@aigne/core 1.61.0-beta → 1.61.0-beta.2

package/CHANGELOG.md

@@ -1,5 +1,19 @@
 # Changelog
 
+## [1.61.0-beta.2](https://github.com/AIGNE-io/aigne-framework/compare/core-v1.61.0-beta.1...core-v1.61.0-beta.2) (2025-09-23)
+
+
+### Bug Fixes
+
+* add prefer input file type option for image model ([#536](https://github.com/AIGNE-io/aigne-framework/issues/536)) ([3cba8a5](https://github.com/AIGNE-io/aigne-framework/commit/3cba8a5562233a1567b49b6dd5c446c0760f5c4c))
+
+## [1.61.0-beta.1](https://github.com/AIGNE-io/aigne-framework/compare/core-v1.61.0-beta...core-v1.61.0-beta.1) (2025-09-23)
+
+
+### Bug Fixes
+
+* standardize file parameter naming across models ([#534](https://github.com/AIGNE-io/aigne-framework/issues/534)) ([f159a9d](https://github.com/AIGNE-io/aigne-framework/commit/f159a9d6af21ec0e99641996b150560929845845))
+
 ## [1.61.0-beta](https://github.com/AIGNE-io/aigne-framework/compare/core-v1.60.3...core-v1.61.0-beta) (2025-09-22)
 
 

package/lib/cjs/agents/ai-agent.d.ts

@@ -3,9 +3,9 @@ import { PromptBuilder } from "../prompt/prompt-builder.js";
 import { Agent, type AgentInvokeOptions, type AgentOptions, type AgentProcessAsyncGenerator, type Message } from "./agent.js";
 import type { ChatModel, ChatModelInput } from "./chat-model.js";
 import type { GuideRailAgentOutput } from "./guide-rail-agent.js";
-import type { FileOutputType } from "./model.js";
+import type { FileType } from "./model.js";
 export declare const DEFAULT_OUTPUT_KEY = "message";
-export declare const DEFAULT_FILE_OUTPUT_KEY = "files";
+export declare const DEFAULT_OUTPUT_FILE_KEY = "files";
 /**
  * Configuration options for an AI Agent
  *
@@ -27,15 +27,15 @@ export interface AIAgentOptions<I extends Message = Message, O extends Message =
      * Pick a message from input to use as the user's message
      */
     inputKey?: string;
-    fileInputKey?: string;
+    inputFileKey?: string;
     /**
      * Custom key to use for text output in the response
      *
      * Defaults to `message` if not specified
      */
     outputKey?: string;
-    fileOutputKey?: string;
-    fileOutputType?: FileOutputType;
+    outputFileKey?: string;
+    outputFileType?: FileType;
     /**
      * Controls how the agent uses tools during execution
      *
@@ -210,7 +210,7 @@ export declare class AIAgent<I extends Message = any, O extends Message = any> e
      * Pick a message from input to use as the user's message
      */
     inputKey?: string;
-    fileInputKey?: string;
+    inputFileKey?: string;
     /**
      * Custom key to use for text output in the response
      *
@@ -219,8 +219,8 @@ export declare class AIAgent<I extends Message = any, O extends Message = any> e
      * {@includeCode ../../test/agents/ai-agent.test.ts#example-ai-agent-custom-output-key}
      */
     outputKey: string;
-    fileOutputKey: string;
-    fileOutputType?: FileOutputType;
+    outputFileKey: string;
+    outputFileType?: FileType;
     /**
      * Controls how the agent uses tools during execution
      *

package/lib/cjs/agents/ai-agent.js

@@ -1,6 +1,6 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.AIAgent = exports.aiAgentOptionsSchema = exports.aiAgentToolChoiceSchema = exports.AIAgentToolChoice = exports.DEFAULT_FILE_OUTPUT_KEY = exports.DEFAULT_OUTPUT_KEY = void 0;
+exports.AIAgent = exports.aiAgentOptionsSchema = exports.aiAgentToolChoiceSchema = exports.AIAgentToolChoice = exports.DEFAULT_OUTPUT_FILE_KEY = exports.DEFAULT_OUTPUT_KEY = void 0;
 const zod_1 = require("zod");
 const prompt_builder_js_1 = require("../prompt/prompt-builder.js");
 const structured_stream_instructions_js_1 = require("../prompt/prompts/structured-stream-instructions.js");
@@ -10,7 +10,7 @@ const type_utils_js_1 = require("../utils/type-utils.js");
 const agent_js_1 = require("./agent.js");
 const types_js_1 = require("./types.js");
 exports.DEFAULT_OUTPUT_KEY = "message";
-exports.DEFAULT_FILE_OUTPUT_KEY = "files";
+exports.DEFAULT_OUTPUT_FILE_KEY = "files";
 /**
  * Tool choice options for AI agents
  *
@@ -111,10 +111,10 @@ class AIAgent extends agent_js_1.Agent {
                 ? prompt_builder_js_1.PromptBuilder.from(options.instructions)
                 : (options.instructions ?? new prompt_builder_js_1.PromptBuilder());
         this.inputKey = options.inputKey;
-        this.fileInputKey = options.fileInputKey;
+        this.inputFileKey = options.inputFileKey;
         this.outputKey = options.outputKey || exports.DEFAULT_OUTPUT_KEY;
-        this.fileOutputKey = options.fileOutputKey || exports.DEFAULT_FILE_OUTPUT_KEY;
-        this.fileOutputType = options.fileOutputType;
+        this.outputFileKey = options.outputFileKey || exports.DEFAULT_OUTPUT_FILE_KEY;
+        this.outputFileType = options.outputFileType;
         this.toolChoice = options.toolChoice;
         this.memoryAgentsAsTools = options.memoryAgentsAsTools;
         this.memoryPromptTemplate = options.memoryPromptTemplate;
@@ -148,7 +148,7 @@ class AIAgent extends agent_js_1.Agent {
      * Pick a message from input to use as the user's message
      */
     inputKey;
-    fileInputKey;
+    inputFileKey;
     /**
      * Custom key to use for text output in the response
      *
@@ -157,8 +157,8 @@ class AIAgent extends agent_js_1.Agent {
      * {@includeCode ../../test/agents/ai-agent.test.ts#example-ai-agent-custom-output-key}
      */
     outputKey;
-    fileOutputKey;
-    fileOutputType;
+    outputFileKey;
+    outputFileType;
     /**
      * Controls how the agent uses tools during execution
      *
@@ -309,7 +309,7 @@ class AIAgent extends agent_js_1.Agent {
                 Object.assign(result, { [outputKey]: text });
             }
             if (files) {
-                Object.assign(result, { [this.fileOutputKey]: files });
+                Object.assign(result, { [this.outputFileKey]: files });
             }
             if (!(0, type_utils_js_1.isEmpty)(result)) {
                 yield { delta: { json: result } };

package/lib/cjs/agents/chat-model.d.ts

@@ -1,12 +1,12 @@
 import { z } from "zod";
 import { type PromiseOrValue } from "../utils/type-utils.js";
 import { type AgentInvokeOptions, type AgentOptions, type AgentProcessResult, type AgentResponse, type AgentResponseStream, type Message } from "./agent.js";
-import { type FileOutputType, type FileUnionContent, Model } from "./model.js";
+import { type FileType, type FileUnionContent, Model } from "./model.js";
 export declare class StructuredOutputError extends Error {
 }
 export interface ChatModelOptions extends Omit<AgentOptions<ChatModelInput, ChatModelOutput>, "model" | "inputSchema" | "outputSchema"> {
     model?: string;
-    modelOptions?: Omit<ModelOptions, "model">;
+    modelOptions?: Omit<ChatModelInputOptions, "model">;
 }
 /**
  * ChatModel is an abstract base class for interacting with Large Language Models (LLMs).
@@ -135,7 +135,7 @@ export interface ChatModelInput extends Message {
      * Specifies the expected response format
      */
     responseFormat?: ChatModelInputResponseFormat;
-    fileOutputType?: FileOutputType;
+    outputFileType?: FileType;
     /**
      * List of tools available for the model to use
      */
@@ -147,7 +147,7 @@ export interface ChatModelInput extends Message {
     /**
      * Model-specific configuration options
      */
-    modelOptions?: ModelOptions;
+    modelOptions?: ChatModelInputOptions;
 }
 /**
  * Message role types
@@ -351,7 +351,7 @@ export type Modality = "text" | "image" | "audio";
  *
  * Contains various parameters for controlling model behavior, such as model name, temperature, etc.
  */
-export interface ModelOptions {
+export interface ChatModelInputOptions {
     /**
      * Model name or version
      */
@@ -377,7 +377,7 @@ export interface ModelOptions {
      */
     parallelToolCalls?: boolean;
     modalities?: Modality[];
-    preferFileInputType?: "file" | "url";
+    preferInputFileType?: "file" | "url";
 }
 /**
  * Output message format for ChatModel

package/lib/cjs/agents/chat-model.js

@@ -186,8 +186,8 @@ class ChatModel extends model_js_1.Model {
                             mimeType: item.mimeType || ChatModel.getMimeType(item.filename || item.path),
                         };
                     }
-                    if ((input.modelOptions?.preferFileInputType ||
-                        this.options?.modelOptions?.preferFileInputType) !== "url") {
+                    if ((input.modelOptions?.preferInputFileType ||
+                        this.options?.modelOptions?.preferInputFileType) !== "url") {
                         if (item.type === "url") {
                             return {
                                 ...item,
@@ -240,7 +240,7 @@ class ChatModel extends model_js_1.Model {
             const files = zod_1.z.array(model_js_1.fileUnionContentSchema).parse(output.files);
             output = {
                 ...output,
-                files: await Promise.all(files.map((file) => this.transformFileOutput(input.fileOutputType, file, options))),
+                files: await Promise.all(files.map((file) => this.transformFileType(input.outputFileType, file, options))),
             };
         }
         // Remove fields with `null` value for validation

package/lib/cjs/agents/image-agent.d.ts

@@ -2,11 +2,11 @@ import { type ZodObject, type ZodType } from "zod";
 import { PromptBuilder } from "../prompt/prompt-builder.js";
 import { Agent, type AgentInvokeOptions, type AgentOptions, type Message } from "./agent.js";
 import { type ImageModelOutput } from "./image-model.js";
-import type { FileOutputType } from "./model.js";
+import { type FileType } from "./model.js";
 export interface ImageAgentOptions<I extends Message = any, O extends ImageModelOutput = any> extends Omit<AgentOptions<I, O>, "outputSchema"> {
     instructions: string | PromptBuilder;
     modelOptions?: Record<string, any>;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
 }
 export declare const imageAgentOptionsSchema: ZodObject<{
     [key in keyof ImageAgentOptions]: ZodType<ImageAgentOptions[key]>;
@@ -17,6 +17,6 @@ export declare class ImageAgent<I extends Message = any, O extends ImageModelOut
     constructor(options: ImageAgentOptions<I, O>);
     instructions: PromptBuilder;
     modelOptions?: Record<string, any>;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
     process(input: I, options: AgentInvokeOptions): Promise<O>;
 }

package/lib/cjs/agents/image-agent.js

@@ -9,9 +9,11 @@ const prompt_builder_js_1 = require("../prompt/prompt-builder.js");
 const type_utils_js_1 = require("../utils/type-utils.js");
 const agent_js_1 = require("./agent.js");
 const image_model_js_1 = require("./image-model.js");
+const model_js_1 = require("./model.js");
 exports.imageAgentOptionsSchema = agent_js_1.agentOptionsSchema.extend({
     instructions: zod_1.default.union([zod_1.default.string(), zod_1.default.custom()]),
     modelOptions: zod_1.default.record(zod_1.default.any()).optional(),
+    outputFileType: model_js_1.fileTypeSchema.optional(),
 });
 class ImageAgent extends agent_js_1.Agent {
     tag = "ImageAgent";
@@ -26,17 +28,17 @@ class ImageAgent extends agent_js_1.Agent {
                 ? prompt_builder_js_1.PromptBuilder.from(options.instructions)
                 : options.instructions;
         this.modelOptions = options.modelOptions;
-        this.outputType = options.outputType;
+        this.outputFileType = options.outputFileType;
     }
     instructions;
     modelOptions;
-    outputType;
+    outputFileType;
     async process(input, options) {
         const imageModel = this.imageModel || options.imageModel || options.context.imageModel;
         if (!imageModel)
             throw new Error("image model is required to run ImageAgent");
         const { prompt } = await this.instructions.buildImagePrompt({ input });
-        return (await this.invokeChildAgent(imageModel, { ...input, ...this.modelOptions, prompt, outputType: this.outputType }, { ...options, streaming: false }));
+        return (await this.invokeChildAgent(imageModel, { ...input, modelOptions: this.modelOptions, prompt, outputFileType: this.outputFileType }, { ...options, streaming: false }));
     }
 }
 exports.ImageAgent = ImageAgent;

package/lib/cjs/agents/image-model.d.ts

@@ -2,12 +2,15 @@ import { type ZodType, z } from "zod";
 import type { PromiseOrValue } from "../utils/type-utils.js";
 import type { AgentInvokeOptions, AgentOptions, AgentProcessResult, AgentResponse, AgentResponseStream, Message } from "./agent.js";
 import { type ChatModelOutputUsage } from "./chat-model.js";
-import { FileOutputType, type FileUnionContent, Model } from "./model.js";
+import { type FileType, type FileUnionContent, Model } from "./model.js";
 export interface ImageModelOptions<I extends ImageModelInput = ImageModelInput, O extends ImageModelOutput = ImageModelOutput> extends Omit<AgentOptions<I, O>, "model"> {
+    model?: string;
+    modelOptions?: Omit<ImageModelInputOptions, "model">;
 }
 export declare abstract class ImageModel<I extends ImageModelInput = ImageModelInput, O extends ImageModelOutput = ImageModelOutput> extends Model<I, O> {
+    options?: ImageModelOptions<I, O> | undefined;
     tag: string;
-    constructor(options?: ImageModelOptions<I, O>);
+    constructor(options?: ImageModelOptions<I, O> | undefined);
     get credential(): PromiseOrValue<{
         url?: string;
         apiKey?: string;
@@ -20,14 +23,18 @@ export declare abstract class ImageModel<I extends ImageModelInput = ImageModelI
 }
 export type ImageModelInputImage = FileUnionContent[];
 export interface ImageModelInput extends Message {
-    model?: string;
-    image?: ImageModelInputImage;
     prompt: string;
+    image?: ImageModelInputImage;
     n?: number;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
+    modelOptions?: ImageModelInputOptions;
+}
+export interface ImageModelInputOptions extends Record<string, unknown> {
+    model?: string;
+    preferInputFileType?: "file" | "url";
 }
 export declare const imageModelInputSchema: z.ZodObject<{
-    model: z.ZodOptional<z.ZodString>;
+    prompt: z.ZodString;
     image: z.ZodOptional<z.ZodArray<z.ZodDiscriminatedUnion<"type", [z.ZodObject<{
         filename: ZodType<string | undefined, z.ZodTypeDef, string | undefined>;
         mimeType: ZodType<string | undefined, z.ZodTypeDef, string | undefined>;
@@ -77,12 +84,12 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     }>]>, "many">>;
-    prompt: z.ZodString;
     n: z.ZodOptional<z.ZodNumber>;
-    outputType: z.ZodOptional<z.ZodNativeEnum<typeof FileOutputType>>;
+    outputFileType: z.ZodOptional<z.ZodEnum<["local", "file"]>>;
+    modelOptions: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
 }, "strip", z.ZodTypeAny, {
     prompt: string;
-    model?: string | undefined;
+    modelOptions?: Record<string, unknown> | undefined;
     image?: ({
         type: "url";
         url: string;
@@ -99,11 +106,11 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     })[] | undefined;
+    outputFileType?: "local" | "file" | undefined;
     n?: number | undefined;
-    outputType?: FileOutputType | undefined;
 }, {
     prompt: string;
-    model?: string | undefined;
+    modelOptions?: Record<string, unknown> | undefined;
     image?: ({
         type: "url";
         url: string;
@@ -120,8 +127,8 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     })[] | undefined;
+    outputFileType?: "local" | "file" | undefined;
     n?: number | undefined;
-    outputType?: FileOutputType | undefined;
 }>;
 export interface ImageModelOutput extends Message {
     images: FileUnionContent[];

package/lib/cjs/agents/image-model.js

@@ -6,13 +6,16 @@ const zod_1 = require("zod");
 const chat_model_js_1 = require("./chat-model.js");
 const model_js_1 = require("./model.js");
 class ImageModel extends model_js_1.Model {
+    options;
     tag = "ImageModelAgent";
     constructor(options) {
         super({
             inputSchema: exports.imageModelInputSchema,
             outputSchema: exports.imageModelOutputSchema,
             ...options,
+            model: undefined,
         });
+        this.options = options;
     }
     get credential() {
         return {};
@@ -35,6 +38,18 @@ class ImageModel extends model_js_1.Model {
                         mimeType: item.mimeType || ImageModel.getMimeType(item.filename || item.path),
                     };
                 }
+                if ((input.modelOptions?.preferInputFileType ||
+                    this.options?.modelOptions?.preferInputFileType) !== "url") {
+                    if (item.type === "url") {
+                        return {
+                            ...item,
+                            type: "file",
+                            data: Buffer.from(await (await this.downloadFile(item.url)).arrayBuffer()).toString("base64"),
+                            url: undefined,
+                            mimeType: item.mimeType || ImageModel.getMimeType(item.filename || item.url),
+                        };
+                    }
+                }
                 return item;
             }));
         }
@@ -54,7 +69,7 @@ class ImageModel extends model_js_1.Model {
             const images = zod_1.z.array(model_js_1.fileUnionContentSchema).parse(output.images);
             output = {
                 ...output,
-                images: await Promise.all(images.map((image) => this.transformFileOutput(input.outputType, image, options))),
+                images: await Promise.all(images.map((image) => this.transformFileType(input.outputFileType, image, options))),
             };
         }
         return super.processAgentOutput(input, output, options);
@@ -62,11 +77,11 @@ class ImageModel extends model_js_1.Model {
 }
 exports.ImageModel = ImageModel;
 exports.imageModelInputSchema = zod_1.z.object({
-    model: zod_1.z.string().optional(),
-    image: zod_1.z.array(model_js_1.fileUnionContentSchema).optional().describe("Images used for editing"),
     prompt: zod_1.z.string(),
+    image: zod_1.z.array(model_js_1.fileUnionContentSchema).optional().describe("Images used for editing"),
     n: zod_1.z.number().int().min(1).optional(),
-    outputType: zod_1.z.nativeEnum(model_js_1.FileOutputType).optional(),
+    outputFileType: model_js_1.fileTypeSchema.optional(),
+    modelOptions: zod_1.z.record(zod_1.z.unknown()).optional(),
 });
 exports.imageModelOutputSchema = zod_1.z.object({
     images: zod_1.z.array(model_js_1.fileUnionContentSchema),

package/lib/cjs/agents/model.d.ts

@@ -1,17 +1,15 @@
 import { z } from "zod";
 import { Agent, type AgentInvokeOptions, type Message } from "./agent.js";
 export declare abstract class Model<I extends Message = any, O extends Message = any> extends Agent<I, O> {
-    transformFileOutput(fileOutputType: FileOutputType.file, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileContent>;
-    transformFileOutput(fileOutputType: FileOutputType.local | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<LocalContent>;
-    transformFileOutput(fileOutputType: FileOutputType | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileUnionContent>;
+    transformFileType(fileType: "file", data: FileUnionContent, options: AgentInvokeOptions): Promise<FileContent>;
+    transformFileType(fileType: "local" | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<LocalContent>;
+    transformFileType(fileType: FileType | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileUnionContent>;
     static getFileExtension(type: string): string | undefined;
     static getMimeType(filename: string): string | undefined;
     downloadFile(url: string): Promise<Response>;
 }
-export declare enum FileOutputType {
-    local = "local",
-    file = "file"
-}
+export type FileType = "local" | "file";
+export declare const fileTypeSchema: z.ZodEnum<["local", "file"]>;
 export interface FileContentBase {
     filename?: string;
     mimeType?: string;

package/lib/cjs/agents/model.js

@@ -3,7 +3,7 @@ var __importDefault = (this && this.__importDefault) || function (mod) {
     return (mod && mod.__esModule) ? mod : { "default": mod };
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.fileUnionContentsSchema = exports.fileUnionContentSchema = exports.localContentSchema = exports.fileContentSchema = exports.urlContentSchema = exports.fileContentBaseSchema = exports.FileOutputType = exports.Model = void 0;
+exports.fileUnionContentsSchema = exports.fileUnionContentSchema = exports.localContentSchema = exports.fileContentSchema = exports.urlContentSchema = exports.fileContentBaseSchema = exports.fileTypeSchema = exports.Model = void 0;
 const index_js_1 = require("@aigne/platform-helpers/nodejs/index.js");
 const mime_1 = __importDefault(require("mime"));
 const ufo_1 = require("ufo");
@@ -13,12 +13,12 @@ const schema_js_1 = require("../loader/schema.js");
 const type_utils_js_1 = require("../utils/type-utils.js");
 const agent_js_1 = require("./agent.js");
 class Model extends agent_js_1.Agent {
-    async transformFileOutput(fileOutputType = FileOutputType.local, data, options) {
-        if (fileOutputType === data.type)
+    async transformFileType(fileType = "local", data, options) {
+        if (fileType === data.type)
             return data;
         const common = (0, type_utils_js_1.pick)(data, "filename", "mimeType");
-        switch (fileOutputType) {
-            case FileOutputType.local: {
+        switch (fileType) {
+            case "local": {
                 const dir = index_js_1.nodejs.path.join(index_js_1.nodejs.os.tmpdir(), options.context.id);
                 await index_js_1.nodejs.fs.mkdir(dir, { recursive: true });
                 const ext = Model.getFileExtension(data.mimeType || data.filename || "");
@@ -41,7 +41,7 @@ class Model extends agent_js_1.Agent {
                 }
                 return { ...common, type: "local", path, mimeType };
             }
-            case FileOutputType.file: {
+            case "file": {
                 let base64;
                 let mimeType = data.mimeType;
                 if (data.type === "local") {
@@ -75,11 +75,7 @@ class Model extends agent_js_1.Agent {
     }
 }
 exports.Model = Model;
-var FileOutputType;
-(function (FileOutputType) {
-    FileOutputType["local"] = "local";
-    FileOutputType["file"] = "file";
-})(FileOutputType || (exports.FileOutputType = FileOutputType = {}));
+exports.fileTypeSchema = zod_1.z.enum(["local", "file"]);
 exports.fileContentBaseSchema = zod_1.z.object({
     filename: (0, schema_js_1.optionalize)(zod_1.z.string()),
     mimeType: (0, schema_js_1.optionalize)(zod_1.z.string()),

package/lib/cjs/prompt/prompt-builder.js

@@ -68,7 +68,7 @@ class PromptBuilder {
             responseFormat: options.agent?.structuredStreamMode
                 ? undefined
                 : this.buildResponseFormat(options),
-            fileOutputType: options.agent?.fileOutputType,
+            outputFileType: options.agent?.outputFileType,
             ...this.buildTools(options),
         };
     }
@@ -87,9 +87,9 @@ class PromptBuilder {
         const messages = (await (typeof this.instructions === "string"
             ? template_js_1.ChatMessagesTemplate.from([template_js_1.SystemMessageTemplate.from(this.instructions)])
             : this.instructions)?.format(options.input, { workingDir: this.workingDir })) ?? [];
-        const fileInputKey = options.agent?.fileInputKey;
-        const files = (0, type_utils_js_1.flat)(fileInputKey
-            ? (0, type_utils_js_1.checkArguments)("Check input files", (0, schema_js_1.optionalize)(model_js_1.fileUnionContentsSchema), input?.[fileInputKey])
+        const inputFileKey = options.agent?.inputFileKey;
+        const files = (0, type_utils_js_1.flat)(inputFileKey
+            ? (0, type_utils_js_1.checkArguments)("Check input files", (0, schema_js_1.optionalize)(model_js_1.fileUnionContentsSchema), input?.[inputFileKey])
             : null);
         const memories = [];
         if (options.agent && options.context) {
@@ -126,9 +126,9 @@ class PromptBuilder {
         const messages = [];
         const other = [];
         const inputKey = options.agent?.inputKey;
-        const fileInputKey = options.agent?.fileInputKey;
+        const inputFileKey = options.agent?.inputFileKey;
         const outputKey = options.agent?.outputKey || ai_agent_js_1.DEFAULT_OUTPUT_KEY;
-        const fileOutputKey = options.agent?.fileOutputKey || ai_agent_js_1.DEFAULT_FILE_OUTPUT_KEY;
+        const outputFileKey = options.agent?.outputFileKey || ai_agent_js_1.DEFAULT_OUTPUT_FILE_KEY;
         const stringOrStringify = (value) => typeof value === "string" ? value : (0, yaml_1.stringify)(value);
         const convertMemoryToMessage = async (content) => {
             const { input, output } = content;
@@ -138,7 +138,7 @@ class PromptBuilder {
             const userMessageContent = [];
             if (typeof input === "object") {
                 const inputMessage = inputKey ? Reflect.get(input, inputKey) : undefined;
-                const inputFiles = fileInputKey ? Reflect.get(input, fileInputKey) : undefined;
+                const inputFiles = inputFileKey ? Reflect.get(input, inputFileKey) : undefined;
                 if (inputMessage) {
                     userMessageContent.push({ type: "text", text: stringOrStringify(inputMessage) });
                 }
@@ -153,7 +153,7 @@ class PromptBuilder {
             const agentMessageContent = [];
             if (typeof output === "object") {
                 const outputMessage = Reflect.get(output, outputKey);
-                const outputFiles = Reflect.get(output, fileOutputKey);
+                const outputFiles = Reflect.get(output, outputFileKey);
                 if (outputMessage) {
                     agentMessageContent.push({ type: "text", text: stringOrStringify(outputMessage) });
                 }

package/lib/dts/agents/ai-agent.d.ts

@@ -3,9 +3,9 @@ import { PromptBuilder } from "../prompt/prompt-builder.js";
 import { Agent, type AgentInvokeOptions, type AgentOptions, type AgentProcessAsyncGenerator, type Message } from "./agent.js";
 import type { ChatModel, ChatModelInput } from "./chat-model.js";
 import type { GuideRailAgentOutput } from "./guide-rail-agent.js";
-import type { FileOutputType } from "./model.js";
+import type { FileType } from "./model.js";
 export declare const DEFAULT_OUTPUT_KEY = "message";
-export declare const DEFAULT_FILE_OUTPUT_KEY = "files";
+export declare const DEFAULT_OUTPUT_FILE_KEY = "files";
 /**
  * Configuration options for an AI Agent
  *
@@ -27,15 +27,15 @@ export interface AIAgentOptions<I extends Message = Message, O extends Message =
      * Pick a message from input to use as the user's message
      */
     inputKey?: string;
-    fileInputKey?: string;
+    inputFileKey?: string;
     /**
      * Custom key to use for text output in the response
      *
      * Defaults to `message` if not specified
      */
     outputKey?: string;
-    fileOutputKey?: string;
-    fileOutputType?: FileOutputType;
+    outputFileKey?: string;
+    outputFileType?: FileType;
     /**
      * Controls how the agent uses tools during execution
      *
@@ -210,7 +210,7 @@ export declare class AIAgent<I extends Message = any, O extends Message = any> e
      * Pick a message from input to use as the user's message
      */
     inputKey?: string;
-    fileInputKey?: string;
+    inputFileKey?: string;
     /**
      * Custom key to use for text output in the response
      *
@@ -219,8 +219,8 @@ export declare class AIAgent<I extends Message = any, O extends Message = any> e
      * {@includeCode ../../test/agents/ai-agent.test.ts#example-ai-agent-custom-output-key}
      */
     outputKey: string;
-    fileOutputKey: string;
-    fileOutputType?: FileOutputType;
+    outputFileKey: string;
+    outputFileType?: FileType;
     /**
      * Controls how the agent uses tools during execution
      *

package/lib/dts/agents/chat-model.d.ts

@@ -1,12 +1,12 @@
 import { z } from "zod";
 import { type PromiseOrValue } from "../utils/type-utils.js";
 import { type AgentInvokeOptions, type AgentOptions, type AgentProcessResult, type AgentResponse, type AgentResponseStream, type Message } from "./agent.js";
-import { type FileOutputType, type FileUnionContent, Model } from "./model.js";
+import { type FileType, type FileUnionContent, Model } from "./model.js";
 export declare class StructuredOutputError extends Error {
 }
 export interface ChatModelOptions extends Omit<AgentOptions<ChatModelInput, ChatModelOutput>, "model" | "inputSchema" | "outputSchema"> {
     model?: string;
-    modelOptions?: Omit<ModelOptions, "model">;
+    modelOptions?: Omit<ChatModelInputOptions, "model">;
 }
 /**
  * ChatModel is an abstract base class for interacting with Large Language Models (LLMs).
@@ -135,7 +135,7 @@ export interface ChatModelInput extends Message {
      * Specifies the expected response format
      */
     responseFormat?: ChatModelInputResponseFormat;
-    fileOutputType?: FileOutputType;
+    outputFileType?: FileType;
     /**
      * List of tools available for the model to use
      */
@@ -147,7 +147,7 @@ export interface ChatModelInput extends Message {
     /**
      * Model-specific configuration options
      */
-    modelOptions?: ModelOptions;
+    modelOptions?: ChatModelInputOptions;
 }
 /**
  * Message role types
@@ -351,7 +351,7 @@ export type Modality = "text" | "image" | "audio";
  *
  * Contains various parameters for controlling model behavior, such as model name, temperature, etc.
  */
-export interface ModelOptions {
+export interface ChatModelInputOptions {
     /**
      * Model name or version
      */
@@ -377,7 +377,7 @@ export interface ModelOptions {
      */
     parallelToolCalls?: boolean;
     modalities?: Modality[];
-    preferFileInputType?: "file" | "url";
+    preferInputFileType?: "file" | "url";
 }
 /**
  * Output message format for ChatModel

package/lib/dts/agents/image-agent.d.ts

@@ -2,11 +2,11 @@ import { type ZodObject, type ZodType } from "zod";
 import { PromptBuilder } from "../prompt/prompt-builder.js";
 import { Agent, type AgentInvokeOptions, type AgentOptions, type Message } from "./agent.js";
 import { type ImageModelOutput } from "./image-model.js";
-import type { FileOutputType } from "./model.js";
+import { type FileType } from "./model.js";
 export interface ImageAgentOptions<I extends Message = any, O extends ImageModelOutput = any> extends Omit<AgentOptions<I, O>, "outputSchema"> {
     instructions: string | PromptBuilder;
     modelOptions?: Record<string, any>;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
 }
 export declare const imageAgentOptionsSchema: ZodObject<{
     [key in keyof ImageAgentOptions]: ZodType<ImageAgentOptions[key]>;
@@ -17,6 +17,6 @@ export declare class ImageAgent<I extends Message = any, O extends ImageModelOut
     constructor(options: ImageAgentOptions<I, O>);
     instructions: PromptBuilder;
     modelOptions?: Record<string, any>;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
     process(input: I, options: AgentInvokeOptions): Promise<O>;
 }

package/lib/dts/agents/image-model.d.ts

@@ -2,12 +2,15 @@ import { type ZodType, z } from "zod";
 import type { PromiseOrValue } from "../utils/type-utils.js";
 import type { AgentInvokeOptions, AgentOptions, AgentProcessResult, AgentResponse, AgentResponseStream, Message } from "./agent.js";
 import { type ChatModelOutputUsage } from "./chat-model.js";
-import { FileOutputType, type FileUnionContent, Model } from "./model.js";
+import { type FileType, type FileUnionContent, Model } from "./model.js";
 export interface ImageModelOptions<I extends ImageModelInput = ImageModelInput, O extends ImageModelOutput = ImageModelOutput> extends Omit<AgentOptions<I, O>, "model"> {
+    model?: string;
+    modelOptions?: Omit<ImageModelInputOptions, "model">;
 }
 export declare abstract class ImageModel<I extends ImageModelInput = ImageModelInput, O extends ImageModelOutput = ImageModelOutput> extends Model<I, O> {
+    options?: ImageModelOptions<I, O> | undefined;
     tag: string;
-    constructor(options?: ImageModelOptions<I, O>);
+    constructor(options?: ImageModelOptions<I, O> | undefined);
     get credential(): PromiseOrValue<{
         url?: string;
         apiKey?: string;
@@ -20,14 +23,18 @@ export declare abstract class ImageModel<I extends ImageModelInput = ImageModelI
 }
 export type ImageModelInputImage = FileUnionContent[];
 export interface ImageModelInput extends Message {
-    model?: string;
-    image?: ImageModelInputImage;
     prompt: string;
+    image?: ImageModelInputImage;
     n?: number;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
+    modelOptions?: ImageModelInputOptions;
+}
+export interface ImageModelInputOptions extends Record<string, unknown> {
+    model?: string;
+    preferInputFileType?: "file" | "url";
 }
 export declare const imageModelInputSchema: z.ZodObject<{
-    model: z.ZodOptional<z.ZodString>;
+    prompt: z.ZodString;
     image: z.ZodOptional<z.ZodArray<z.ZodDiscriminatedUnion<"type", [z.ZodObject<{
         filename: ZodType<string | undefined, z.ZodTypeDef, string | undefined>;
         mimeType: ZodType<string | undefined, z.ZodTypeDef, string | undefined>;
@@ -77,12 +84,12 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     }>]>, "many">>;
-    prompt: z.ZodString;
     n: z.ZodOptional<z.ZodNumber>;
-    outputType: z.ZodOptional<z.ZodNativeEnum<typeof FileOutputType>>;
+    outputFileType: z.ZodOptional<z.ZodEnum<["local", "file"]>>;
+    modelOptions: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
 }, "strip", z.ZodTypeAny, {
     prompt: string;
-    model?: string | undefined;
+    modelOptions?: Record<string, unknown> | undefined;
     image?: ({
         type: "url";
         url: string;
@@ -99,11 +106,11 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     })[] | undefined;
+    outputFileType?: "local" | "file" | undefined;
     n?: number | undefined;
-    outputType?: FileOutputType | undefined;
 }, {
     prompt: string;
-    model?: string | undefined;
+    modelOptions?: Record<string, unknown> | undefined;
     image?: ({
         type: "url";
         url: string;
@@ -120,8 +127,8 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     })[] | undefined;
+    outputFileType?: "local" | "file" | undefined;
     n?: number | undefined;
-    outputType?: FileOutputType | undefined;
 }>;
 export interface ImageModelOutput extends Message {
     images: FileUnionContent[];

package/lib/dts/agents/model.d.ts

@@ -1,17 +1,15 @@
 import { z } from "zod";
 import { Agent, type AgentInvokeOptions, type Message } from "./agent.js";
 export declare abstract class Model<I extends Message = any, O extends Message = any> extends Agent<I, O> {
-    transformFileOutput(fileOutputType: FileOutputType.file, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileContent>;
-    transformFileOutput(fileOutputType: FileOutputType.local | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<LocalContent>;
-    transformFileOutput(fileOutputType: FileOutputType | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileUnionContent>;
+    transformFileType(fileType: "file", data: FileUnionContent, options: AgentInvokeOptions): Promise<FileContent>;
+    transformFileType(fileType: "local" | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<LocalContent>;
+    transformFileType(fileType: FileType | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileUnionContent>;
     static getFileExtension(type: string): string | undefined;
     static getMimeType(filename: string): string | undefined;
     downloadFile(url: string): Promise<Response>;
 }
-export declare enum FileOutputType {
-    local = "local",
-    file = "file"
-}
+export type FileType = "local" | "file";
+export declare const fileTypeSchema: z.ZodEnum<["local", "file"]>;
 export interface FileContentBase {
     filename?: string;
     mimeType?: string;

package/lib/esm/agents/ai-agent.d.ts

@@ -3,9 +3,9 @@ import { PromptBuilder } from "../prompt/prompt-builder.js";
 import { Agent, type AgentInvokeOptions, type AgentOptions, type AgentProcessAsyncGenerator, type Message } from "./agent.js";
 import type { ChatModel, ChatModelInput } from "./chat-model.js";
 import type { GuideRailAgentOutput } from "./guide-rail-agent.js";
-import type { FileOutputType } from "./model.js";
+import type { FileType } from "./model.js";
 export declare const DEFAULT_OUTPUT_KEY = "message";
-export declare const DEFAULT_FILE_OUTPUT_KEY = "files";
+export declare const DEFAULT_OUTPUT_FILE_KEY = "files";
 /**
  * Configuration options for an AI Agent
  *
@@ -27,15 +27,15 @@ export interface AIAgentOptions<I extends Message = Message, O extends Message =
      * Pick a message from input to use as the user's message
      */
     inputKey?: string;
-    fileInputKey?: string;
+    inputFileKey?: string;
     /**
      * Custom key to use for text output in the response
      *
      * Defaults to `message` if not specified
      */
     outputKey?: string;
-    fileOutputKey?: string;
-    fileOutputType?: FileOutputType;
+    outputFileKey?: string;
+    outputFileType?: FileType;
     /**
      * Controls how the agent uses tools during execution
      *
@@ -210,7 +210,7 @@ export declare class AIAgent<I extends Message = any, O extends Message = any> e
      * Pick a message from input to use as the user's message
      */
     inputKey?: string;
-    fileInputKey?: string;
+    inputFileKey?: string;
     /**
      * Custom key to use for text output in the response
      *
@@ -219,8 +219,8 @@ export declare class AIAgent<I extends Message = any, O extends Message = any> e
      * {@includeCode ../../test/agents/ai-agent.test.ts#example-ai-agent-custom-output-key}
      */
     outputKey: string;
-    fileOutputKey: string;
-    fileOutputType?: FileOutputType;
+    outputFileKey: string;
+    outputFileType?: FileType;
     /**
      * Controls how the agent uses tools during execution
      *

package/lib/esm/agents/ai-agent.js

@@ -7,7 +7,7 @@ import { checkArguments, isEmpty } from "../utils/type-utils.js";
 import { Agent, agentOptionsSchema, isAgentResponseDelta, } from "./agent.js";
 import { isTransferAgentOutput } from "./types.js";
 export const DEFAULT_OUTPUT_KEY = "message";
-export const DEFAULT_FILE_OUTPUT_KEY = "files";
+export const DEFAULT_OUTPUT_FILE_KEY = "files";
 /**
  * Tool choice options for AI agents
  *
@@ -108,10 +108,10 @@ export class AIAgent extends Agent {
                 ? PromptBuilder.from(options.instructions)
                 : (options.instructions ?? new PromptBuilder());
         this.inputKey = options.inputKey;
-        this.fileInputKey = options.fileInputKey;
+        this.inputFileKey = options.inputFileKey;
         this.outputKey = options.outputKey || DEFAULT_OUTPUT_KEY;
-        this.fileOutputKey = options.fileOutputKey || DEFAULT_FILE_OUTPUT_KEY;
-        this.fileOutputType = options.fileOutputType;
+        this.outputFileKey = options.outputFileKey || DEFAULT_OUTPUT_FILE_KEY;
+        this.outputFileType = options.outputFileType;
         this.toolChoice = options.toolChoice;
         this.memoryAgentsAsTools = options.memoryAgentsAsTools;
         this.memoryPromptTemplate = options.memoryPromptTemplate;
@@ -145,7 +145,7 @@ export class AIAgent extends Agent {
      * Pick a message from input to use as the user's message
      */
     inputKey;
-    fileInputKey;
+    inputFileKey;
     /**
      * Custom key to use for text output in the response
      *
@@ -154,8 +154,8 @@ export class AIAgent extends Agent {
      * {@includeCode ../../test/agents/ai-agent.test.ts#example-ai-agent-custom-output-key}
      */
     outputKey;
-    fileOutputKey;
-    fileOutputType;
+    outputFileKey;
+    outputFileType;
     /**
      * Controls how the agent uses tools during execution
      *
@@ -306,7 +306,7 @@ export class AIAgent extends Agent {
                 Object.assign(result, { [outputKey]: text });
             }
             if (files) {
-                Object.assign(result, { [this.fileOutputKey]: files });
+                Object.assign(result, { [this.outputFileKey]: files });
             }
             if (!isEmpty(result)) {
                 yield { delta: { json: result } };

package/lib/esm/agents/chat-model.d.ts

@@ -1,12 +1,12 @@
 import { z } from "zod";
 import { type PromiseOrValue } from "../utils/type-utils.js";
 import { type AgentInvokeOptions, type AgentOptions, type AgentProcessResult, type AgentResponse, type AgentResponseStream, type Message } from "./agent.js";
-import { type FileOutputType, type FileUnionContent, Model } from "./model.js";
+import { type FileType, type FileUnionContent, Model } from "./model.js";
 export declare class StructuredOutputError extends Error {
 }
 export interface ChatModelOptions extends Omit<AgentOptions<ChatModelInput, ChatModelOutput>, "model" | "inputSchema" | "outputSchema"> {
     model?: string;
-    modelOptions?: Omit<ModelOptions, "model">;
+    modelOptions?: Omit<ChatModelInputOptions, "model">;
 }
 /**
  * ChatModel is an abstract base class for interacting with Large Language Models (LLMs).
@@ -135,7 +135,7 @@ export interface ChatModelInput extends Message {
      * Specifies the expected response format
      */
     responseFormat?: ChatModelInputResponseFormat;
-    fileOutputType?: FileOutputType;
+    outputFileType?: FileType;
     /**
      * List of tools available for the model to use
      */
@@ -147,7 +147,7 @@ export interface ChatModelInput extends Message {
     /**
      * Model-specific configuration options
      */
-    modelOptions?: ModelOptions;
+    modelOptions?: ChatModelInputOptions;
 }
 /**
  * Message role types
@@ -351,7 +351,7 @@ export type Modality = "text" | "image" | "audio";
  *
  * Contains various parameters for controlling model behavior, such as model name, temperature, etc.
  */
-export interface ModelOptions {
+export interface ChatModelInputOptions {
     /**
      * Model name or version
      */
@@ -377,7 +377,7 @@ export interface ModelOptions {
      */
     parallelToolCalls?: boolean;
     modalities?: Modality[];
-    preferFileInputType?: "file" | "url";
+    preferInputFileType?: "file" | "url";
 }
 /**
  * Output message format for ChatModel

package/lib/esm/agents/chat-model.js

@@ -149,8 +149,8 @@ export class ChatModel extends Model {
                             mimeType: item.mimeType || ChatModel.getMimeType(item.filename || item.path),
                         };
                     }
-                    if ((input.modelOptions?.preferFileInputType ||
-                        this.options?.modelOptions?.preferFileInputType) !== "url") {
+                    if ((input.modelOptions?.preferInputFileType ||
+                        this.options?.modelOptions?.preferInputFileType) !== "url") {
                         if (item.type === "url") {
                             return {
                                 ...item,
@@ -203,7 +203,7 @@ export class ChatModel extends Model {
             const files = z.array(fileUnionContentSchema).parse(output.files);
             output = {
                 ...output,
-                files: await Promise.all(files.map((file) => this.transformFileOutput(input.fileOutputType, file, options))),
+                files: await Promise.all(files.map((file) => this.transformFileType(input.outputFileType, file, options))),
             };
         }
         // Remove fields with `null` value for validation

package/lib/esm/agents/image-agent.d.ts

@@ -2,11 +2,11 @@ import { type ZodObject, type ZodType } from "zod";
 import { PromptBuilder } from "../prompt/prompt-builder.js";
 import { Agent, type AgentInvokeOptions, type AgentOptions, type Message } from "./agent.js";
 import { type ImageModelOutput } from "./image-model.js";
-import type { FileOutputType } from "./model.js";
+import { type FileType } from "./model.js";
 export interface ImageAgentOptions<I extends Message = any, O extends ImageModelOutput = any> extends Omit<AgentOptions<I, O>, "outputSchema"> {
     instructions: string | PromptBuilder;
     modelOptions?: Record<string, any>;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
 }
 export declare const imageAgentOptionsSchema: ZodObject<{
     [key in keyof ImageAgentOptions]: ZodType<ImageAgentOptions[key]>;
@@ -17,6 +17,6 @@ export declare class ImageAgent<I extends Message = any, O extends ImageModelOut
     constructor(options: ImageAgentOptions<I, O>);
     instructions: PromptBuilder;
     modelOptions?: Record<string, any>;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
     process(input: I, options: AgentInvokeOptions): Promise<O>;
 }

package/lib/esm/agents/image-agent.js

@@ -3,9 +3,11 @@ import { PromptBuilder } from "../prompt/prompt-builder.js";
 import { checkArguments } from "../utils/type-utils.js";
 import { Agent, agentOptionsSchema, } from "./agent.js";
 import { imageModelOutputSchema } from "./image-model.js";
+import { fileTypeSchema } from "./model.js";
 export const imageAgentOptionsSchema = agentOptionsSchema.extend({
     instructions: z.union([z.string(), z.custom()]),
     modelOptions: z.record(z.any()).optional(),
+    outputFileType: fileTypeSchema.optional(),
 });
 export class ImageAgent extends Agent {
     tag = "ImageAgent";
@@ -20,16 +22,16 @@ export class ImageAgent extends Agent {
                 ? PromptBuilder.from(options.instructions)
                 : options.instructions;
         this.modelOptions = options.modelOptions;
-        this.outputType = options.outputType;
+        this.outputFileType = options.outputFileType;
     }
     instructions;
     modelOptions;
-    outputType;
+    outputFileType;
     async process(input, options) {
         const imageModel = this.imageModel || options.imageModel || options.context.imageModel;
         if (!imageModel)
             throw new Error("image model is required to run ImageAgent");
         const { prompt } = await this.instructions.buildImagePrompt({ input });
-        return (await this.invokeChildAgent(imageModel, { ...input, ...this.modelOptions, prompt, outputType: this.outputType }, { ...options, streaming: false }));
+        return (await this.invokeChildAgent(imageModel, { ...input, modelOptions: this.modelOptions, prompt, outputFileType: this.outputFileType }, { ...options, streaming: false }));
     }
 }

package/lib/esm/agents/image-model.d.ts

@@ -2,12 +2,15 @@ import { type ZodType, z } from "zod";
 import type { PromiseOrValue } from "../utils/type-utils.js";
 import type { AgentInvokeOptions, AgentOptions, AgentProcessResult, AgentResponse, AgentResponseStream, Message } from "./agent.js";
 import { type ChatModelOutputUsage } from "./chat-model.js";
-import { FileOutputType, type FileUnionContent, Model } from "./model.js";
+import { type FileType, type FileUnionContent, Model } from "./model.js";
 export interface ImageModelOptions<I extends ImageModelInput = ImageModelInput, O extends ImageModelOutput = ImageModelOutput> extends Omit<AgentOptions<I, O>, "model"> {
+    model?: string;
+    modelOptions?: Omit<ImageModelInputOptions, "model">;
 }
 export declare abstract class ImageModel<I extends ImageModelInput = ImageModelInput, O extends ImageModelOutput = ImageModelOutput> extends Model<I, O> {
+    options?: ImageModelOptions<I, O> | undefined;
     tag: string;
-    constructor(options?: ImageModelOptions<I, O>);
+    constructor(options?: ImageModelOptions<I, O> | undefined);
     get credential(): PromiseOrValue<{
         url?: string;
         apiKey?: string;
@@ -20,14 +23,18 @@ export declare abstract class ImageModel<I extends ImageModelInput = ImageModelI
 }
 export type ImageModelInputImage = FileUnionContent[];
 export interface ImageModelInput extends Message {
-    model?: string;
-    image?: ImageModelInputImage;
     prompt: string;
+    image?: ImageModelInputImage;
     n?: number;
-    outputType?: FileOutputType;
+    outputFileType?: FileType;
+    modelOptions?: ImageModelInputOptions;
+}
+export interface ImageModelInputOptions extends Record<string, unknown> {
+    model?: string;
+    preferInputFileType?: "file" | "url";
 }
 export declare const imageModelInputSchema: z.ZodObject<{
-    model: z.ZodOptional<z.ZodString>;
+    prompt: z.ZodString;
     image: z.ZodOptional<z.ZodArray<z.ZodDiscriminatedUnion<"type", [z.ZodObject<{
         filename: ZodType<string | undefined, z.ZodTypeDef, string | undefined>;
         mimeType: ZodType<string | undefined, z.ZodTypeDef, string | undefined>;
@@ -77,12 +84,12 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     }>]>, "many">>;
-    prompt: z.ZodString;
     n: z.ZodOptional<z.ZodNumber>;
-    outputType: z.ZodOptional<z.ZodNativeEnum<typeof FileOutputType>>;
+    outputFileType: z.ZodOptional<z.ZodEnum<["local", "file"]>>;
+    modelOptions: z.ZodOptional<z.ZodRecord<z.ZodString, z.ZodUnknown>>;
 }, "strip", z.ZodTypeAny, {
     prompt: string;
-    model?: string | undefined;
+    modelOptions?: Record<string, unknown> | undefined;
     image?: ({
         type: "url";
         url: string;
@@ -99,11 +106,11 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     })[] | undefined;
+    outputFileType?: "local" | "file" | undefined;
     n?: number | undefined;
-    outputType?: FileOutputType | undefined;
 }, {
     prompt: string;
-    model?: string | undefined;
+    modelOptions?: Record<string, unknown> | undefined;
     image?: ({
         type: "url";
         url: string;
@@ -120,8 +127,8 @@ export declare const imageModelInputSchema: z.ZodObject<{
         filename?: string | undefined;
         mimeType?: string | undefined;
     })[] | undefined;
+    outputFileType?: "local" | "file" | undefined;
     n?: number | undefined;
-    outputType?: FileOutputType | undefined;
 }>;
 export interface ImageModelOutput extends Message {
     images: FileUnionContent[];

package/lib/esm/agents/image-model.js

@@ -1,15 +1,18 @@
 import { nodejs } from "@aigne/platform-helpers/nodejs/index.js";
 import { z } from "zod";
 import { chatModelOutputUsageSchema } from "./chat-model.js";
-import { FileOutputType, fileUnionContentSchema, Model } from "./model.js";
+import { fileTypeSchema, fileUnionContentSchema, Model, } from "./model.js";
 export class ImageModel extends Model {
+    options;
     tag = "ImageModelAgent";
     constructor(options) {
         super({
             inputSchema: imageModelInputSchema,
             outputSchema: imageModelOutputSchema,
             ...options,
+            model: undefined,
         });
+        this.options = options;
     }
     get credential() {
         return {};
@@ -32,6 +35,18 @@ export class ImageModel extends Model {
                         mimeType: item.mimeType || ImageModel.getMimeType(item.filename || item.path),
                     };
                 }
+                if ((input.modelOptions?.preferInputFileType ||
+                    this.options?.modelOptions?.preferInputFileType) !== "url") {
+                    if (item.type === "url") {
+                        return {
+                            ...item,
+                            type: "file",
+                            data: Buffer.from(await (await this.downloadFile(item.url)).arrayBuffer()).toString("base64"),
+                            url: undefined,
+                            mimeType: item.mimeType || ImageModel.getMimeType(item.filename || item.url),
+                        };
+                    }
+                }
                 return item;
             }));
         }
@@ -51,18 +66,18 @@ export class ImageModel extends Model {
             const images = z.array(fileUnionContentSchema).parse(output.images);
             output = {
                 ...output,
-                images: await Promise.all(images.map((image) => this.transformFileOutput(input.outputType, image, options))),
+                images: await Promise.all(images.map((image) => this.transformFileType(input.outputFileType, image, options))),
             };
         }
         return super.processAgentOutput(input, output, options);
     }
 }
 export const imageModelInputSchema = z.object({
-    model: z.string().optional(),
-    image: z.array(fileUnionContentSchema).optional().describe("Images used for editing"),
     prompt: z.string(),
+    image: z.array(fileUnionContentSchema).optional().describe("Images used for editing"),
     n: z.number().int().min(1).optional(),
-    outputType: z.nativeEnum(FileOutputType).optional(),
+    outputFileType: fileTypeSchema.optional(),
+    modelOptions: z.record(z.unknown()).optional(),
 });
 export const imageModelOutputSchema = z.object({
     images: z.array(fileUnionContentSchema),

package/lib/esm/agents/model.d.ts

@@ -1,17 +1,15 @@
 import { z } from "zod";
 import { Agent, type AgentInvokeOptions, type Message } from "./agent.js";
 export declare abstract class Model<I extends Message = any, O extends Message = any> extends Agent<I, O> {
-    transformFileOutput(fileOutputType: FileOutputType.file, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileContent>;
-    transformFileOutput(fileOutputType: FileOutputType.local | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<LocalContent>;
-    transformFileOutput(fileOutputType: FileOutputType | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileUnionContent>;
+    transformFileType(fileType: "file", data: FileUnionContent, options: AgentInvokeOptions): Promise<FileContent>;
+    transformFileType(fileType: "local" | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<LocalContent>;
+    transformFileType(fileType: FileType | undefined, data: FileUnionContent, options: AgentInvokeOptions): Promise<FileUnionContent>;
     static getFileExtension(type: string): string | undefined;
     static getMimeType(filename: string): string | undefined;
     downloadFile(url: string): Promise<Response>;
 }
-export declare enum FileOutputType {
-    local = "local",
-    file = "file"
-}
+export type FileType = "local" | "file";
+export declare const fileTypeSchema: z.ZodEnum<["local", "file"]>;
 export interface FileContentBase {
     filename?: string;
     mimeType?: string;

package/lib/esm/agents/model.js

@@ -7,12 +7,12 @@ import { optionalize } from "../loader/schema.js";
 import { pick } from "../utils/type-utils.js";
 import { Agent } from "./agent.js";
 export class Model extends Agent {
-    async transformFileOutput(fileOutputType = FileOutputType.local, data, options) {
-        if (fileOutputType === data.type)
+    async transformFileType(fileType = "local", data, options) {
+        if (fileType === data.type)
             return data;
         const common = pick(data, "filename", "mimeType");
-        switch (fileOutputType) {
-            case FileOutputType.local: {
+        switch (fileType) {
+            case "local": {
                 const dir = nodejs.path.join(nodejs.os.tmpdir(), options.context.id);
                 await nodejs.fs.mkdir(dir, { recursive: true });
                 const ext = Model.getFileExtension(data.mimeType || data.filename || "");
@@ -35,7 +35,7 @@ export class Model extends Agent {
                 }
                 return { ...common, type: "local", path, mimeType };
             }
-            case FileOutputType.file: {
+            case "file": {
                 let base64;
                 let mimeType = data.mimeType;
                 if (data.type === "local") {
@@ -68,11 +68,7 @@ export class Model extends Agent {
         return response;
     }
 }
-export var FileOutputType;
-(function (FileOutputType) {
-    FileOutputType["local"] = "local";
-    FileOutputType["file"] = "file";
-})(FileOutputType || (FileOutputType = {}));
+export const fileTypeSchema = z.enum(["local", "file"]);
 export const fileContentBaseSchema = z.object({
     filename: optionalize(z.string()),
     mimeType: optionalize(z.string()),

package/lib/esm/prompt/prompt-builder.js

@@ -3,7 +3,7 @@ import { stringify } from "yaml";
 import { ZodObject } from "zod";
 import { zodToJsonSchema } from "zod-to-json-schema";
 import { Agent } from "../agents/agent.js";
-import { DEFAULT_FILE_OUTPUT_KEY, DEFAULT_OUTPUT_KEY } from "../agents/ai-agent.js";
+import { DEFAULT_OUTPUT_FILE_KEY, DEFAULT_OUTPUT_KEY } from "../agents/ai-agent.js";
 import { fileUnionContentsSchema } from "../agents/model.js";
 import { optionalize } from "../loader/schema.js";
 import { outputSchemaToResponseFormatSchema } from "../utils/json-schema.js";
@@ -65,7 +65,7 @@ export class PromptBuilder {
             responseFormat: options.agent?.structuredStreamMode
                 ? undefined
                 : this.buildResponseFormat(options),
-            fileOutputType: options.agent?.fileOutputType,
+            outputFileType: options.agent?.outputFileType,
             ...this.buildTools(options),
         };
     }
@@ -84,9 +84,9 @@ export class PromptBuilder {
         const messages = (await (typeof this.instructions === "string"
             ? ChatMessagesTemplate.from([SystemMessageTemplate.from(this.instructions)])
             : this.instructions)?.format(options.input, { workingDir: this.workingDir })) ?? [];
-        const fileInputKey = options.agent?.fileInputKey;
-        const files = flat(fileInputKey
-            ? checkArguments("Check input files", optionalize(fileUnionContentsSchema), input?.[fileInputKey])
+        const inputFileKey = options.agent?.inputFileKey;
+        const files = flat(inputFileKey
+            ? checkArguments("Check input files", optionalize(fileUnionContentsSchema), input?.[inputFileKey])
             : null);
         const memories = [];
         if (options.agent && options.context) {
@@ -123,9 +123,9 @@ export class PromptBuilder {
         const messages = [];
         const other = [];
         const inputKey = options.agent?.inputKey;
-        const fileInputKey = options.agent?.fileInputKey;
+        const inputFileKey = options.agent?.inputFileKey;
         const outputKey = options.agent?.outputKey || DEFAULT_OUTPUT_KEY;
-        const fileOutputKey = options.agent?.fileOutputKey || DEFAULT_FILE_OUTPUT_KEY;
+        const outputFileKey = options.agent?.outputFileKey || DEFAULT_OUTPUT_FILE_KEY;
         const stringOrStringify = (value) => typeof value === "string" ? value : stringify(value);
         const convertMemoryToMessage = async (content) => {
             const { input, output } = content;
@@ -135,7 +135,7 @@ export class PromptBuilder {
             const userMessageContent = [];
             if (typeof input === "object") {
                 const inputMessage = inputKey ? Reflect.get(input, inputKey) : undefined;
-                const inputFiles = fileInputKey ? Reflect.get(input, fileInputKey) : undefined;
+                const inputFiles = inputFileKey ? Reflect.get(input, inputFileKey) : undefined;
                 if (inputMessage) {
                     userMessageContent.push({ type: "text", text: stringOrStringify(inputMessage) });
                 }
@@ -150,7 +150,7 @@ export class PromptBuilder {
             const agentMessageContent = [];
             if (typeof output === "object") {
                 const outputMessage = Reflect.get(output, outputKey);
-                const outputFiles = Reflect.get(output, fileOutputKey);
+                const outputFiles = Reflect.get(output, outputFileKey);
                 if (outputMessage) {
                     agentMessageContent.push({ type: "text", text: stringOrStringify(outputMessage) });
                 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@aigne/core",
-  "version": "1.61.0-beta",
+  "version": "1.61.0-beta.2",
   "description": "The functional core of agentic AI",
   "publishConfig": {
     "access": "public"