0.3.0

changelog.md

@@ -1,5 +1,27 @@
 # Changelog
 
+## 2025-02-05 - 0.3.0 - feat(integration-xai)
+Add support for X.AI provider with chat and document processing capabilities.
+
+- Introduced XAIProvider class for integrating X.AI features.
+- Implemented chat streaming and synchronous chat for X.AI.
+- Enabled document processing capabilities with PDF conversion in X.AI.
+
+## 2025-02-03 - 0.2.0 - feat(provider.anthropic)
+Add support for vision and document processing in Anthropic provider
+
+- Implemented vision tasks for Anthropic provider using Claude-3-opus-20240229 model.
+- Implemented document processing for Anthropic provider, supporting conversion of PDF documents to images and analysis with Claude-3-opus-20240229 model.
+- Updated documentation to reflect the new capabilities of the Anthropic provider.
+
+## 2025-02-03 - 0.1.0 - feat(providers)
+Add vision and document processing capabilities to providers
+
+- OpenAI and Ollama providers now support vision tasks using GPT-4 Vision and Llava models respectively.
+- Document processing has been implemented for OpenAI and Ollama providers, converting PDFs to images for analysis.
+- Introduced abstract methods for vision and document processing in the MultiModalModel class.
+- Updated the readme file with examples for vision and document processing.
+
 ## 2025-02-03 - 0.0.19 - fix(core)
 Enhanced chat streaming and error handling across providers
 

license

@@ -0,0 +1,19 @@
+Copyright (c) 2024 Task Venture Capital GmbH (hello@task.vc)
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@push.rocks/smartai",
-  "version": "0.0.19",
+  "version": "0.3.0",
   "private": false,
   "description": "A TypeScript library for integrating and interacting with multiple AI models, offering capabilities for chat and potentially audio responses.",
   "main": "dist_ts/index.js",

readme.md

@@ -17,16 +17,24 @@ This command installs the package and adds it to your project's dependencies.
 @push.rocks/smartai supports multiple AI providers, each with its own unique capabilities:
 
 ### OpenAI
-- Models: GPT-4, GPT-3.5-turbo
-- Features: Chat, Streaming, Audio Generation
+- Models: GPT-4, GPT-3.5-turbo, GPT-4-vision-preview
+- Features: Chat, Streaming, Audio Generation, Vision, Document Processing
 - Configuration:
   ```typescript
   openaiToken: 'your-openai-token'
   ```
 
+### X.AI
+- Models: Grok-2-latest
+- Features: Chat, Streaming, Document Processing
+- Configuration:
+  ```typescript
+  xaiToken: 'your-xai-token'
+  ```
+
 ### Anthropic
 - Models: Claude-3-opus-20240229
-- Features: Chat, Streaming
+- Features: Chat, Streaming, Vision, Document Processing
 - Configuration:
   ```typescript
   anthropicToken: 'your-anthropic-token'
@@ -49,12 +57,13 @@ This command installs the package and adds it to your project's dependencies.
   ```
 
 ### Ollama
-- Models: Configurable (default: llama2)
-- Features: Chat, Streaming
+- Models: Configurable (default: llama2, llava for vision/documents)
+- Features: Chat, Streaming, Vision, Document Processing
 - Configuration:
   ```typescript
   baseUrl: 'http://localhost:11434' // Optional
   model: 'llama2' // Optional
+  visionModel: 'llava' // Optional, for vision and document tasks
   ```
 
 ## Usage
@@ -74,6 +83,7 @@ import { SmartAi } from '@push.rocks/smartai';
 
 const smartAi = new SmartAi({
   openaiToken: 'your-openai-token',
+  xaiToken: 'your-xai-token',
   anthropicToken: 'your-anthropic-token',
   perplexityToken: 'your-perplexity-token',
   groqToken: 'your-groq-token',
@@ -147,15 +157,61 @@ const audioStream = await smartAi.openaiProvider.audio({
 
 ### Document Processing
 
-For providers that support document processing (currently OpenAI):
+For providers that support document processing (OpenAI, Ollama, and Anthropic):
 
 ```typescript
+// Using OpenAI
 const result = await smartAi.openaiProvider.document({
   systemMessage: 'Classify the document type',
   userMessage: 'What type of document is this?',
   messageHistory: [],
   pdfDocuments: [pdfBuffer] // Uint8Array of PDF content
 });
+
+// Using Ollama with llava
+const analysis = await smartAi.ollamaProvider.document({
+  systemMessage: 'You are a document analysis assistant',
+  userMessage: 'Extract the key information from this document',
+  messageHistory: [],
+  pdfDocuments: [pdfBuffer] // Uint8Array of PDF content
+});
+
+// Using Anthropic with Claude 3
+const anthropicAnalysis = await smartAi.anthropicProvider.document({
+  systemMessage: 'You are a document analysis assistant',
+  userMessage: 'Please analyze this document and extract key information',
+  messageHistory: [],
+  pdfDocuments: [pdfBuffer] // Uint8Array of PDF content
+});
+```
+
+Both providers will:
+1. Convert PDF documents to images
+2. Process each page using their vision models
+3. Return a comprehensive analysis based on the system message and user query
+
+### Vision Processing
+
+For providers that support vision tasks (OpenAI, Ollama, and Anthropic):
+
+```typescript
+// Using OpenAI's GPT-4 Vision
+const description = await smartAi.openaiProvider.vision({
+  image: imageBuffer, // Buffer containing the image data
+  prompt: 'What do you see in this image?'
+});
+
+// Using Ollama's Llava model
+const analysis = await smartAi.ollamaProvider.vision({
+  image: imageBuffer,
+  prompt: 'Analyze this image in detail'
+});
+
+// Using Anthropic's Claude 3
+const anthropicAnalysis = await smartAi.anthropicProvider.vision({
+  image: imageBuffer,
+  prompt: 'Please analyze this image and describe what you see'
+});
 ```
 
 ## Error Handling

ts/00_commitinfo_data.ts

@@ -3,6 +3,6 @@
  */
 export const commitinfo = {
   name: '@push.rocks/smartai',
-  version: '0.0.19',
+  version: '0.3.0',
   description: 'A TypeScript library for integrating and interacting with multiple AI models, offering capabilities for chat and potentially audio responses.'
 }

ts/abstract.classes.multimodal.ts

@@ -62,4 +62,25 @@ export abstract class MultiModalModel {
    * @throws Error if the provider doesn't support audio generation
    */
   public abstract audio(optionsArg: { message: string }): Promise<NodeJS.ReadableStream>;
+
+  /**
+   * Vision-language processing
+   * @param optionsArg Options containing the image and prompt for analysis
+   * @returns Promise resolving to the model's description or analysis of the image
+   * @throws Error if the provider doesn't support vision tasks
+   */
+  public abstract vision(optionsArg: { image: Buffer; prompt: string }): Promise<string>;
+
+  /**
+   * Document analysis and processing
+   * @param optionsArg Options containing system message, user message, PDF documents, and message history
+   * @returns Promise resolving to the model's analysis of the documents
+   * @throws Error if the provider doesn't support document processing
+   */
+  public abstract document(optionsArg: {
+    systemMessage: string;
+    userMessage: string;
+    pdfDocuments: Uint8Array[];
+    messageHistory: ChatMessage[];
+  }): Promise<{ message: any }>;
 }

ts/provider.anthropic.ts

@@ -2,6 +2,9 @@ import * as plugins from './plugins.js';
 import * as paths from './paths.js';
 import { MultiModalModel } from './abstract.classes.multimodal.js';
 import type { ChatOptions, ChatResponse, ChatMessage } from './abstract.classes.multimodal.js';
+import type { ImageBlockParam, TextBlockParam } from '@anthropic-ai/sdk/resources/messages';
+
+type ContentBlock = ImageBlockParam | TextBlockParam;
 
 export interface IAnthropicProviderOptions {
   anthropicToken: string;
@@ -130,4 +133,108 @@ export class AnthropicProvider extends MultiModalModel {
     // Anthropic does not provide an audio API, so this method is not implemented.
     throw new Error('Audio generation is not yet supported by Anthropic.');
   }
+
+  public async vision(optionsArg: { image: Buffer; prompt: string }): Promise<string> {
+    const base64Image = optionsArg.image.toString('base64');
+    
+    const content: ContentBlock[] = [
+      {
+        type: 'text',
+        text: optionsArg.prompt
+      },
+      {
+        type: 'image',
+        source: {
+          type: 'base64',
+          media_type: 'image/jpeg',
+          data: base64Image
+        }
+      }
+    ];
+
+    const result = await this.anthropicApiClient.messages.create({
+      model: 'claude-3-opus-20240229',
+      messages: [{
+        role: 'user',
+        content
+      }],
+      max_tokens: 1024
+    });
+
+    // Extract text content from the response
+    let message = '';
+    for (const block of result.content) {
+      if ('text' in block) {
+        message += block.text;
+      }
+    }
+    return message;
+  }
+
+  public async document(optionsArg: {
+    systemMessage: string;
+    userMessage: string;
+    pdfDocuments: Uint8Array[];
+    messageHistory: ChatMessage[];
+  }): Promise<{ message: any }> {
+    // Convert PDF documents to images using SmartPDF
+    const smartpdfInstance = new plugins.smartpdf.SmartPdf();
+    let documentImageBytesArray: Uint8Array[] = [];
+
+    for (const pdfDocument of optionsArg.pdfDocuments) {
+      const documentImageArray = await smartpdfInstance.convertPDFToPngBytes(pdfDocument);
+      documentImageBytesArray = documentImageBytesArray.concat(documentImageArray);
+    }
+
+    // Convert message history to Anthropic format
+    const messages = optionsArg.messageHistory.map(msg => ({
+      role: msg.role === 'assistant' ? 'assistant' as const : 'user' as const,
+      content: msg.content
+    }));
+
+    // Create content array with text and images
+    const content: ContentBlock[] = [
+      {
+        type: 'text',
+        text: optionsArg.userMessage
+      }
+    ];
+
+    // Add each document page as an image
+    for (const imageBytes of documentImageBytesArray) {
+      content.push({
+        type: 'image',
+        source: {
+          type: 'base64',
+          media_type: 'image/jpeg',
+          data: Buffer.from(imageBytes).toString('base64')
+        }
+      });
+    }
+
+    const result = await this.anthropicApiClient.messages.create({
+      model: 'claude-3-opus-20240229',
+      system: optionsArg.systemMessage,
+      messages: [
+        ...messages,
+        { role: 'user', content }
+      ],
+      max_tokens: 4096
+    });
+
+    // Extract text content from the response
+    let message = '';
+    for (const block of result.content) {
+      if ('text' in block) {
+        message += block.text;
+      }
+    }
+
+    return {
+      message: {
+        role: 'assistant',
+        content: message
+      }
+    };
+  }
 }

ts/provider.groq.ts

@@ -176,4 +176,17 @@ export class GroqProvider extends MultiModalModel {
     // Groq does not provide an audio API, so this method is not implemented.
     throw new Error('Audio generation is not yet supported by Groq.');
   }
+
+  public async vision(optionsArg: { image: Buffer; prompt: string }): Promise<string> {
+    throw new Error('Vision tasks are not yet supported by Groq.');
+  }
+
+  public async document(optionsArg: {
+    systemMessage: string;
+    userMessage: string;
+    pdfDocuments: Uint8Array[];
+    messageHistory: ChatMessage[];
+  }): Promise<{ message: any }> {
+    throw new Error('Document processing is not yet supported by Groq.');
+  }
 }

ts/provider.ollama.ts

@@ -6,18 +6,21 @@ import type { ChatOptions, ChatResponse, ChatMessage } from './abstract.classes.
 export interface IOllamaProviderOptions {
   baseUrl?: string;
   model?: string;
+  visionModel?: string; // Model to use for vision tasks (e.g. 'llava')
 }
 
 export class OllamaProvider extends MultiModalModel {
   private options: IOllamaProviderOptions;
   private baseUrl: string;
   private model: string;
+  private visionModel: string;
 
   constructor(optionsArg: IOllamaProviderOptions = {}) {
     super();
     this.options = optionsArg;
     this.baseUrl = optionsArg.baseUrl || 'http://localhost:11434';
     this.model = optionsArg.model || 'llama2';
+    this.visionModel = optionsArg.visionModel || 'llava';
   }
 
   async start() {
@@ -167,4 +170,83 @@ export class OllamaProvider extends MultiModalModel {
   public async audio(optionsArg: { message: string }): Promise<NodeJS.ReadableStream> {
     throw new Error('Audio generation is not supported by Ollama.');
   }
+
+  public async vision(optionsArg: { image: Buffer; prompt: string }): Promise<string> {
+    const base64Image = optionsArg.image.toString('base64');
+    
+    const response = await fetch(`${this.baseUrl}/api/chat`, {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+      },
+      body: JSON.stringify({
+        model: this.visionModel,
+        messages: [{
+          role: 'user',
+          content: optionsArg.prompt,
+          images: [base64Image]
+        }],
+        stream: false
+      }),
+    });
+
+    if (!response.ok) {
+      throw new Error(`Ollama API error: ${response.statusText}`);
+    }
+
+    const result = await response.json();
+    return result.message.content;
+  }
+
+  public async document(optionsArg: {
+    systemMessage: string;
+    userMessage: string;
+    pdfDocuments: Uint8Array[];
+    messageHistory: ChatMessage[];
+  }): Promise<{ message: any }> {
+    // Convert PDF documents to images using SmartPDF
+    const smartpdfInstance = new plugins.smartpdf.SmartPdf();
+    let documentImageBytesArray: Uint8Array[] = [];
+
+    for (const pdfDocument of optionsArg.pdfDocuments) {
+      const documentImageArray = await smartpdfInstance.convertPDFToPngBytes(pdfDocument);
+      documentImageBytesArray = documentImageBytesArray.concat(documentImageArray);
+    }
+
+    // Convert images to base64
+    const base64Images = documentImageBytesArray.map(bytes => Buffer.from(bytes).toString('base64'));
+
+    // Send request to Ollama with images
+    const response = await fetch(`${this.baseUrl}/api/chat`, {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+      },
+      body: JSON.stringify({
+        model: this.visionModel,
+        messages: [
+          { role: 'system', content: optionsArg.systemMessage },
+          ...optionsArg.messageHistory,
+          {
+            role: 'user',
+            content: optionsArg.userMessage,
+            images: base64Images
+          }
+        ],
+        stream: false
+      }),
+    });
+
+    if (!response.ok) {
+      throw new Error(`Ollama API error: ${response.statusText}`);
+    }
+
+    const result = await response.json();
+    return {
+      message: {
+        role: 'assistant',
+        content: result.message.content
+      }
+    };
+  }
 }

ts/provider.openai.ts

@@ -192,4 +192,27 @@ export class OpenAiProvider extends MultiModalModel {
       message: result.choices[0].message,
     };
   }
+
+  public async vision(optionsArg: { image: Buffer; prompt: string }): Promise<string> {
+    const result = await this.openAiApiClient.chat.completions.create({
+      model: 'gpt-4-vision-preview',
+      messages: [
+        {
+          role: 'user',
+          content: [
+            { type: 'text', text: optionsArg.prompt },
+            {
+              type: 'image_url',
+              image_url: {
+                url: `data:image/jpeg;base64,${optionsArg.image.toString('base64')}`
+              }
+            }
+          ]
+        }
+      ],
+      max_tokens: 300
+    });
+
+    return result.choices[0].message.content || '';
+  }
 }

ts/provider.perplexity.ts

@@ -155,4 +155,17 @@ export class PerplexityProvider extends MultiModalModel {
   public async audio(optionsArg: { message: string }): Promise<NodeJS.ReadableStream> {
     throw new Error('Audio generation is not supported by Perplexity.');
   }
+
+  public async vision(optionsArg: { image: Buffer; prompt: string }): Promise<string> {
+    throw new Error('Vision tasks are not supported by Perplexity.');
+  }
+
+  public async document(optionsArg: {
+    systemMessage: string;
+    userMessage: string;
+    pdfDocuments: Uint8Array[];
+    messageHistory: ChatMessage[];
+  }): Promise<{ message: any }> {
+    throw new Error('Document processing is not supported by Perplexity.');
+  }
 }

ts/provider.xai.ts

@@ -0,0 +1,183 @@
+import * as plugins from './plugins.js';
+import * as paths from './paths.js';
+import { MultiModalModel } from './abstract.classes.multimodal.js';
+import type { ChatOptions, ChatResponse, ChatMessage } from './abstract.classes.multimodal.js';
+import type { ChatCompletionMessageParam } from 'openai/resources/chat/completions';
+
+export interface IXAIProviderOptions {
+  xaiToken: string;
+}
+
+export class XAIProvider extends MultiModalModel {
+  private options: IXAIProviderOptions;
+  public openAiApiClient: plugins.openai.default;
+  public smartpdfInstance: plugins.smartpdf.SmartPdf;
+
+  constructor(optionsArg: IXAIProviderOptions) {
+    super();
+    this.options = optionsArg;
+  }
+
+  public async start() {
+    this.openAiApiClient = new plugins.openai.default({
+      apiKey: this.options.xaiToken,
+      baseURL: 'https://api.x.ai/v1',
+    });
+    this.smartpdfInstance = new plugins.smartpdf.SmartPdf();
+  }
+
+  public async stop() {}
+
+  public async chatStream(input: ReadableStream<Uint8Array>): Promise<ReadableStream<string>> {
+    // Create a TextDecoder to handle incoming chunks
+    const decoder = new TextDecoder();
+    let buffer = '';
+    let currentMessage: { role: string; content: string; } | null = null;
+
+    // Create a TransformStream to process the input
+    const transform = new TransformStream<Uint8Array, string>({
+      async transform(chunk, controller) {
+        buffer += decoder.decode(chunk, { stream: true });
+
+        // Try to parse complete JSON messages from the buffer
+        while (true) {
+          const newlineIndex = buffer.indexOf('\n');
+          if (newlineIndex === -1) break;
+
+          const line = buffer.slice(0, newlineIndex);
+          buffer = buffer.slice(newlineIndex + 1);
+
+          if (line.trim()) {
+            try {
+              const message = JSON.parse(line);
+              currentMessage = {
+                role: message.role || 'user',
+                content: message.content || '',
+              };
+            } catch (e) {
+              console.error('Failed to parse message:', e);
+            }
+          }
+        }
+
+        // If we have a complete message, send it to X.AI
+        if (currentMessage) {
+          const stream = await this.openAiApiClient.chat.completions.create({
+            model: 'grok-2-latest',
+            messages: [{ role: currentMessage.role, content: currentMessage.content }],
+            stream: true,
+          });
+
+          // Process each chunk from X.AI
+          for await (const chunk of stream) {
+            const content = chunk.choices[0]?.delta?.content;
+            if (content) {
+              controller.enqueue(content);
+            }
+          }
+
+          currentMessage = null;
+        }
+      },
+
+      flush(controller) {
+        if (buffer) {
+          try {
+            const message = JSON.parse(buffer);
+            controller.enqueue(message.content || '');
+          } catch (e) {
+            console.error('Failed to parse remaining buffer:', e);
+          }
+        }
+      }
+    });
+
+    // Connect the input to our transform stream
+    return input.pipeThrough(transform);
+  }
+
+  public async chat(optionsArg: {
+    systemMessage: string;
+    userMessage: string;
+    messageHistory: { role: string; content: string; }[];
+  }): Promise<{ role: 'assistant'; message: string; }> {
+    // Prepare messages array with system message, history, and user message
+    const messages: ChatCompletionMessageParam[] = [
+      { role: 'system', content: optionsArg.systemMessage },
+      ...optionsArg.messageHistory.map(msg => ({
+        role: msg.role as 'system' | 'user' | 'assistant',
+        content: msg.content
+      })),
+      { role: 'user', content: optionsArg.userMessage }
+    ];
+
+    // Call X.AI's chat completion API
+    const completion = await this.openAiApiClient.chat.completions.create({
+      model: 'grok-2-latest',
+      messages: messages,
+      stream: false,
+    });
+
+    // Return the assistant's response
+    return {
+      role: 'assistant',
+      message: completion.choices[0]?.message?.content || ''
+    };
+  }
+
+  public async audio(optionsArg: { message: string }): Promise<NodeJS.ReadableStream> {
+    throw new Error('Audio generation is not supported by X.AI');
+  }
+
+  public async vision(optionsArg: { image: Buffer; prompt: string }): Promise<string> {
+    throw new Error('Vision tasks are not supported by X.AI');
+  }
+
+  public async document(optionsArg: {
+    systemMessage: string;
+    userMessage: string;
+    pdfDocuments: Uint8Array[];
+    messageHistory: { role: string; content: string; }[];
+  }): Promise<{ message: any }> {
+    // First convert PDF documents to images
+    let pdfDocumentImageBytesArray: Uint8Array[] = [];
+    
+    for (const pdfDocument of optionsArg.pdfDocuments) {
+      const documentImageArray = await this.smartpdfInstance.convertPDFToPngBytes(pdfDocument);
+      pdfDocumentImageBytesArray = pdfDocumentImageBytesArray.concat(documentImageArray);
+    }
+
+    // Convert images to base64 for inclusion in the message
+    const imageBase64Array = pdfDocumentImageBytesArray.map(bytes => 
+      Buffer.from(bytes).toString('base64')
+    );
+
+    // Combine document images into the user message
+    const enhancedUserMessage = `
+      ${optionsArg.userMessage}
+      
+      Document contents (as images):
+      ${imageBase64Array.map((img, i) => `Image ${i + 1}: <image data>`).join('\n')}
+    `;
+
+    // Use chat completion to analyze the documents
+    const messages: ChatCompletionMessageParam[] = [
+      { role: 'system', content: optionsArg.systemMessage },
+      ...optionsArg.messageHistory.map(msg => ({
+        role: msg.role as 'system' | 'user' | 'assistant',
+        content: msg.content
+      })),
+      { role: 'user', content: enhancedUserMessage }
+    ];
+
+    const completion = await this.openAiApiClient.chat.completions.create({
+      model: 'grok-2-latest',
+      messages: messages,
+      stream: false,
+    });
+
+    return {
+      message: completion.choices[0]?.message?.content || ''
+    };
+  }
+}