claude-autopm 1.30.1 → 2.1.0

package/lib/ai-providers/ClaudeProvider.js

@@ -0,0 +1,423 @@
+/**
+ * ClaudeProvider - Anthropic Claude API Integration
+ *
+ * Extends AbstractAIProvider to provide Claude-specific AI capabilities
+ * with full backward compatibility for existing code.
+ *
+ * Documentation Queries:
+ * - mcp://context7/anthropic/sdk - Anthropic SDK patterns
+ * - mcp://context7/anthropic/streaming - Streaming best practices
+ * - mcp://context7/nodejs/async-generators - Async generator patterns
+ *
+ * @extends AbstractAIProvider
+ *
+ * @example
+ * // Legacy usage (backward compatible)
+ * const provider = new ClaudeProvider('sk-ant-...');
+ * const result = await provider.complete('Hello');
+ *
+ * @example
+ * // Enhanced usage with config
+ * const provider = new ClaudeProvider({
+ *   apiKey: 'sk-ant-...',
+ *   model: 'claude-sonnet-4-20250514',
+ *   temperature: 0.7,
+ *   maxTokens: 4096
+ * });
+ *
+ * @example
+ * // Using environment variable
+ * process.env.ANTHROPIC_API_KEY = 'sk-ant-...';
+ * const provider = new ClaudeProvider({});
+ */
+
+const Anthropic = require('@anthropic-ai/sdk');
+const AbstractAIProvider = require('./AbstractAIProvider');
+const AIProviderError = require('../errors/AIProviderError');
+
+/**
+ * ClaudeProvider class for Anthropic Claude API integration
+ *
+ * @class ClaudeProvider
+ * @extends AbstractAIProvider
+ */
+class ClaudeProvider extends AbstractAIProvider {
+  /**
+   * Create a new ClaudeProvider instance
+   *
+   * Supports two constructor signatures for backward compatibility:
+   * 1. new ClaudeProvider('api-key')  <- Legacy
+   * 2. new ClaudeProvider({ apiKey: 'key', ...options })  <- Enhanced
+   *
+   * @param {string|Object} [config={}] - API key string or configuration object
+   * @param {string} [config.apiKey] - Anthropic API key (or use ANTHROPIC_API_KEY env var)
+   * @param {string} [config.model] - Model to use (default: claude-sonnet-4-20250514)
+   * @param {number} [config.maxTokens] - Maximum tokens (default: 4096)
+   * @param {number} [config.temperature] - Temperature (default: 0.7)
+   *
+   * @throws {Error} If API key is not provided and ANTHROPIC_API_KEY env var is not set
+   */
+  constructor(config = {}) {
+    // Handle backward compatibility: string API key as first parameter
+    if (typeof config === 'string') {
+      config = { apiKey: config };
+    }
+
+    // Call parent constructor (handles config normalization and defaults)
+    super(config);
+
+    // Validate API key is available after parent constructor
+    if (!this.apiKey) {
+      throw new Error(
+        'API key is required for ClaudeProvider. ' +
+        'Provide it via constructor or set ANTHROPIC_API_KEY environment variable.'
+      );
+    }
+
+    // Initialize Anthropic client with API key
+    this.client = new Anthropic({ apiKey: this.apiKey });
+  }
+
+  // ============================================================
+  // ABSTRACT METHOD IMPLEMENTATIONS (Required by AbstractAIProvider)
+  // ============================================================
+
+  /**
+   * Get the default model identifier for Claude
+   *
+   * @returns {string} Default Claude model
+   */
+  getDefaultModel() {
+    return 'claude-sonnet-4-20250514';
+  }
+
+  /**
+   * Get the environment variable name for API key
+   *
+   * @returns {string} Environment variable name
+   */
+  getApiKeyEnvVar() {
+    return 'ANTHROPIC_API_KEY';
+  }
+
+  /**
+   * Complete a prompt synchronously (wait for full response)
+   *
+   * Uses parent's _mergeOptions to combine instance config with method options.
+   * Method-level options take precedence over instance config.
+   *
+   * Automatically applies rate limiting if configured in constructor.
+   *
+   * @param {string} prompt - The prompt to complete
+   * @param {Object} [options={}] - Optional configuration
+   * @param {string} [options.model] - Model to use (overrides instance model)
+   * @param {number} [options.maxTokens] - Maximum tokens (overrides instance maxTokens)
+   * @param {number} [options.temperature] - Temperature (overrides instance temperature)
+   *
+   * @returns {Promise<string>} The completed text
+   * @throws {AIProviderError} On API errors
+   */
+  async complete(prompt, options = {}) {
+    return this._withRateLimit(async () => {
+      // Merge instance config with method options
+      const finalOptions = this._mergeOptions(options);
+
+      try {
+        const response = await this.client.messages.create({
+          model: finalOptions.model,
+          max_tokens: finalOptions.maxTokens,
+          temperature: finalOptions.temperature,
+          messages: [{ role: 'user', content: prompt }]
+        });
+
+        // Extract text from response
+        if (response.content && response.content.length > 0) {
+          return response.content[0].text || '';
+        }
+
+        return '';
+      } catch (error) {
+        // Format error using parent's formatError + Claude-specific mapping
+        throw this.formatError(error);
+      }
+    });
+  }
+
+  /**
+   * Stream a prompt response (async generator for real-time feedback)
+   *
+   * Uses parent's _mergeOptions to combine instance config with method options.
+   * Method-level options take precedence over instance config.
+   *
+   * Automatically applies rate limiting if configured in constructor.
+   *
+   * @param {string} prompt - The prompt to complete
+   * @param {Object} [options={}] - Optional configuration
+   * @param {string} [options.model] - Model to use (overrides instance model)
+   * @param {number} [options.maxTokens] - Maximum tokens (overrides instance maxTokens)
+   * @param {number} [options.temperature] - Temperature (overrides instance temperature)
+   *
+   * @yields {string} Text chunks as they arrive
+   * @throws {AIProviderError} On API errors
+   */
+  async *stream(prompt, options = {}) {
+    // Apply rate limiting before initiating stream
+    if (this.rateLimiter) {
+      await this.rateLimiter.removeTokens(1);
+    }
+
+    // Merge instance config with method options
+    const finalOptions = this._mergeOptions(options);
+
+    try {
+      const stream = await this.client.messages.create({
+        model: finalOptions.model,
+        max_tokens: finalOptions.maxTokens,
+        temperature: finalOptions.temperature,
+        stream: true,
+        messages: [{ role: 'user', content: prompt }]
+      });
+
+      // Yield text deltas as they arrive
+      for await (const event of stream) {
+        if (
+          event.type === 'content_block_delta' &&
+          event.delta &&
+          event.delta.type === 'text_delta'
+        ) {
+          yield event.delta.text;
+        }
+      }
+    } catch (error) {
+      // Format error using parent's formatError + Claude-specific mapping
+      throw this.formatError(error);
+    }
+  }
+
+  // ============================================================
+  // CAPABILITY OVERRIDES
+  // ============================================================
+
+  /**
+   * Check if Claude supports streaming
+   *
+   * @returns {boolean} True (Claude supports streaming)
+   */
+  supportsStreaming() {
+    return true;
+  }
+
+  /**
+   * Check if Claude supports function calling
+   *
+   * @returns {boolean} True (Claude supports tools/function calling)
+   */
+  supportsFunctionCalling() {
+    return true;
+  }
+
+  /**
+   * Check if Claude supports chat format
+   *
+   * @returns {boolean} True (Claude has native chat format)
+   */
+  supportsChat() {
+    return true;
+  }
+
+  /**
+   * Check if Claude supports vision/image inputs
+   *
+   * @returns {boolean} False (not implemented yet)
+   */
+  supportsVision() {
+    return false;
+  }
+
+  // ============================================================
+  // ENHANCED ERROR HANDLING
+  // ============================================================
+
+  /**
+   * Format Anthropic API errors into AIProviderError
+   *
+   * Maps Anthropic-specific error codes and HTTP statuses to AIProviderError codes.
+   * Falls back to parent's formatError for unknown errors.
+   *
+   * @param {Error} error - The error to format
+   * @returns {AIProviderError} Formatted error with appropriate code and metadata
+   */
+  formatError(error) {
+    // Already an AIProviderError, return as-is
+    if (error instanceof AIProviderError) {
+      return error;
+    }
+
+    // Map Anthropic HTTP status codes to AIProviderError codes
+    if (error.status) {
+      switch (error.status) {
+        case 401:
+          return new AIProviderError(
+            AIProviderError.INVALID_API_KEY,
+            'Invalid Anthropic API key or authentication failed',
+            true,
+            401
+          );
+
+        case 429:
+          return new AIProviderError(
+            AIProviderError.RATE_LIMIT,
+            'Claude API rate limit exceeded. Please retry after a delay.',
+            true,
+            429
+          );
+
+        case 500:
+        case 502:
+        case 503:
+        case 504:
+          return new AIProviderError(
+            AIProviderError.SERVICE_UNAVAILABLE,
+            `Claude API service temporarily unavailable (${error.status})`,
+            true,
+            error.status
+          );
+
+        case 400:
+          return new AIProviderError(
+            AIProviderError.INVALID_REQUEST,
+            error.message || 'Invalid request parameters',
+            true,
+            400
+          );
+
+        default:
+          // Unknown status code, wrap in generic error
+          return new AIProviderError(
+            'UNKNOWN_ERROR',
+            `Claude API error (${error.status}): ${error.message}`,
+            true,
+            error.status
+          );
+      }
+    }
+
+    // No status code, check for specific error types
+    if (error.message) {
+      const lowerMessage = error.message.toLowerCase();
+
+      if (lowerMessage.includes('api key') || lowerMessage.includes('authentication')) {
+        return new AIProviderError(
+          AIProviderError.INVALID_API_KEY,
+          error.message,
+          true
+        );
+      }
+
+      if (lowerMessage.includes('rate limit')) {
+        return new AIProviderError(
+          AIProviderError.RATE_LIMIT,
+          error.message,
+          true
+        );
+      }
+
+      if (lowerMessage.includes('network') || lowerMessage.includes('connection')) {
+        return new AIProviderError(
+          AIProviderError.NETWORK_ERROR,
+          error.message,
+          true
+        );
+      }
+
+      if (lowerMessage.includes('context') || lowerMessage.includes('too long')) {
+        return new AIProviderError(
+          AIProviderError.CONTEXT_LENGTH_EXCEEDED,
+          error.message,
+          true
+        );
+      }
+    }
+
+    // Fall back to parent's error handling
+    return super.formatError(error);
+  }
+
+  // ============================================================
+  // ENHANCED CHAT SUPPORT
+  // ============================================================
+
+  /**
+   * Chat completion with message history
+   *
+   * Overrides parent's fallback implementation to use Claude's native chat format.
+   * Converts system role to user role (Claude requirement).
+   *
+   * Automatically applies rate limiting if configured in constructor.
+   *
+   * @param {Array<{role: string, content: string}>} messages - Chat messages
+   * @param {Object} [options={}] - Optional configuration
+   *
+   * @returns {Promise<string>} Completion response
+   * @throws {AIProviderError} On API errors
+   */
+  async chat(messages, options = {}) {
+    return this._withRateLimit(async () => {
+      // Merge instance config with method options
+      const finalOptions = this._mergeOptions(options);
+
+      try {
+        // Convert system role to user (Claude doesn't support system role in messages array)
+        const claudeMessages = messages.map(msg => ({
+          role: msg.role === 'system' ? 'user' : msg.role,
+          content: msg.content
+        }));
+
+        const response = await this.client.messages.create({
+          model: finalOptions.model,
+          max_tokens: finalOptions.maxTokens,
+          temperature: finalOptions.temperature,
+          messages: claudeMessages
+        });
+
+        // Extract text from response
+        if (response.content && response.content.length > 0) {
+          return response.content[0].text || '';
+        }
+
+        return '';
+      } catch (error) {
+        throw this.formatError(error);
+      }
+    });
+  }
+
+  // ============================================================
+  // BACKWARD COMPATIBILITY METHODS
+  // ============================================================
+
+  /**
+   * Get the current model being used (legacy method for backward compatibility)
+   *
+   * Note: This returns the default model, not the instance model.
+   * For instance model, use provider.model property.
+   *
+   * @returns {string} The default model name
+   * @deprecated Use getDefaultModel() or access provider.model instead
+   */
+  getModel() {
+    return this.getDefaultModel();
+  }
+
+  /**
+   * Test the API connection (inherited from AbstractAIProvider)
+   *
+   * This method is provided by the parent class and works by making a test
+   * complete() call. Included here for documentation completeness.
+   *
+   * @returns {Promise<boolean>} True if connection is successful
+   */
+  // testConnection() is inherited from AbstractAIProvider
+}
+
+module.exports = ClaudeProvider;