@jaypie/llm 1.2.1 → 1.2.3

package/dist/esm/operate/OperateLoop.d.ts

@@ -1,4 +1,4 @@
-import { LlmHistory, LlmInputMessage, LlmOperateOptions, LlmOperateResponse } from "../types/LlmProvider.interface.js";
+import { LlmHistory, LlmInputMessage, LlmOperateInput, LlmOperateOptions, LlmOperateResponse } from "../types/LlmProvider.interface.js";
 import { ProviderAdapter } from "./adapters/ProviderAdapter.interface.js";
 import { HookRunner } from "./hooks/index.js";
 import { InputProcessor } from "./input/index.js";
@@ -28,7 +28,7 @@ export declare class OperateLoop {
     /**
      * Execute the operate loop for multi-turn conversations with tool calling.
      */
-    execute(input: string | LlmHistory | LlmInputMessage, options?: LlmOperateOptions): Promise<LlmOperateResponse>;
+    execute(input: string | LlmHistory | LlmInputMessage | LlmOperateInput, options?: LlmOperateOptions): Promise<LlmOperateResponse>;
     private initializeState;
     private createContext;
     private buildInitialRequest;

package/dist/esm/operate/StreamLoop.d.ts

@@ -1,4 +1,4 @@
-import { LlmHistory, LlmInputMessage, LlmOperateOptions } from "../types/LlmProvider.interface.js";
+import { LlmHistory, LlmInputMessage, LlmOperateInput, LlmOperateOptions } from "../types/LlmProvider.interface.js";
 import { LlmStreamChunk } from "../types/LlmStreamChunk.interface.js";
 import { ProviderAdapter } from "./adapters/ProviderAdapter.interface.js";
 import { HookRunner } from "./hooks/index.js";
@@ -24,7 +24,7 @@ export declare class StreamLoop {
      * Execute the streaming loop for multi-turn conversations with tool calling.
      * Yields stream chunks as they become available.
      */
-    execute(input: string | LlmHistory | LlmInputMessage, options?: LlmOperateOptions): AsyncIterable<LlmStreamChunk>;
+    execute(input: string | LlmHistory | LlmInputMessage | LlmOperateInput, options?: LlmOperateOptions): AsyncIterable<LlmStreamChunk>;
     private initializeState;
     private createContext;
     private buildInitialRequest;

package/dist/esm/operate/adapters/AnthropicAdapter.d.ts

@@ -13,7 +13,7 @@ import { BaseProviderAdapter } from "./ProviderAdapter.interface.js";
  */
 export declare class AnthropicAdapter extends BaseProviderAdapter {
     readonly name: "anthropic";
-    readonly defaultModel: "claude-opus-4-1";
+    readonly defaultModel: "claude-sonnet-4-5";
     buildRequest(request: OperateRequest): Anthropic.MessageCreateParams;
     formatTools(toolkit: Toolkit, outputSchema?: JsonObject): ProviderToolDefinition[];
     formatOutputSchema(schema: JsonObject | NaturalSchema | z.ZodType): JsonObject;

package/dist/esm/operate/adapters/GeminiAdapter.d.ts

@@ -13,7 +13,7 @@ import { GeminiPart, GeminiRawResponse, GeminiRequest } from "../../providers/ge
  */
 export declare class GeminiAdapter extends BaseProviderAdapter {
     readonly name: "gemini";
-    readonly defaultModel: "gemini-2.5-flash";
+    readonly defaultModel: "gemini-3-pro-preview";
     buildRequest(request: OperateRequest): GeminiRequest;
     formatTools(toolkit: Toolkit, outputSchema?: JsonObject): ProviderToolDefinition[];
     formatOutputSchema(schema: JsonObject | NaturalSchema | z.ZodType): JsonObject;
@@ -30,6 +30,7 @@ export declare class GeminiAdapter extends BaseProviderAdapter {
     hasStructuredOutput(response: unknown): boolean;
     extractStructuredOutput(response: unknown): JsonObject | undefined;
     private convertMessagesToContents;
+    private convertContentToParts;
     private mapRole;
     private hasToolCalls;
     private getFinishReason;

package/dist/esm/operate/adapters/OpenAiAdapter.d.ts

@@ -12,7 +12,7 @@ import { BaseProviderAdapter } from "./ProviderAdapter.interface.js";
  */
 export declare class OpenAiAdapter extends BaseProviderAdapter {
     readonly name: "openai";
-    readonly defaultModel: "gpt-5";
+    readonly defaultModel: "gpt-5.1";
     buildRequest(request: OperateRequest): unknown;
     formatTools(toolkit: Toolkit, _outputSchema?: JsonObject): ProviderToolDefinition[];
     formatOutputSchema(schema: JsonObject | NaturalSchema | z.ZodType): JsonObject;

package/dist/esm/operate/adapters/OpenRouterAdapter.d.ts

@@ -7,7 +7,7 @@ import { ClassifiedError, OperateRequest, ParsedResponse, ProviderToolDefinition
 import { BaseProviderAdapter } from "./ProviderAdapter.interface.js";
 interface OpenRouterMessage {
     role: "system" | "user" | "assistant" | "tool";
-    content?: string | null;
+    content?: string | OpenRouterContentPart[] | null;
     toolCalls?: OpenRouterToolCall[];
     toolCallId?: string;
 }
@@ -39,6 +39,9 @@ interface OpenRouterUsage {
     prompt_tokens?: number;
     completion_tokens?: number;
     total_tokens?: number;
+    completionTokensDetails?: {
+        reasoningTokens?: number;
+    };
 }
 interface OpenRouterResponse {
     id: string;
@@ -66,6 +69,13 @@ interface OpenRouterRequest {
     };
     user?: string;
 }
+/**
+ * OpenRouter content part types (text only - images/files not supported)
+ */
+type OpenRouterContentPart = {
+    type: "text";
+    text: string;
+};
 /**
  * OpenRouterAdapter implements the ProviderAdapter interface for OpenRouter's API.
  * OpenRouter provides a unified API to access hundreds of AI models through OpenAI-compatible endpoints.
@@ -74,7 +84,7 @@ interface OpenRouterRequest {
  */
 export declare class OpenRouterAdapter extends BaseProviderAdapter {
     readonly name: "openrouter";
-    readonly defaultModel: "openai/gpt-4o";
+    readonly defaultModel: "z-ai/glm-4.7";
     buildRequest(request: OperateRequest): OpenRouterRequest;
     formatTools(toolkit: Toolkit, outputSchema?: JsonObject): ProviderToolDefinition[];
     formatOutputSchema(schema: JsonObject | NaturalSchema | z.ZodType): JsonObject;

package/dist/esm/operate/input/InputProcessor.d.ts

@@ -1,4 +1,4 @@
-import { LlmHistory, LlmInputMessage, LlmOperateOptions } from "../../types/LlmProvider.interface.js";
+import { LlmHistory, LlmInputMessage, LlmOperateInput, LlmOperateOptions } from "../../types/LlmProvider.interface.js";
 export interface ProcessedInput {
     /** The processed history with all messages formatted */
     history: LlmHistory;
@@ -15,11 +15,11 @@ export declare class InputProcessor {
     /**
      * Process input with placeholders, history merging, and system message handling
      *
-     * @param input - The raw input (string, message, or history)
+     * @param input - The raw input (string, message, history, or LlmOperateInput)
      * @param options - The operate options containing data, history, system, etc.
      * @returns Processed input with all transformations applied
      */
-    process(input: string | LlmHistory | LlmInputMessage, options?: LlmOperateOptions): ProcessedInput;
+    process(input: string | LlmHistory | LlmInputMessage | LlmOperateInput, options?: LlmOperateOptions): Promise<ProcessedInput>;
     /**
      * Format input and apply placeholders if data is provided
      */

package/dist/esm/operate/response/ResponseBuilder.d.ts

@@ -45,6 +45,18 @@ export declare class ResponseBuilder {
      * Append items to the output
      */
     appendToOutput(...items: LlmOutput): this;
+    /**
+     * Set the reasoning array
+     */
+    setReasoning(reasoning: string[]): this;
+    /**
+     * Append reasoning text items
+     */
+    appendToReasoning(...items: string[]): this;
+    /**
+     * Get the current reasoning array
+     */
+    getReasoning(): string[];
     /**
      * Add a raw provider response
      */
@@ -78,7 +90,8 @@ export declare class ResponseBuilder {
      */
     incomplete(): this;
     /**
-     * Build and return the final response object
+     * Build and return the final response object.
+     * Automatically extracts reasoning from history if not already set.
      */
     build(): LlmOperateResponse;
 }

package/dist/esm/types/LlmOperateInput.guards.d.ts

@@ -0,0 +1,17 @@
+import type { LlmOperateInput, LlmOperateInputContent, LlmOperateInputFile, LlmOperateInputImage } from "./LlmProvider.interface.js";
+/**
+ * Type guard to check if an item is an LlmOperateInputFile
+ */
+export declare function isLlmOperateInputFile(item: unknown): item is LlmOperateInputFile;
+/**
+ * Type guard to check if an item is an LlmOperateInputImage
+ */
+export declare function isLlmOperateInputImage(item: unknown): item is LlmOperateInputImage;
+/**
+ * Type guard to check if an item is an LlmOperateInputContent
+ */
+export declare function isLlmOperateInputContent(item: unknown): item is LlmOperateInputContent;
+/**
+ * Type guard to check if input is an LlmOperateInput array
+ */
+export declare function isLlmOperateInput(input: unknown): input is LlmOperateInput;

package/dist/esm/types/LlmProvider.interface.d.ts

@@ -56,6 +56,51 @@ export interface LlmInputMessage {
     role: LlmMessageRole;
     type?: LlmMessageType.Message;
 }
+/**
+ * File input that can be loaded from local filesystem, S3, or provided as base64 data.
+ * For PDFs, specific pages can be extracted via the `pages` property.
+ * If the file extension is an image type, it will be treated as an image internally.
+ */
+export interface LlmOperateInputFile {
+    /** Path or filename (extension determines type via /\.(\w+)$/) */
+    file: string;
+    /** S3 bucket name (if present, load from S3) */
+    bucket?: string;
+    /** Specific pages to extract from PDF (omit = all pages) */
+    pages?: number[];
+    /** Base64 data (if present, use directly instead of loading from file) */
+    data?: string;
+}
+/**
+ * Image input that can be loaded from local filesystem, S3, or provided as base64 data.
+ */
+export interface LlmOperateInputImage {
+    /** Path or filename */
+    image: string;
+    /** S3 bucket name (if present, load from S3) */
+    bucket?: string;
+    /** Base64 data (if present, use directly instead of loading from file) */
+    data?: string;
+}
+/**
+ * Content item in LlmOperateInput array.
+ * Can be a string (text), file object, or image object.
+ */
+export type LlmOperateInputContent = string | LlmOperateInputFile | LlmOperateInputImage;
+/**
+ * Simplified input format for operate() that supports automatic file resolution.
+ * Files/images are loaded from local filesystem, S3, or provided data.
+ *
+ * @example
+ * ```typescript
+ * const input: LlmOperateInput = [
+ *   "Extract text from these documents",
+ *   { file: "document.pdf", bucket: "my-bucket", pages: [1, 2, 3] },
+ *   { image: "photo.png" },  // loads from local filesystem
+ * ];
+ * ```
+ */
+export type LlmOperateInput = LlmOperateInputContent[];
 export interface LlmOutputContentText {
     annotations?: AnyValue[];
     text: string;
@@ -188,13 +233,14 @@ export interface LlmOperateResponse {
     model?: string;
     output: LlmOutput;
     provider?: string;
+    reasoning: string[];
     responses: JsonReturn[];
     status: LlmResponseStatus;
     usage: LlmUsage;
 }
 export interface LlmProvider {
-    operate?(input: string | LlmHistory | LlmInputMessage, options?: LlmOperateOptions): Promise<LlmOperateResponse>;
+    operate?(input: string | LlmHistory | LlmInputMessage | LlmOperateInput, options?: LlmOperateOptions): Promise<LlmOperateResponse>;
     send(message: string, options?: LlmMessageOptions): Promise<string | JsonObject>;
-    stream?(input: string | LlmHistory | LlmInputMessage, options?: LlmOperateOptions): AsyncIterable<LlmStreamChunk>;
+    stream?(input: string | LlmHistory | LlmInputMessage | LlmOperateInput, options?: LlmOperateOptions): AsyncIterable<LlmStreamChunk>;
 }
 export {};

package/dist/esm/upload/detectMimeType.d.ts

@@ -0,0 +1,16 @@
+/**
+ * Extract file extension from a path or filename
+ */
+export declare function getFileExtension(filePath: string): string | undefined;
+/**
+ * Determine MIME type from file extension
+ */
+export declare function getMimeType(filePath: string): string | undefined;
+/**
+ * Check if a file path has an image extension
+ */
+export declare function isImageExtension(filePath: string): boolean;
+/**
+ * Check if a file path has a PDF extension
+ */
+export declare function isPdfExtension(filePath: string): boolean;

package/dist/esm/upload/extractPdfPages.d.ts

@@ -0,0 +1,13 @@
+/**
+ * Extract specific pages from a PDF buffer
+ * @param pdfBytes - Buffer containing the PDF data
+ * @param pages - Array of page numbers to extract (1-indexed)
+ * @returns Buffer containing a new PDF with only the specified pages
+ */
+export declare function extractPdfPages(pdfBytes: Buffer, pages: number[]): Promise<Buffer>;
+/**
+ * Get the total number of pages in a PDF
+ * @param pdfBytes - Buffer containing the PDF data
+ * @returns Total number of pages
+ */
+export declare function getPdfPageCount(pdfBytes: Buffer): Promise<number>;

package/dist/esm/upload/index.d.ts

@@ -0,0 +1,5 @@
+export { getFileExtension, getMimeType, isImageExtension, isPdfExtension, } from "./detectMimeType.js";
+export { extractPdfPages, getPdfPageCount } from "./extractPdfPages.js";
+export { loadLocalFile } from "./loadLocalFile.js";
+export { loadS3File } from "./loadS3File.js";
+export { resolveOperateInput } from "./resolveOperateInput.js";

package/dist/esm/upload/loadLocalFile.d.ts

@@ -0,0 +1,6 @@
+/**
+ * Load a file from the local filesystem
+ * @param filePath - Path to the file (relative paths resolve from process.cwd())
+ * @returns Buffer containing the file contents
+ */
+export declare function loadLocalFile(filePath: string): Promise<Buffer>;

package/dist/esm/upload/loadS3File.d.ts

@@ -0,0 +1,7 @@
+/**
+ * Load a file from S3
+ * @param bucket - S3 bucket name
+ * @param key - S3 object key
+ * @returns Buffer containing the file contents
+ */
+export declare function loadS3File(bucket: string, key: string): Promise<Buffer>;

package/dist/esm/upload/resolveOperateInput.d.ts

@@ -0,0 +1,12 @@
+import { LlmInputMessage, LlmOperateInput } from "../types/LlmProvider.interface.js";
+/**
+ * Resolve LlmOperateInput to LlmInputMessage
+ *
+ * This function takes the simplified LlmOperateInput format and resolves
+ * all files/images from their sources (provided data, S3, or local filesystem),
+ * converting them to the standard LlmInputMessage format.
+ *
+ * @param input - LlmOperateInput array
+ * @returns LlmInputMessage ready for LLM providers
+ */
+export declare function resolveOperateInput(input: LlmOperateInput): Promise<LlmInputMessage>;

package/dist/esm/util/extractReasoning.d.ts

@@ -0,0 +1,14 @@
+import { LlmHistory } from "../types/LlmProvider.interface.js";
+/**
+ * Extracts reasoning text from LLM history items.
+ *
+ * Reasoning items may have text in different locations depending on the provider:
+ * - OpenAI: `summary` array with objects containing `text` property (type: "reasoning")
+ * - OpenAI: `content` property on reasoning items (type: "reasoning")
+ * - Anthropic: `thinking` property on thinking items (type: "thinking")
+ * - OpenRouter/z-ai: `reasoning` property on message items
+ *
+ * @param history - The LLM history array to extract reasoning from
+ * @returns Array of reasoning text strings
+ */
+export declare function extractReasoning(history: LlmHistory): string[];

package/dist/esm/util/index.d.ts

@@ -1,4 +1,5 @@
 export * from "./determineModelProvider.js";
+export * from "./extractReasoning.js";
 export * from "./formatOperateInput.js";
 export * from "./formatOperateMessage.js";
 export * from "./logger.js";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jaypie/llm",
-  "version": "1.2.1",
+  "version": "1.2.3",
   "description": "Large language model utilities",
   "repository": {
     "type": "git",
@@ -36,6 +36,7 @@
     "@jaypie/errors": "^1.2.1",
     "openai": "^6.9.1",
     "openmeteo": "^1.2.1",
+    "pdf-lib": "^1.17.1",
     "random": "^5.3.0",
     "z-schema": "^6.0.2",
     "zod": "^4.1.13"