AI Code Toolkit

/** * LlmProxyService * * DESIGN PATTERNS: * - Proxy pattern: Provides a unified interface for multiple LLM backends * - Factory pattern: Creates appropriate service based on tool type * - Strategy pattern: Delegates to specific service implementation * * CODING STANDARDS: * - Service class names use PascalCase with 'Service' suffix * - Method names use camelCase with descriptive verbs * - Return types should be explicit (never use implicit any) * - Use async/await for asynchronous operations * * AVOID: * - Direct coupling to specific service implementations outside factory * - Exposing internal service instances */ import { type LlmToolId, LLM_TOOL_CLAUDE_CODE, LLM_TOOL_CODEX, LLM_TOOL_GEMINI_CLI, LLM_TOOL_GITHUB_COPILOT, isValidLlmTool, LlmToolConfig, } from '../constants/llmTools'; import type { CodingAgentService, LlmInvocationParams, LlmInvocationResponse } from '../types'; import { ClaudeCodeService } from './ClaudeCodeService'; import { CodexService } from './CodexService'; import { GeminiCliService } from './GeminiCliService'; import { GitHubCopilotService } from './GitHubCopilotService'; interface LlmProxyServiceOptions { /** The LLM tool to use */ llmTool: LlmToolId; /** Workspace root directory */ workspaceRoot?: string; /** Default timeout in milliseconds */ defaultTimeout?: number; /** Tool-specific configuration (spread to service constructor) */ toolConfig?: Record<string, unknown>; } /** * Proxy service that provides a unified interface for multiple LLM backends * Use this service when you need to support multiple LLM tools interchangeably * * @example * ```ts * import { LlmProxyService, LLM_TOOL_CLAUDE_CODE } from '@agiflowai/coding-agent-bridge'; * * const llm = new LlmProxyService({ llmTool: LLM_TOOL_CLAUDE_CODE }); * const response = await llm.invokeAsLlm({ prompt: 'Hello' }); * ``` */ export class LlmProxyService { private service: CodingAgentService; private readonly llmTool: LlmToolId; constructor(options: LlmProxyServiceOptions) { if (!isValidLlmTool(options.llmTool)) { throw new Error( `Invalid LLM tool: ${options.llmTool}. Supported tools: ${Object.keys(LlmToolConfig).join(', ')}`, ); } this.llmTool = options.llmTool; this.service = this.createService(options); } /** * Get the current LLM tool identifier */ getLlmTool(): LlmToolId { return this.llmTool; } /** * Invoke the LLM with the given parameters * Delegates to the appropriate underlying service */ async invokeAsLlm(params: LlmInvocationParams): Promise<LlmInvocationResponse> { return this.service.invokeAsLlm(params); } /** * Check if the underlying LLM service is enabled/available */ async isEnabled(): Promise<boolean> { return this.service.isEnabled(); } /** * Create the appropriate service based on the tool type */ private createService(options: LlmProxyServiceOptions): CodingAgentService { const { llmTool, workspaceRoot, defaultTimeout, toolConfig } = options; switch (llmTool) { case LLM_TOOL_CLAUDE_CODE: return new ClaudeCodeService({ workspaceRoot, defaultTimeout, toolConfig, }); case LLM_TOOL_CODEX: return new CodexService({ workspaceRoot, defaultTimeout, toolConfig, }); case LLM_TOOL_GEMINI_CLI: return new GeminiCliService({ workspaceRoot, toolConfig, }); case LLM_TOOL_GITHUB_COPILOT: return new GitHubCopilotService({ workspaceRoot, defaultTimeout, toolConfig, }); default: { // TypeScript exhaustiveness check const _exhaustiveCheck: never = llmTool; throw new Error(`Unsupported LLM tool: ${_exhaustiveCheck}`); } } } }