@ax-llm/ax 19.0.42 → 19.0.43

package/index.d.cts

@@ -191,6 +191,16 @@ type AxModelInfo = {
     characterIsToken?: boolean;
     promptTokenCostPer1M?: number;
     completionTokenCostPer1M?: number;
+    cacheReadTokenCostPer1M?: number;
+    cacheWriteTokenCostPer1M?: number;
+    /** Prompt token cost for requests exceeding longContextThreshold */
+    longContextPromptTokenCostPer1M?: number;
+    /** Completion token cost for requests exceeding longContextThreshold */
+    longContextCompletionTokenCostPer1M?: number;
+    /** Cache read token cost for requests exceeding longContextThreshold */
+    longContextCacheReadTokenCostPer1M?: number;
+    /** Total input token count (including cached) above which long-context rates apply */
+    longContextThreshold?: number;
     aliases?: string[];
     supported?: {
         thinkingBudget?: boolean;
@@ -433,6 +443,7 @@ type AxDebugChatResponseUsage = AxModelUsage & {
     mutableChatContextCharacters?: number;
     chatContextCharacters?: number;
     totalPromptCharacters?: number;
+    estimatedCost?: number;
 };
 type AxChatResponse = {
     sessionId?: string;
@@ -971,6 +982,7 @@ interface AxAIService<TModel = unknown, TEmbedModel = unknown, TModelKey = strin
     getLastUsedModelConfig(): AxModelConfig | undefined;
     chat(req: Readonly<AxChatRequest<TModel | TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxChatResponse | ReadableStream<AxChatResponse>>;
     embed(req: Readonly<AxEmbedRequest<TEmbedModel | TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxEmbedResponse>;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     setOptions(options: Readonly<AxAIServiceOptions>): void;
     getOptions(): Readonly<AxAIServiceOptions>;
 }
@@ -3883,7 +3895,6 @@ declare class AxBaseAI<TModel, TEmbedModel, TChatRequest, TEmbedRequest, TChatRe
     private calculateResponseSize;
     private detectMultimodalContent;
     private calculateContextWindowUsage;
-    private estimateCost;
     private estimateCostByName;
     private recordFunctionCallMetrics;
     private recordTimeoutMetric;
@@ -3891,6 +3902,7 @@ declare class AxBaseAI<TModel, TEmbedModel, TChatRequest, TEmbedRequest, TChatRe
     private recordChatMetrics;
     private recordEmbedMetrics;
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     chat(req: Readonly<AxChatRequest<TModel | TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxChatResponse | ReadableStream<AxChatResponse>>;
     private _chat1;
     private cleanupFunctionSchema;
@@ -4527,7 +4539,9 @@ interface AxAIOpenAIArgs<TName = 'openai', TModel = AxAIOpenAIModel, TEmbedModel
     modelInfo?: AxModelInfo[];
     config?: Partial<AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq>['config']>;
 }
-type ChatReqUpdater<TModel, TChatReq extends AxAIOpenAIChatRequest<TModel>> = (req: Readonly<TChatReq>) => TChatReq;
+type ChatReqUpdater<TModel, TChatReq extends AxAIOpenAIChatRequest<TModel>> = (req: Readonly<TChatReq>, config: Readonly<AxAIServiceOptions>) => TChatReq;
+type ChatRespProcessor = (resp: AxChatResponse) => AxChatResponse;
+type ChatStreamRespProcessor = (resp: AxChatResponse, state: object) => AxChatResponse;
 interface AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq extends AxAIOpenAIChatRequest<TModel>> {
     apiKey: string;
     apiURL?: string;
@@ -4538,10 +4552,12 @@ interface AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq extends Ax
     modelInfo: Readonly<AxModelInfo[]>;
     models?: AxAIInputModelList<TModel, TEmbedModel, TModelKey>;
     chatReqUpdater?: ChatReqUpdater<TModel, TChatReq>;
+    chatRespProcessor?: ChatRespProcessor;
+    chatStreamRespProcessor?: ChatStreamRespProcessor;
     supportFor: AxAIFeatures | ((model: TModel) => AxAIFeatures);
 }
 declare class AxAIOpenAIBase<TModel, TEmbedModel, TModelKey, TChatReq extends AxAIOpenAIChatRequest<TModel> = AxAIOpenAIChatRequest<TModel>> extends AxBaseAI<TModel, TEmbedModel, AxAIOpenAIChatRequest<TModel>, AxAIOpenAIEmbedRequest<TEmbedModel>, AxAIOpenAIChatResponse, AxAIOpenAIChatResponseDelta, AxAIOpenAIEmbedResponse, TModelKey> {
-    constructor({ apiKey, config, options, apiURL, modelInfo, models, chatReqUpdater, supportFor, }: Readonly<Omit<AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq>, 'name'>>);
+    constructor({ apiKey, config, options, apiURL, modelInfo, models, chatReqUpdater, chatRespProcessor, chatStreamRespProcessor, supportFor, }: Readonly<Omit<AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq>, 'name'>>);
 }
 declare class AxAIOpenAI<TModelKey = string> extends AxAIOpenAIBase<AxAIOpenAIModel, AxAIOpenAIEmbedModel, TModelKey> {
     constructor({ apiKey, apiURL, config, options, models, modelInfo, }: Readonly<Omit<AxAIOpenAIArgs<'openai', AxAIOpenAIModel, AxAIOpenAIEmbedModel, TModelKey>, 'name'>>);
@@ -4611,6 +4627,7 @@ declare class AxBalancer<TServices extends readonly AxAIService<any, any, any>[]
     getId(): string;
     getFeatures(model?: string): AxAIFeatures;
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     private canRetryService;
     private handleFailure;
     private handleSuccess;
@@ -5677,6 +5694,7 @@ declare class AxMockAIService<TModelKey> implements AxAIService<unknown, unknown
     getFeatures(_model?: string): AxAIFeatures;
     getModelList(): AxAIModelList<TModelKey> | undefined;
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(): number;
     chat(req: Readonly<AxChatRequest<unknown>>, _options?: Readonly<AxAIServiceOptions>): Promise<AxChatResponse | ReadableStream<AxChatResponse>>;
     embed(req: Readonly<AxEmbedRequest>, _options?: Readonly<AxAIServiceOptions>): Promise<AxEmbedResponse>;
     setOptions(options: Readonly<AxAIServiceOptions>): void;
@@ -5744,6 +5762,7 @@ declare class AxMultiServiceRouter<TServices extends readonly (AxAIService | AxA
      * or falls back to the first service if none has been used.
      */
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     /**
      * Sets options on all underlying services.
      */
@@ -5776,6 +5795,9 @@ declare class AxMultiServiceRouter<TServices extends readonly (AxAIService | AxA
  * Configuration type for Ollama AI service
  */
 type AxAIOllamaAIConfig = AxAIOpenAIConfig<string, string>;
+type AxAIOllamaChatRequest = AxAIOpenAIChatRequest<string> & {
+    think?: boolean;
+};
 /**
  * Creates default configuration for Ollama AI service
  * @returns Default configuration object with nous-hermes2 model and all-minilm embed model
@@ -5800,7 +5822,7 @@ type AxAIOllamaArgs<TModelKey> = AxAIOpenAIArgs<'ollama', string, string, TModel
  * Provides access to locally hosted Ollama models with OpenAI-compatible API
  * @template TModelKey - Type for model key
  */
-declare class AxAIOllama<TModelKey> extends AxAIOpenAIBase<string, string, TModelKey> {
+declare class AxAIOllama<TModelKey> extends AxAIOpenAIBase<string, string, TModelKey, AxAIOllamaChatRequest> {
     /**
      * Creates a new Ollama AI service instance
      * @param args - Configuration arguments for the Ollama service
@@ -7116,6 +7138,7 @@ declare class AxAI<TModelKey = string> implements AxAIService<any, any, TModelKe
     getLastUsedEmbedModel(): any;
     getLastUsedModelConfig(): AxModelConfig | undefined;
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     chat(req: Readonly<AxChatRequest<TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxChatResponse | ReadableStream<AxChatResponse>>;
     embed(req: Readonly<AxEmbedRequest<TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxEmbedResponse>;
     setOptions(options: Readonly<AxAIServiceOptions>): void;

package/index.d.ts

@@ -191,6 +191,16 @@ type AxModelInfo = {
     characterIsToken?: boolean;
     promptTokenCostPer1M?: number;
     completionTokenCostPer1M?: number;
+    cacheReadTokenCostPer1M?: number;
+    cacheWriteTokenCostPer1M?: number;
+    /** Prompt token cost for requests exceeding longContextThreshold */
+    longContextPromptTokenCostPer1M?: number;
+    /** Completion token cost for requests exceeding longContextThreshold */
+    longContextCompletionTokenCostPer1M?: number;
+    /** Cache read token cost for requests exceeding longContextThreshold */
+    longContextCacheReadTokenCostPer1M?: number;
+    /** Total input token count (including cached) above which long-context rates apply */
+    longContextThreshold?: number;
     aliases?: string[];
     supported?: {
         thinkingBudget?: boolean;
@@ -433,6 +443,7 @@ type AxDebugChatResponseUsage = AxModelUsage & {
     mutableChatContextCharacters?: number;
     chatContextCharacters?: number;
     totalPromptCharacters?: number;
+    estimatedCost?: number;
 };
 type AxChatResponse = {
     sessionId?: string;
@@ -971,6 +982,7 @@ interface AxAIService<TModel = unknown, TEmbedModel = unknown, TModelKey = strin
     getLastUsedModelConfig(): AxModelConfig | undefined;
     chat(req: Readonly<AxChatRequest<TModel | TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxChatResponse | ReadableStream<AxChatResponse>>;
     embed(req: Readonly<AxEmbedRequest<TEmbedModel | TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxEmbedResponse>;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     setOptions(options: Readonly<AxAIServiceOptions>): void;
     getOptions(): Readonly<AxAIServiceOptions>;
 }
@@ -3883,7 +3895,6 @@ declare class AxBaseAI<TModel, TEmbedModel, TChatRequest, TEmbedRequest, TChatRe
     private calculateResponseSize;
     private detectMultimodalContent;
     private calculateContextWindowUsage;
-    private estimateCost;
     private estimateCostByName;
     private recordFunctionCallMetrics;
     private recordTimeoutMetric;
@@ -3891,6 +3902,7 @@ declare class AxBaseAI<TModel, TEmbedModel, TChatRequest, TEmbedRequest, TChatRe
     private recordChatMetrics;
     private recordEmbedMetrics;
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     chat(req: Readonly<AxChatRequest<TModel | TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxChatResponse | ReadableStream<AxChatResponse>>;
     private _chat1;
     private cleanupFunctionSchema;
@@ -4527,7 +4539,9 @@ interface AxAIOpenAIArgs<TName = 'openai', TModel = AxAIOpenAIModel, TEmbedModel
     modelInfo?: AxModelInfo[];
     config?: Partial<AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq>['config']>;
 }
-type ChatReqUpdater<TModel, TChatReq extends AxAIOpenAIChatRequest<TModel>> = (req: Readonly<TChatReq>) => TChatReq;
+type ChatReqUpdater<TModel, TChatReq extends AxAIOpenAIChatRequest<TModel>> = (req: Readonly<TChatReq>, config: Readonly<AxAIServiceOptions>) => TChatReq;
+type ChatRespProcessor = (resp: AxChatResponse) => AxChatResponse;
+type ChatStreamRespProcessor = (resp: AxChatResponse, state: object) => AxChatResponse;
 interface AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq extends AxAIOpenAIChatRequest<TModel>> {
     apiKey: string;
     apiURL?: string;
@@ -4538,10 +4552,12 @@ interface AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq extends Ax
     modelInfo: Readonly<AxModelInfo[]>;
     models?: AxAIInputModelList<TModel, TEmbedModel, TModelKey>;
     chatReqUpdater?: ChatReqUpdater<TModel, TChatReq>;
+    chatRespProcessor?: ChatRespProcessor;
+    chatStreamRespProcessor?: ChatStreamRespProcessor;
     supportFor: AxAIFeatures | ((model: TModel) => AxAIFeatures);
 }
 declare class AxAIOpenAIBase<TModel, TEmbedModel, TModelKey, TChatReq extends AxAIOpenAIChatRequest<TModel> = AxAIOpenAIChatRequest<TModel>> extends AxBaseAI<TModel, TEmbedModel, AxAIOpenAIChatRequest<TModel>, AxAIOpenAIEmbedRequest<TEmbedModel>, AxAIOpenAIChatResponse, AxAIOpenAIChatResponseDelta, AxAIOpenAIEmbedResponse, TModelKey> {
-    constructor({ apiKey, config, options, apiURL, modelInfo, models, chatReqUpdater, supportFor, }: Readonly<Omit<AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq>, 'name'>>);
+    constructor({ apiKey, config, options, apiURL, modelInfo, models, chatReqUpdater, chatRespProcessor, chatStreamRespProcessor, supportFor, }: Readonly<Omit<AxAIOpenAIBaseArgs<TModel, TEmbedModel, TModelKey, TChatReq>, 'name'>>);
 }
 declare class AxAIOpenAI<TModelKey = string> extends AxAIOpenAIBase<AxAIOpenAIModel, AxAIOpenAIEmbedModel, TModelKey> {
     constructor({ apiKey, apiURL, config, options, models, modelInfo, }: Readonly<Omit<AxAIOpenAIArgs<'openai', AxAIOpenAIModel, AxAIOpenAIEmbedModel, TModelKey>, 'name'>>);
@@ -4611,6 +4627,7 @@ declare class AxBalancer<TServices extends readonly AxAIService<any, any, any>[]
     getId(): string;
     getFeatures(model?: string): AxAIFeatures;
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     private canRetryService;
     private handleFailure;
     private handleSuccess;
@@ -5677,6 +5694,7 @@ declare class AxMockAIService<TModelKey> implements AxAIService<unknown, unknown
     getFeatures(_model?: string): AxAIFeatures;
     getModelList(): AxAIModelList<TModelKey> | undefined;
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(): number;
     chat(req: Readonly<AxChatRequest<unknown>>, _options?: Readonly<AxAIServiceOptions>): Promise<AxChatResponse | ReadableStream<AxChatResponse>>;
     embed(req: Readonly<AxEmbedRequest>, _options?: Readonly<AxAIServiceOptions>): Promise<AxEmbedResponse>;
     setOptions(options: Readonly<AxAIServiceOptions>): void;
@@ -5744,6 +5762,7 @@ declare class AxMultiServiceRouter<TServices extends readonly (AxAIService | AxA
      * or falls back to the first service if none has been used.
      */
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     /**
      * Sets options on all underlying services.
      */
@@ -5776,6 +5795,9 @@ declare class AxMultiServiceRouter<TServices extends readonly (AxAIService | AxA
  * Configuration type for Ollama AI service
  */
 type AxAIOllamaAIConfig = AxAIOpenAIConfig<string, string>;
+type AxAIOllamaChatRequest = AxAIOpenAIChatRequest<string> & {
+    think?: boolean;
+};
 /**
  * Creates default configuration for Ollama AI service
  * @returns Default configuration object with nous-hermes2 model and all-minilm embed model
@@ -5800,7 +5822,7 @@ type AxAIOllamaArgs<TModelKey> = AxAIOpenAIArgs<'ollama', string, string, TModel
  * Provides access to locally hosted Ollama models with OpenAI-compatible API
  * @template TModelKey - Type for model key
  */
-declare class AxAIOllama<TModelKey> extends AxAIOpenAIBase<string, string, TModelKey> {
+declare class AxAIOllama<TModelKey> extends AxAIOpenAIBase<string, string, TModelKey, AxAIOllamaChatRequest> {
     /**
      * Creates a new Ollama AI service instance
      * @param args - Configuration arguments for the Ollama service
@@ -7116,6 +7138,7 @@ declare class AxAI<TModelKey = string> implements AxAIService<any, any, TModelKe
     getLastUsedEmbedModel(): any;
     getLastUsedModelConfig(): AxModelConfig | undefined;
     getMetrics(): AxAIServiceMetrics;
+    getEstimatedCost(modelUsage?: AxModelUsage): number;
     chat(req: Readonly<AxChatRequest<TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxChatResponse | ReadableStream<AxChatResponse>>;
     embed(req: Readonly<AxEmbedRequest<TModelKey>>, options?: Readonly<AxAIServiceOptions>): Promise<AxEmbedResponse>;
     setOptions(options: Readonly<AxAIServiceOptions>): void;