Merge pull request #1074 from narengogi/chore/anthropic-pdf

anthropic pdf

Author: VisargD

src/providers/anthropic/chatComplete.ts

@@ -63,11 +63,40 @@ interface AnthropicTextContentItem {
   text: string;
 }
 
+interface AnthropicUrlPdfContentItem {
+  type: string;
+  source: {
+    type: string;
+    url: string;
+  };
+}
+
+interface AnthropicBase64PdfContentItem {
+  type: string;
+  source: {
+    type: string;
+    data: string;
+    media_type: string;
+  };
+}
+
+interface AnthropicPlainTextContentItem {
+  type: string;
+  source: {
+    type: string;
+    data: string;
+    media_type: string;
+  };
+}
+
 type AnthropicMessageContentItem =
   | AnthropicToolResultContentItem
   | AnthropicBase64ImageContentItem
   | AnthropicUrlImageContentItem
-  | AnthropicTextContentItem;
+  | AnthropicTextContentItem
+  | AnthropicUrlPdfContentItem
+  | AnthropicBase64PdfContentItem
+  | AnthropicPlainTextContentItem;
 
 interface AnthropicMessage extends Message, PromptCache {
   content: AnthropicMessageContentItem[];
@@ -166,6 +195,35 @@ const transformAndAppendImageContentItem = (
   }
 };
 
+const transformAndAppendFileContentItem = (
+  item: ContentType,
+  transformedMessage: AnthropicMessage
+) => {
+  const mimeType =
+    (item.file?.mime_type as keyof typeof fileExtensionMimeTypeMap) ||
+    fileExtensionMimeTypeMap.pdf;
+  if (item.file?.file_url) {
+    transformedMessage.content.push({
+      type: 'document',
+      source: {
+        type: 'url',
+        url: item.file.file_url,
+      },
+    });
+  } else if (item.file?.file_data) {
+    const contentType =
+      mimeType === fileExtensionMimeTypeMap.txt ? 'text' : 'base64';
+    transformedMessage.content.push({
+      type: 'document',
+      source: {
+        type: contentType,
+        data: item.file.file_data,
+        media_type: mimeType,
+      },
+    });
+  }
+};
+
 export const AnthropicChatCompleteConfig: ProviderConfig = {
   model: {
     param: 'model',
@@ -205,6 +263,8 @@ export const AnthropicChatCompleteConfig: ProviderConfig = {
                   });
                 } else if (item.type === 'image_url') {
                   transformAndAppendImageContentItem(item, transformedMessage);
+                } else if (item.type === 'file') {
+                  transformAndAppendFileContentItem(item, transformedMessage);
                 }
               });
               messages.push(transformedMessage as AnthropicMessage);

src/providers/bedrock/chatComplete.ts

@@ -1,4 +1,9 @@
-import { BEDROCK, documentMimeTypes, imagesMimeTypes } from '../../globals';
+import {
+  BEDROCK,
+  documentMimeTypes,
+  fileExtensionMimeTypeMap,
+  imagesMimeTypes,
+} from '../../globals';
 import {
   Message,
   Params,
@@ -183,6 +188,32 @@ const getMessageContent = (message: Message) => {
             },
           });
         }
+      } else if (item.type === 'file') {
+        const mimeType = item.file?.mime_type || fileExtensionMimeTypeMap.pdf;
+        const fileFormat = mimeType.split('/')[1];
+        if (item.file?.file_url) {
+          out.push({
+            document: {
+              format: fileFormat,
+              name: item.file.file_name || crypto.randomUUID(),
+              source: {
+                s3Location: {
+                  uri: item.file.file_url,
+                },
+              },
+            },
+          });
+        } else if (item.file?.file_data) {
+          out.push({
+            document: {
+              format: fileFormat,
+              name: item.file.file_name || crypto.randomUUID(),
+              source: {
+                bytes: item.file.file_data,
+              },
+            },
+          });
+        }
       }
 
       if (item.cache_control) {
@@ -402,7 +433,10 @@ type BedrockContentItem = {
     format: string;
     name: string;
     source: {
-      bytes: string;
+      bytes?: string;
+      s3Location?: {
+        uri: string;
+      };
     };
   };
   cachePoint?: {

src/types/requestBody.ts

@@ -216,6 +216,7 @@ export interface Config {
 }
 
 /**
+ * TODO: make this a union type
  * A message content type.
  * @interface
  */
@@ -230,6 +231,17 @@ export interface ContentType extends PromptCache {
     mime_type?: string;
   };
   data?: string;
+  file?: {
+    file_data?: string;
+    file_id?: string;
+    file_name?: string;
+    file_url?: string;
+    mime_type?: string;
+  };
+  input_audio?: {
+    data: string;
+    format: string; //defaults to auto
+  };
 }
 
 export interface ToolCall {