Cb/standard outputs (#8587)

Author: hntrl

langchain-core/src/_standard/_scratchpad.ts

@@ -41,6 +41,19 @@ if (openaiMessage.type === "user") {
   }
 }
 
+// @ts-expect-error - demo
+const file: ContentBlock.Multimodal.File = {
+  type: "file",
+  fileId: "123",
+  // uncommenting this will cause a type error
+  // url: "asddas",
+  mimeType: "image/png",
+  metadata: {
+    width: 100,
+    height: 100,
+  },
+};
+
 function normalReturn(): BaseMessage {
   return {
     type: "user",
@@ -62,14 +75,14 @@ if (normalMessage.type === "user") {
 
 // Test with multiple tool call interfaces
 
-interface ToolCallA extends ContentBlock.Tools.ToolCallContentBlock {
+interface ToolCallA extends ContentBlock.Tools.ToolCall {
   name: "tool_a";
   args: {
     bar: string;
   };
 }
 
-interface ToolCallB extends ContentBlock.Tools.ToolCallContentBlock {
+interface ToolCallB extends ContentBlock.Tools.ToolCall {
   name: "tool_b";
   args: {
     foo: string;

langchain-core/src/_standard/content/base.ts

@@ -0,0 +1,3 @@
+export interface ContentBlock {
+  readonly type: string;
+}

langchain-core/src/_standard/content/index.ts

@@ -1,14 +1,11 @@
-import { ToolContentBlocks } from "./tools.js";
-import { MultimodalContentBlocks } from "./multimodal.js";
-
-export interface ContentBlock {
-  readonly type: string;
-}
+import { ContentBlock as BaseContentBlock } from "./base.js";
+import { Tools as ToolsContentBlock } from "./tools.js";
+import { Multimodal as MultimodalContentBlock } from "./multimodal.js";
 
 // eslint-disable-next-line @typescript-eslint/no-namespace
 export declare namespace ContentBlock {
   /** Content block for citation */
-  interface Citation extends ContentBlock {
+  interface Citation extends BaseContentBlock {
     /**
      * Type of the content block
      */
@@ -44,7 +41,7 @@ export declare namespace ContentBlock {
   }
 
   /** Content block for text */
-  interface TextContentBlock extends ContentBlock {
+  interface Text extends BaseContentBlock {
     /**
      * Type of the content block
      */
@@ -66,7 +63,7 @@ export declare namespace ContentBlock {
   }
 
   /** Content block for reasoning output */
-  interface ReasoningContentBlock extends ContentBlock {
+  interface Reasoning extends BaseContentBlock {
     /**
      * Type of the content block
      */
@@ -104,12 +101,11 @@ export declare namespace ContentBlock {
     signature?: string;
   }
 
-  export import Tools = ToolContentBlocks;
-  export import Multimodal = MultimodalContentBlocks;
-
-  export type Standard =
-    | TextContentBlock
-    | ReasoningContentBlock
-    | Tools.Standard
-    | Multimodal.Standard;
+  export import Tools = ToolsContentBlock;
+  export import Multimodal = MultimodalContentBlock;
+  export type Types =
+    | Text
+    | Reasoning
+    | Tools.ContentBlock
+    | Multimodal.ContentBlock;
 }

langchain-core/src/_standard/content/multimodal.ts

@@ -1,138 +1,84 @@
-import { ContentBlock } from "./index.js";
+import { ContentBlock as BaseContentBlock } from "./base.js";
 
 // eslint-disable-next-line @typescript-eslint/no-namespace
-export declare namespace MultimodalContentBlocks {
-  /** Content block for multimodal data */
-  export interface DataContentBlock<
-    TMetadata extends Record<string, unknown> = Record<string, unknown>
-  > extends ContentBlock {
+export declare namespace Multimodal {
+  type DataRecordFileId = {
+    /**
+     * ID of the image file. e.g. from the OpenAI file api
+     */
+    fileId: string;
+    url?: never;
+    data?: never;
+  };
+  type DataRecordUrl = {
+    fileId?: never;
     /**
-     * MIME type of the file. Required for base64 encoding.
+     * URL of the data file
      */
-    mimeType?: string;
+    url: string;
+    data?: never;
+  };
+  type DataRecordData = {
+    fileId?: never;
+    url?: never;
     /**
-     * Metadata of the file
+     * Base64 encoded string or binary data of the data
      */
-    metadata?: TMetadata;
-  }
+    data: string | Uint8Array;
+  };
+  type DataRecord = DataRecordFileId | DataRecordUrl | DataRecordData;
+
+  /** Content block for multimodal data */
+  export type Data<
+    TMetadata extends Record<string, unknown> = Record<string, unknown>
+  > = BaseContentBlock &
+    DataRecord & {
+      /**
+       * MIME type of the file. Required for base64 encoding.
+       */
+      mimeType?: string;
+      /**
+       * Metadata of the file
+       */
+      metadata?: TMetadata;
+      /**
+       * Content block identifier for multimodal content, e.g. image, video, audio, file or plain text. This can be either:
+       *  - generated by the provider (e.g., an OpenAI block ID)
+       *  - generated by LangChain upon creation
+       */
+      id?: string;
+    };
 
   /** Content block for image data */
-  export interface ImageContentBlock extends DataContentBlock {
+  export type Image = Data & {
     /**
      * Type of the content block
      */
     readonly type: "image";
-    /**
-     * Image content block identifier, which can be either
-     *  - generated by the provider (e.g., an OpenAI image block ID)
-     *  - generated by LangChain upon creation
-     */
-    id?: string;
-    /**
-     * ID of the image file. e.g. from the OpenAI file api
-     */
-    fileId?: string;
-    /**
-     * MIME type of the image file. Required for base64 encoding.
-     */
-    mimeType?: string;
-    /**
-     * URL of the image file
-     */
-    url?: string;
-    /**
-     * Base64 encoded data of the image file
-     */
-    base64?: string;
-  }
+  };
   /** Content block for video data */
-  export interface VideoContentBlock extends DataContentBlock {
+  export type Video = Data & {
     /**
      * Type of the content block
      */
     readonly type: "video";
-    /**
-     * Video content block identifier, which can be either
-     *  - generated by the provider (e.g., an OpenAI video block ID)
-     *  - generated by LangChain upon creation
-     */
-    id?: string;
-    /**
-     * ID of the video file. e.g. from the OpenAI file api
-     */
-    fileId?: string;
-    /**
-     * MIME type of the video file. Required for base64 encoding.
-     */
-    mimeType?: string;
-    /**
-     * URL of the video file
-     */
-    url?: string;
-    /**
-     * Base64 encoded data of the video file
-     */
-    base64?: string;
-  }
+  };
 
   /** Content block for audio data */
-  export interface AudioContentBlock extends DataContentBlock {
+  export type Audio = Data & {
     /**
      * Type of the content block
      */
     readonly type: "audio";
-    /**
-     * Audio content block identifier, which can be either
-     *  - generated by the provider (e.g., an OpenAI audio block ID)
-     *  - generated by LangChain upon creation
-     */
-    id?: string;
-    /**
-     * ID of the audio file. e.g. from the OpenAI file api
-     */
-    fileId?: string;
-    /**
-     * MIME type of the audio file. Required for base64 encoding.
-     */
-    mimeType?: string;
-    /**
-     * URL of the audio file
-     */
-    url?: string;
-    /**
-     * Base64 encoded data of the audio file
-     */
-    base64?: string;
-  }
+  };
 
   /** Content block for plain text data */
-  export interface PlainTextContentBlock extends DataContentBlock {
+  export type PlainText = Data & {
     /**
      * Type of the content block
+     * maybe only "text"?
      */
     readonly type: "text-plain";
-    /**
-     * Plain text content block identifier, which can be either
-     *  - generated by the provider (e.g., an OpenAI file ID)
-     *  - generated by LangChain upon creation
-     */
-    id?: string;
-    /**
-     * ID of the file. e.g. from the OpenAI file api
-     */
-    fileId?: string;
-    /**
-     * MIME type of the file. Required for base64 encoding.
-     */
-    mimeType?: "text/plain";
-    /**
-     * URL of the plain text file
-     */
-    url?: string;
-    /**
-     * Base64 encoded data of the file
-     */
-    base64?: string;
     /**
      * Plaintext content. This is optional if the data is base64 encoded.
      */
@@ -145,42 +91,15 @@ export declare namespace MultimodalContentBlocks {
      * Context for the text, e.g. a description or a summary of the text's content
      */
     context?: string;
-  }
+  };
 
   /** Content block for file data */
-  export interface FileContentBlock extends DataContentBlock {
+  export type File = Data & {
     /**
      * Type of the content block
      */
     readonly type: "file";
-    /**
-     * Search call identifier, which can be either
-     *  - generated by the provider (e.g., an OpenAI search call ID)
-     *  - generated by LangChain upon creation
-     */
-    id?: string;
-    /**
-     * ID of the file. e.g. from the OpenAI file api
-     */
-    fileId?: string;
-    /**
-     * MIME type of the file. Required for base64 encoding.
-     */
-    mimeType?: string;
-    /**
-     * URL of the file
-     */
-    url?: string;
-    /**
-     * Base64 encoded data of the file
-     */
-    base64?: string;
-  }
+  };
 
-  export type Standard =
-    | ImageContentBlock
-    | VideoContentBlock
-    | AudioContentBlock
-    | PlainTextContentBlock
-    | FileContentBlock;
+  export type ContentBlock = Image | Video | Audio | PlainText | File;
 }