@librechat/agents 2.4.84 → 2.4.86

package/src/messages/format.ts

@@ -17,28 +17,28 @@ import type {
 } from '@/types';
 import { Providers, ContentTypes } from '@/common';
 
-interface VisionMessageParams {
+interface MediaMessageParams {
   message: {
     role: string;
     content: string;
     name?: string;
     [key: string]: any;
   };
-  image_urls: MessageContentImageUrl[];
+  mediaParts: MessageContentComplex[];
   endpoint?: Providers;
 }
 
 /**
- * Formats a message to OpenAI Vision API payload format.
+ * Formats a message with media content (images, documents, videos, audios) to API payload format.
  *
- * @param {VisionMessageParams} params - The parameters for formatting.
- * @returns {Object} - The formatted message.
+ * @param params - The parameters for formatting.
+ * @returns - The formatted message.
  */
-export const formatVisionMessage = ({
+export const formatMediaMessage = ({
   message,
-  image_urls,
   endpoint,
-}: VisionMessageParams): {
+  mediaParts,
+}: MediaMessageParams): {
   role: string;
   content: MessageContentComplex[];
   name?: string;
@@ -57,7 +57,7 @@ export const formatVisionMessage = ({
 
   if (endpoint === Providers.ANTHROPIC) {
     result.content = [
-      ...image_urls,
+      ...mediaParts,
       { type: ContentTypes.TEXT, text: message.content },
     ] as MessageContentComplex[];
     return result;
@@ -65,7 +65,7 @@ export const formatVisionMessage = ({
 
   result.content = [
     { type: ContentTypes.TEXT, text: message.content },
-    ...image_urls,
+    ...mediaParts,
   ] as MessageContentComplex[];
 
   return result;
@@ -78,6 +78,9 @@ interface MessageInput {
   text?: string;
   content?: string | MessageContentComplex[];
   image_urls?: MessageContentImageUrl[];
+  documents?: MessageContentComplex[];
+  videos?: MessageContentComplex[];
+  audios?: MessageContentComplex[];
   lc_id?: string[];
   [key: string]: any;
 }
@@ -100,14 +103,14 @@ interface FormattedMessage {
 /**
  * Formats a message to OpenAI payload format based on the provided options.
  *
- * @param {FormatMessageParams} params - The parameters for formatting.
- * @returns {FormattedMessage | HumanMessage | AIMessage | SystemMessage} - The formatted message.
+ * @param params - The parameters for formatting.
+ * @returns - The formatted message.
  */
 export const formatMessage = ({
   message,
   userName,
-  assistantName,
   endpoint,
+  assistantName,
   langChain = false,
 }: FormatMessageParams):
   | FormattedMessage
@@ -135,21 +138,7 @@ export const formatMessage = ({
     content,
   };
 
-  const { image_urls } = message;
-  if (Array.isArray(image_urls) && image_urls.length > 0 && role === 'user') {
-    return formatVisionMessage({
-      message: {
-        ...formattedMessage,
-        content:
-          typeof formattedMessage.content === 'string'
-            ? formattedMessage.content
-            : '',
-      },
-      image_urls,
-      endpoint,
-    });
-  }
-
+  // Set name fields first
   if (_name != null && _name) {
     formattedMessage.name = _name;
   }
@@ -179,6 +168,45 @@ export const formatMessage = ({
     }
   }
 
+  const { image_urls, documents, videos, audios } = message;
+  const mediaParts: MessageContentComplex[] = [];
+
+  if (Array.isArray(documents) && documents.length > 0) {
+    mediaParts.push(...documents);
+  }
+
+  if (Array.isArray(videos) && videos.length > 0) {
+    mediaParts.push(...videos);
+  }
+
+  if (Array.isArray(audios) && audios.length > 0) {
+    mediaParts.push(...audios);
+  }
+
+  if (Array.isArray(image_urls) && image_urls.length > 0) {
+    mediaParts.push(...image_urls);
+  }
+
+  if (mediaParts.length > 0 && role === 'user') {
+    const mediaMessage = formatMediaMessage({
+      message: {
+        ...formattedMessage,
+        content:
+          typeof formattedMessage.content === 'string'
+            ? formattedMessage.content
+            : '',
+      },
+      mediaParts,
+      endpoint,
+    });
+
+    if (!langChain) {
+      return mediaMessage;
+    }
+
+    return new HumanMessage(mediaMessage);
+  }
+
   if (!langChain) {
     return formattedMessage;
   }
@@ -195,9 +223,9 @@ export const formatMessage = ({
 /**
  * Formats an array of messages for LangChain.
  *
- * @param {Array<MessageInput>} messages - The array of messages to format.
- * @param {Omit<FormatMessageParams, 'message' | 'langChain'>} formatOptions - The options for formatting each message.
- * @returns {Array<HumanMessage | AIMessage | SystemMessage>} - The array of formatted LangChain messages.
+ * @param messages - The array of messages to format.
+ * @param formatOptions - The options for formatting each message.
+ * @returns - The array of formatted LangChain messages.
  */
 export const formatLangChainMessages = (
   messages: Array<MessageInput>,
@@ -228,8 +256,8 @@ interface LangChainMessage {
 /**
  * Formats a LangChain message object by merging properties from `lc_kwargs` or `kwargs` and `additional_kwargs`.
  *
- * @param {LangChainMessage} message - The message object to format.
- * @returns {Record<string, any>} The formatted LangChain message.
+ * @param message - The message object to format.
+ * @returns - The formatted LangChain message.
  */
 export const formatFromLangChain = (
   message: LangChainMessage
@@ -357,10 +385,10 @@ function formatAssistantMessage(
 /**
  * Formats an array of messages for LangChain, handling tool calls and creating ToolMessage instances.
  *
- * @param {TPayload} payload - The array of messages to format.
- * @param {Record<number, number>} [indexTokenCountMap] - Optional map of message indices to token counts.
- * @param {Set<string>} [tools] - Optional set of tool names that are allowed in the request.
- * @returns {Object} - Object containing formatted messages and updated indexTokenCountMap if provided.
+ * @param payload - The array of messages to format.
+ * @param indexTokenCountMap - Optional map of message indices to token counts.
+ * @param tools - Optional set of tool names that are allowed in the request.
+ * @returns - Object containing formatted messages and updated indexTokenCountMap if provided.
  */
 export const formatAgentMessages = (
   payload: TPayload,
@@ -539,8 +567,8 @@ export const formatAgentMessages = (
 
 /**
  * Formats an array of messages for LangChain, making sure all content fields are strings
- * @param {Array<HumanMessage | AIMessage | SystemMessage | ToolMessage>} payload - The array of messages to format.
- * @returns {Array<HumanMessage | AIMessage | SystemMessage | ToolMessage>} - The array of formatted LangChain messages, including ToolMessages for tool calls.
+ * @param payload - The array of messages to format.
+ * @returns - The array of formatted LangChain messages, including ToolMessages for tool calls.
  */
 export const formatContentStrings = (
   payload: Array<BaseMessage>

package/src/messages/formatMessage.test.ts

@@ -1,8 +1,43 @@
-import { HumanMessage, AIMessage, SystemMessage } from '@langchain/core/messages';
-import { formatMessage, formatLangChainMessages, formatFromLangChain } from './format';
+import {
+  HumanMessage,
+  AIMessage,
+  SystemMessage,
+} from '@langchain/core/messages';
+import type { MessageContentComplex } from '@/types';
+import {
+  formatMessage,
+  formatLangChainMessages,
+  formatFromLangChain,
+  formatMediaMessage,
+} from './format';
+import { Providers } from '@/common';
 
 const NO_PARENT = '00000000-0000-0000-0000-000000000000';
 
+/**
+ * Type for formatted message results with media content
+ */
+interface FormattedMediaMessage {
+  role: string;
+  content: MessageContentComplex[];
+  name?: string;
+}
+
+/**
+ * Type guard to check if result is a FormattedMediaMessage
+ */
+function isFormattedMediaMessage(
+  result: unknown
+): result is FormattedMediaMessage {
+  return (
+    typeof result === 'object' &&
+    result !== null &&
+    'role' in result &&
+    'content' in result &&
+    Array.isArray((result as FormattedMediaMessage).content)
+  );
+}
+
 describe('formatMessage', () => {
   it('formats user message', () => {
     const input = {
@@ -187,6 +222,387 @@ describe('formatMessage', () => {
   });
 });
 
+describe('formatMediaMessage', () => {
+  it('formats message with images for default provider', () => {
+    const message = {
+      role: 'user',
+      content: 'Check out this image',
+      name: 'John',
+    };
+    const mediaParts = [
+      {
+        type: 'image_url',
+        image_url: { url: 'https://example.com/image1.jpg' },
+      },
+      {
+        type: 'image_url',
+        image_url: { url: 'https://example.com/image2.jpg' },
+      },
+    ];
+
+    const result = formatMediaMessage({ message, mediaParts });
+
+    expect(result.role).toBe('user');
+    expect(result.name).toBe('John');
+    expect(Array.isArray(result.content)).toBe(true);
+    expect(result.content).toHaveLength(3);
+    expect(result.content[0]).toEqual({
+      type: 'text',
+      text: 'Check out this image',
+    });
+    expect(result.content[1]).toEqual(mediaParts[0]);
+    expect(result.content[2]).toEqual(mediaParts[1]);
+  });
+
+  it('formats message with images for Anthropic (media first)', () => {
+    const message = {
+      role: 'user',
+      content: 'Check out this image',
+    };
+    const mediaParts = [
+      {
+        type: 'image_url',
+        image_url: { url: 'https://example.com/image.jpg' },
+      },
+    ];
+
+    const result = formatMediaMessage({
+      message,
+      mediaParts,
+      endpoint: Providers.ANTHROPIC,
+    });
+
+    expect(result.content).toHaveLength(2);
+    expect(result.content[0]).toEqual(mediaParts[0]);
+    expect(result.content[1]).toEqual({
+      type: 'text',
+      text: 'Check out this image',
+    });
+  });
+
+  it('formats message with multiple media types', () => {
+    const message = {
+      role: 'user',
+      content: 'Check out these files',
+    };
+    const mediaParts = [
+      { type: 'document', document: { url: 'https://example.com/doc.pdf' } },
+      { type: 'video', video: { url: 'https://example.com/video.mp4' } },
+      { type: 'audio', audio: { url: 'https://example.com/audio.mp3' } },
+      {
+        type: 'image_url',
+        image_url: { url: 'https://example.com/image.jpg' },
+      },
+    ];
+
+    const result = formatMediaMessage({ message, mediaParts });
+
+    expect(result.content).toHaveLength(5);
+    expect(result.content[0]).toEqual({
+      type: 'text',
+      text: 'Check out these files',
+    });
+    expect(result.content[1]).toEqual(mediaParts[0]);
+    expect(result.content[2]).toEqual(mediaParts[1]);
+    expect(result.content[3]).toEqual(mediaParts[2]);
+    expect(result.content[4]).toEqual(mediaParts[3]);
+  });
+});
+
+describe('formatMessage with media', () => {
+  it('formats user message with image_urls (backward compatibility)', () => {
+    const input = {
+      message: {
+        sender: 'user',
+        text: 'Check out this image',
+        image_urls: [
+          {
+            type: 'image_url' as const,
+            image_url: { url: 'https://example.com/image.jpg' },
+          },
+        ],
+      },
+      userName: 'John',
+    };
+
+    const result = formatMessage(input);
+
+    expect(isFormattedMediaMessage(result)).toBe(true);
+    if (isFormattedMediaMessage(result)) {
+      expect(result.role).toBe('user');
+      expect(result.name).toBe('John');
+      expect(Array.isArray(result.content)).toBe(true);
+      expect(result.content).toHaveLength(2);
+      expect(result.content[0]).toEqual({
+        type: 'text',
+        text: 'Check out this image',
+      });
+      expect(result.content[1]).toEqual(input.message.image_urls[0]);
+    }
+  });
+
+  it('formats user message with documents', () => {
+    const input = {
+      message: {
+        role: 'user',
+        content: 'Review this document',
+        documents: [
+          {
+            type: 'document',
+            document: { url: 'https://example.com/report.pdf' },
+          },
+        ],
+      },
+    };
+
+    const result = formatMessage(input);
+
+    expect(isFormattedMediaMessage(result)).toBe(true);
+    if (isFormattedMediaMessage(result)) {
+      expect(result.role).toBe('user');
+      expect(Array.isArray(result.content)).toBe(true);
+      expect(result.content).toHaveLength(2);
+      expect(result.content[0]).toEqual({
+        type: 'text',
+        text: 'Review this document',
+      });
+      expect(result.content[1]).toEqual(input.message.documents[0]);
+    }
+  });
+
+  it('formats user message with videos', () => {
+    const input = {
+      message: {
+        role: 'user',
+        content: 'Watch this video',
+        videos: [
+          { type: 'video', video: { url: 'https://example.com/demo.mp4' } },
+        ],
+      },
+    };
+
+    const result = formatMessage(input);
+
+    expect(isFormattedMediaMessage(result)).toBe(true);
+    if (isFormattedMediaMessage(result)) {
+      expect(result.role).toBe('user');
+      expect(Array.isArray(result.content)).toBe(true);
+      expect(result.content).toHaveLength(2);
+      expect(result.content[0]).toEqual({
+        type: 'text',
+        text: 'Watch this video',
+      });
+      expect(result.content[1]).toEqual(input.message.videos[0]);
+    }
+  });
+
+  it('formats user message with audios', () => {
+    const input = {
+      message: {
+        role: 'user',
+        content: 'Listen to this',
+        audios: [
+          { type: 'audio', audio: { url: 'https://example.com/podcast.mp3' } },
+        ],
+      },
+    };
+
+    const result = formatMessage(input);
+
+    expect(isFormattedMediaMessage(result)).toBe(true);
+    if (isFormattedMediaMessage(result)) {
+      expect(result.role).toBe('user');
+      expect(Array.isArray(result.content)).toBe(true);
+      expect(result.content).toHaveLength(2);
+      expect(result.content[0]).toEqual({
+        type: 'text',
+        text: 'Listen to this',
+      });
+      expect(result.content[1]).toEqual(input.message.audios[0]);
+    }
+  });
+
+  it('formats user message with all media types in correct order', () => {
+    const input = {
+      message: {
+        role: 'user',
+        content: 'Check out all these files',
+        documents: [
+          {
+            type: 'document',
+            document: { url: 'https://example.com/doc.pdf' },
+          },
+        ],
+        videos: [
+          { type: 'video', video: { url: 'https://example.com/video.mp4' } },
+        ],
+        audios: [
+          { type: 'audio', audio: { url: 'https://example.com/audio.mp3' } },
+        ],
+        image_urls: [
+          {
+            type: 'image_url' as const,
+            image_url: { url: 'https://example.com/image.jpg' },
+          },
+        ],
+      },
+    };
+
+    const result = formatMessage(input);
+
+    expect(isFormattedMediaMessage(result)).toBe(true);
+    if (isFormattedMediaMessage(result)) {
+      expect(result.role).toBe('user');
+      expect(Array.isArray(result.content)).toBe(true);
+      expect(result.content).toHaveLength(5);
+      // Text first
+      expect(result.content[0]).toEqual({
+        type: 'text',
+        text: 'Check out all these files',
+      });
+      // Then documents, videos, audios, images
+      expect(result.content[1]).toEqual(input.message.documents[0]);
+      expect(result.content[2]).toEqual(input.message.videos[0]);
+      expect(result.content[3]).toEqual(input.message.audios[0]);
+      expect(result.content[4]).toEqual(input.message.image_urls[0]);
+    }
+  });
+
+  it('formats user message with multiple files of the same type', () => {
+    const input = {
+      message: {
+        role: 'user',
+        content: 'Review these documents',
+        documents: [
+          {
+            type: 'document',
+            document: { url: 'https://example.com/doc1.pdf' },
+          },
+          {
+            type: 'document',
+            document: { url: 'https://example.com/doc2.pdf' },
+          },
+          {
+            type: 'document',
+            document: { url: 'https://example.com/doc3.pdf' },
+          },
+        ],
+      },
+    };
+
+    const result = formatMessage(input);
+
+    expect(isFormattedMediaMessage(result)).toBe(true);
+    if (isFormattedMediaMessage(result)) {
+      expect(result.content).toHaveLength(4);
+      expect(result.content[0].type).toBe('text');
+      expect(result.content[1]).toEqual(input.message.documents[0]);
+      expect(result.content[2]).toEqual(input.message.documents[1]);
+      expect(result.content[3]).toEqual(input.message.documents[2]);
+    }
+  });
+
+  it('respects Anthropic provider ordering (media before text)', () => {
+    const input = {
+      message: {
+        role: 'user',
+        content: 'Check this out',
+        documents: [
+          {
+            type: 'document',
+            document: { url: 'https://example.com/doc.pdf' },
+          },
+        ],
+        image_urls: [
+          {
+            type: 'image_url' as const,
+            image_url: { url: 'https://example.com/image.jpg' },
+          },
+        ],
+      },
+      endpoint: Providers.ANTHROPIC,
+    };
+
+    const result = formatMessage(input);
+
+    expect(isFormattedMediaMessage(result)).toBe(true);
+    if (isFormattedMediaMessage(result)) {
+      expect(result.content).toHaveLength(3);
+      // Media first for Anthropic
+      expect(result.content[0]).toEqual(input.message.documents[0]);
+      expect(result.content[1]).toEqual(input.message.image_urls[0]);
+      expect(result.content[2]).toEqual({
+        type: 'text',
+        text: 'Check this out',
+      });
+    }
+  });
+
+  it('does not format media for assistant messages', () => {
+    const input = {
+      message: {
+        role: 'assistant',
+        content: 'Here is a response',
+        documents: [
+          {
+            type: 'document',
+            document: { url: 'https://example.com/doc.pdf' },
+          },
+        ],
+      },
+    };
+
+    const result = formatMessage(input);
+
+    expect(result).toMatchObject({
+      role: 'assistant',
+      content: 'Here is a response',
+    });
+  });
+
+  it('handles empty media arrays gracefully', () => {
+    const input = {
+      message: {
+        role: 'user',
+        content: 'Just text',
+        documents: [],
+        videos: [],
+        audios: [],
+        image_urls: [],
+      },
+    };
+
+    const result = formatMessage(input);
+
+    expect(result).toMatchObject({
+      role: 'user',
+      content: 'Just text',
+    });
+  });
+
+  it('formats media with langChain flag', () => {
+    const input = {
+      message: {
+        role: 'user',
+        content: 'Check this image',
+        image_urls: [
+          {
+            type: 'image_url' as const,
+            image_url: { url: 'https://example.com/image.jpg' },
+          },
+        ],
+      },
+      langChain: true,
+    };
+
+    const result = formatMessage(input);
+
+    expect(result).toBeInstanceOf(HumanMessage);
+    expect(Array.isArray(result.lc_kwargs.content)).toBe(true);
+    expect(result.lc_kwargs.content).toHaveLength(2);
+  });
+});
+
 describe('formatLangChainMessages', () => {
   it('formats an array of messages for LangChain', () => {
     const messages = [

package/src/stream.ts

@@ -394,9 +394,13 @@ export function createContentAggregator(): t.ContentAggregatorResult {
 
   const updateContent = (
     index: number,
-    contentPart: t.MessageContentComplex,
+    contentPart?: t.MessageContentComplex,
     finalUpdate = false
   ): void => {
+    if (!contentPart) {
+      console.warn('No content part found in \'updateContent\'');
+      return;
+    }
     const partType = contentPart.type ?? '';
     if (!partType) {
       console.warn('No content type found in content part');
@@ -574,7 +578,10 @@ export function createContentAggregator(): t.ContentAggregatorResult {
       event === GraphEvents.ON_AGENT_UPDATE &&
       (data as t.AgentUpdate | undefined)?.agent_update
     ) {
-      const contentPart = data as t.AgentUpdate;
+      const contentPart = data as t.AgentUpdate | undefined;
+      if (!contentPart) {
+        return;
+      }
       updateContent(contentPart.agent_update.index, contentPart);
     } else if (event === GraphEvents.ON_REASONING_DELTA) {
       const reasoningDelta = data as t.ReasoningDeltaEvent;