall, type ToolCallDelta, type ToolChoice, type ToolDefinition, type TrainParams, type TrainingJob, type TranscribeParams, type TranscriptWord, UnprocessableEntityError, type UsageSummary, type VideoContent, type VideoGenerateParams, type VideoGenerationResult, type VideoUploadOptions, type VideoUploadResult, type VisionModel, type VoiceInfo, type VoiceSettings, type WebDetectParams, type WebDetectionResult, type WebEntity, type WebSearchOptions, type ZeroShotDetectParams, toBrokeredCredential };
+export { APIConnectionError, APIConnectionTimeoutError, APIError, type ApiEnvelope, type AudioOutputFormat, AuthenticationError, type AutoLabelParams, type AutoLabelResult, type AutoTrainParams, BadRequestError, type BoundingBox, BrokeredCredential, CanaryCodersAI, type CanaryCodersAIOptions, CanaryCodersAI as CanaryLLM, type CanaryCodersAIOptions as CanaryLLMOptions, type ChatProvider, type ChatStreamEvent, type ChatStreamOptions, type Citation, CompatTarget, type CompleteParams, type CompletionResult, ConflictError, type ContentPart, type ConversationQuestion, type ConversationSession, type ConversationSessionRecord, type ConversationTemplate, type ConversationTemplateParams, type ConversationTemplateType, type ConversationTemplateUpdate, type ConversationTool, type ConversationVoiceSettings, type CostInfo, type CreateConversationSessionParams, CreateRealtimeSessionParams, type DetectParams, type Detection, type DialogueInput, type DialogueParams, type DialogueResult, type DocumentContent, type EmbeddingParams, type EmbeddingResult, type FaceDetectParams, type FaceDetectionResult, FinalizeRealtimeParams, type GeneratedImage, type GeneratedVideo, type ImageContent, type ImageGenerateParams, type ImageGenerationResult, type ImageRecognitionResult, InternalServerError, Job, type JobSnapshot, type JobStatus, type JobStreamOptions, type KeyInfo, type KeyValidation, type Message, type MessageContent, type MessageRole, type ModelCapability, type ModelInfo, type MusicParams, type MusicResult, NotFoundError, type OcrGranularity, type OcrLanguages, type OcrLine, type OcrRecognitionLevel, type OcrRecognizeParams, type OcrResult, type OcrWord, PermissionError, type PollOptions, type PortalPeriod, type ProviderType, RateLimitError, RealtimeSession, RealtimeSessionRecord, type RedactParams, type RedactPreset, type RedactResult, type RedactStyle, type RedactionDetection, type ResponseFormat, type STTResult, type SignedUrlParams, type SignedUrlResult, type SoundEffectParams, type SoundEffectResult, type SpeechParams, type StreamAdapterOptions, type StreamChunk, type StreamProtocol, type TTSResult, type TaskKind, type TextContent, type ThinkingMode, type TokenUsage, type ToolCall, type ToolCallDelta, type ToolChoice, type ToolDefinition, type TrainParams, type TrainingJob, type TranscribeParams, type TranscriptWord, UnprocessableEntityError, type UsageSummary, type VideoContent, type VideoGenerateParams, type VideoGenerationResult, type VideoUploadOptions, type VideoUploadResult, type VisionModel, type VoiceInfo, type VoiceSettings, type WebDetectParams, type WebDetectionResult, type WebEntity, type WebSearchOptions, type ZeroShotDetectParams, toBrokeredCredential };

package/dist/index.d.ts

@@ -66,6 +66,18 @@ interface ApiEnvelope<T> {
     data: T;
     message?: string;
 }
+/**
+ * What a call cost you. Present on every task result (chat and media) and on
+ * the final chunk of a stream. Absent only on results from older gateways.
+ */
+interface CostInfo {
+    /** Customer price in USD for this call; 0 when it ran on your own provider key (BYOK). */
+    customerPrice: number;
+    currency: "USD";
+    billing: "metered" | "byok";
+    /** True when served from the completion cache. */
+    cached?: boolean;
+}
 
 type ConversationTemplateType = "interview" | "conversation";
 interface ConversationQuestion {
@@ -330,6 +342,8 @@ interface CompletionResult {
     toolCalls?: ToolCall[];
     citations?: Citation[];
     metadata?: Record<string, unknown>;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 /** Raw chunk shape emitted by the native `/queue/stream` endpoint. */
 interface StreamChunk {
@@ -338,6 +352,8 @@ interface StreamChunk {
     finishReason?: string;
     toolCallDeltas?: ToolCallDelta[];
     metadata?: Record<string, unknown>;
+    /** Final chunk only: cost of the call. */
+    cost?: CostInfo;
 }
 /** Normalized streaming event, identical across every wire protocol. */
 type ChatStreamEvent = {
@@ -366,6 +382,7 @@ type ChatStreamEvent = {
     type: "done";
     finishReason?: string;
     usage?: TokenUsage;
+    cost?: CostInfo;
     raw?: unknown;
 } | {
     type: "raw";
@@ -488,6 +505,8 @@ interface ImageGenerationResult {
     images: GeneratedImage[];
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface VideoGenerateParams {
     provider: "gemini" | "vertex" | "xai";
@@ -513,6 +532,8 @@ interface VideoGenerationResult {
     videos: GeneratedVideo[];
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface VideoUploadResult {
     fileId: string;
@@ -548,6 +569,8 @@ interface TTSResult {
     model: string;
     provider: string;
     characterCount?: number;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface TranscribeParams {
     provider: "elevenlabs" | "mlxaudio";
@@ -576,6 +599,8 @@ interface STTResult {
     words?: TranscriptWord[];
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface SoundEffectParams {
     text: string;
@@ -591,6 +616,8 @@ interface SoundEffectResult {
     mimeType: string;
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface MusicParams {
     prompt: string;
@@ -605,6 +632,8 @@ interface MusicResult {
     mimeType: string;
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface DialogueInput {
     text: string;
@@ -626,6 +655,8 @@ interface DialogueResult {
     mimeType: string;
     model: string;
     provider: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface EmbeddingParams {
     provider: "lmstudio";
@@ -649,6 +680,8 @@ interface EmbeddingResult {
         inputTokens: number;
         totalTokens: number;
     };
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 
 declare class AudioResource extends BaseResource {
@@ -909,12 +942,16 @@ interface ImageRecognitionResult {
     model: string;
     width?: number;
     height?: number;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface FaceDetectionResult {
     faces: Detection[];
     /** base64-encoded image with faces blurred (when `blur: true`) */
     blurredImage?: string;
     model: string;
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface WebEntity {
     description?: string;
@@ -924,6 +961,8 @@ interface WebDetectionResult {
     entities: WebEntity[];
     pages?: unknown[];
     matchingImages?: unknown[];
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface AutoLabelParams {
     images: string[];
@@ -1033,6 +1072,8 @@ interface OcrResult {
         width: number;
         height: number;
     };
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface RedactParams {
     /** base64-encoded image bytes (max ~50MB) */
@@ -1082,6 +1123,8 @@ interface RedactResult {
         width: number;
         height: number;
     };
+    /** Cost of this call (absent on older gateways). */
+    cost?: CostInfo;
 }
 interface OcrLanguages {
     recognitionLevels: Record<string, string[]>;
@@ -1205,4 +1248,4 @@ interface StreamAdapterOptions extends SSEOptions {
     includeRaw?: boolean;
 }
 
-export { APIConnectionError, APIConnectionTimeoutError, APIError, type ApiEnvelope, type AudioOutputFormat, AuthenticationError, type AutoLabelParams, type AutoLabelResult, type AutoTrainParams, BadRequestError, type BoundingBox, BrokeredCredential, CanaryCodersAI, type CanaryCodersAIOptions, CanaryCodersAI as CanaryLLM, type CanaryCodersAIOptions as CanaryLLMOptions, type ChatProvider, type ChatStreamEvent, type ChatStreamOptions, type Citation, CompatTarget, type CompleteParams, type CompletionResult, ConflictError, type ContentPart, type ConversationQuestion, type ConversationSession, type ConversationSessionRecord, type ConversationTemplate, type ConversationTemplateParams, type ConversationTemplateType, type ConversationTemplateUpdate, type ConversationTool, type ConversationVoiceSettings, type CreateConversationSessionParams, CreateRealtimeSessionParams, type DetectParams, type Detection, type DialogueInput, type DialogueParams, type DialogueResult, type DocumentContent, type EmbeddingParams, type EmbeddingResult, type FaceDetectParams, type FaceDetectionResult, FinalizeRealtimeParams, type GeneratedImage, type GeneratedVideo, type ImageContent, type ImageGenerateParams, type ImageGenerationResult, type ImageRecognitionResult, InternalServerError, Job, type JobSnapshot, type JobStatus, type JobStreamOptions, type KeyInfo, type KeyValidation, type Message, type MessageContent, type MessageRole, type ModelCapability, type ModelInfo, type MusicParams, type MusicResult, NotFoundError, type OcrGranularity, type OcrLanguages, type OcrLine, type OcrRecognitionLevel, type OcrRecognizeParams, type OcrResult, type OcrWord, PermissionError, type PollOptions, type PortalPeriod, type ProviderType, RateLimitError, RealtimeSession, RealtimeSessionRecord, type RedactParams, type RedactPreset, type RedactResult, type RedactStyle, type RedactionDetection, type ResponseFormat, type STTResult, type SignedUrlParams, type SignedUrlResult, type SoundEffectParams, type SoundEffectResult, type SpeechParams, type StreamAdapterOptions, type StreamChunk, type StreamProtocol, type TTSResult, type TaskKind, type TextContent, type ThinkingMode, type TokenUsage, type ToolCall, type ToolCallDelta, type ToolChoice, type ToolDefinition, type TrainParams, type TrainingJob, type TranscribeParams, type TranscriptWord, UnprocessableEntityError, type UsageSummary, type VideoContent, type VideoGenerateParams, type VideoGenerationResult, type VideoUploadOptions, type VideoUploadResult, type VisionModel, type VoiceInfo, type VoiceSettings, type WebDetectParams, type WebDetectionResult, type WebEntity, type WebSearchOptions, type ZeroShotDetectParams, CanaryCodersAI as default, toBrokeredCredential };
+export { APIConnectionError, APIConnectionTimeoutError, APIError, type ApiEnvelope, type AudioOutputFormat, AuthenticationError, type AutoLabelParams, type AutoLabelResult, type AutoTrainParams, BadRequestError, type BoundingBox, BrokeredCredential, CanaryCodersAI, type CanaryCodersAIOptions, CanaryCodersAI as CanaryLLM, type CanaryCodersAIOptions as CanaryLLMOptions, type ChatProvider, type ChatStreamEvent, type ChatStreamOptions, type Citation, CompatTarget, type CompleteParams, type CompletionResult, ConflictError, type ContentPart, type ConversationQuestion, type ConversationSession, type ConversationSessionRecord, type ConversationTemplate, type ConversationTemplateParams, type ConversationTemplateType, type ConversationTemplateUpdate, type ConversationTool, type ConversationVoiceSettings, type CostInfo, type CreateConversationSessionParams, CreateRealtimeSessionParams, type DetectParams, type Detection, type DialogueInput, type DialogueParams, type DialogueResult, type DocumentContent, type EmbeddingParams, type EmbeddingResult, type FaceDetectParams, type FaceDetectionResult, FinalizeRealtimeParams, type GeneratedImage, type GeneratedVideo, type ImageContent, type ImageGenerateParams, type ImageGenerationResult, type ImageRecognitionResult, InternalServerError, Job, type JobSnapshot, type JobStatus, type JobStreamOptions, type KeyInfo, type KeyValidation, type Message, type MessageContent, type MessageRole, type ModelCapability, type ModelInfo, type MusicParams, type MusicResult, NotFoundError, type OcrGranularity, type OcrLanguages, type OcrLine, type OcrRecognitionLevel, type OcrRecognizeParams, type OcrResult, type OcrWord, PermissionError, type PollOptions, type PortalPeriod, type ProviderType, RateLimitError, RealtimeSession, RealtimeSessionRecord, type RedactParams, type RedactPreset, type RedactResult, type RedactStyle, type RedactionDetection, type ResponseFormat, type STTResult, type SignedUrlParams, type SignedUrlResult, type SoundEffectParams, type SoundEffectResult, type SpeechParams, type StreamAdapterOptions, type StreamChunk, type StreamProtocol, type TTSResult, type TaskKind, type TextContent, type ThinkingMode, type TokenUsage, type ToolCall, type ToolCallDelta, type ToolChoice, type ToolDefinition, type TrainParams, type TrainingJob, type TranscribeParams, type TranscriptWord, UnprocessableEntityError, type UsageSummary, type VideoContent, type VideoGenerateParams, type VideoGenerationResult, type VideoUploadOptions, type VideoUploadResult, type VisionModel, type VoiceInfo, type VoiceSettings, type WebDetectParams, type WebDetectionResult, type WebEntity, type WebSearchOptions, type ZeroShotDetectParams, CanaryCodersAI as default, toBrokeredCredential };

package/dist/index.js

@@ -678,6 +678,7 @@ async function* nativeQueueAdapter(frames, includeRaw) {
   const raw = (v) => includeRaw ? v : void 0;
   let lastUsage;
   let lastFinish;
+  let lastCost;
   for await (const frame of frames) {
     switch (frame.event) {
       case "start":
@@ -686,11 +687,12 @@ async function* nativeQueueAdapter(frames, includeRaw) {
       case "error":
         throw streamError(tryParse(frame.data));
       case "done":
-        yield { type: "done", finishReason: lastFinish, usage: lastUsage };
+        yield { type: "done", finishReason: lastFinish, usage: lastUsage, cost: lastCost };
         return;
       case "chunk": {
         const chunk = tryParse(frame.data);
         if (!chunk) break;
+        if (chunk.cost) lastCost = chunk.cost;
         if (chunk.usage) {
           lastUsage = chunk.usage;
           yield { type: "usage", usage: chunk.usage, raw: raw(chunk) };
@@ -718,7 +720,7 @@ async function* nativeQueueAdapter(frames, includeRaw) {
       }
     }
   }
-  yield { type: "done", finishReason: lastFinish, usage: lastUsage };
+  yield { type: "done", finishReason: lastFinish, usage: lastUsage, cost: lastCost };
 }
 async function* openAIAdapter(frames, includeRaw) {
   const raw = (v) => includeRaw ? v : void 0;