metona-ai-desktop/electron/harness/adapters/agnes-ai.adapter.ts

/**
 * Agnes AI Provider Adapter
 *
 * OpenAI 兼容 API。支持 Tool Calling、Thinking 模式、多模态（图片）。
 *
 * 独立继承 BaseAdapter，通过 shared/openai-format 和 shared/sse-stream 复用
 * OpenAI 兼容格式构建和 SSE 流式解析逻辑。不与其他 Provider Adapter 耦合。
 *
 * 与 DeepSeek 的差异：
 * - Thinking 模式使用 chat_template_kwargs（非 thinking 字段）
 * - 默认 max_tokens 更大（65536 vs 8192）
 *
 * @see apis/agnes-ai-api-docs-20260625.html
 */

import { BaseAdapter } from './base-adapter';
import type { MetonaRequest, MetonaResponse, MetonaStreamEvent } from '../types';
import { MetonaFinishReason } from '../types';
import { buildOpenAICompatibleMessages, buildOpenAICompatibleTools } from './shared/openai-format';
import { parseSSEStream, parseOpenAICompatibleResponse } from './shared/sse-stream';

export class AgnesAdapter extends BaseAdapter {
  override readonly provider: string = 'agnes';
  readonly supportedModels = ['agnes-2.0-flash'];
  readonly supportsToolCalling = true;
  readonly supportsThinking = true;

  // ===== POST /chat/completions (非流式) =====

  async chat(request: MetonaRequest): Promise<MetonaResponse> {
    const body = this.toNativeRequest(request, false);

    const response = await fetch(`${this.config.baseURL}/chat/completions`, {
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
        Authorization: `Bearer ${this.config.apiKey}`,
        ...this.config.headers,
      },
      body: JSON.stringify(body),
      signal: AbortSignal.timeout(this.config.timeoutMs ?? 300_000),
    });

    if (!response.ok) {
      const errorBody = await response.text().catch(() => '');
      throw new Error(`Agnes AI API error: ${response.status} ${response.statusText} - ${errorBody}`);
    }

    const data = await response.json() as Record<string, unknown>;
    const parsed = parseOpenAICompatibleResponse(data, request.meta.requestId, this.provider, this.config.defaultModel);

    return {
      meta: {
        requestId: request.meta.requestId,
        provider: this.provider,
        model: (data.model as string) ?? this.config.defaultModel,
        latencyMs: 0,
        timestamp: Date.now(),
      },
      content: parsed.content,
      reasoningContent: parsed.reasoningContent,
      toolCalls: parsed.toolCalls,
      usage: parsed.usage,
      finishReason: parsed.finishReason as MetonaFinishReason,
    };
  }

  // ===== POST /chat/completions (流式) =====

  async *chatStream(request: MetonaRequest): AsyncIterable<MetonaStreamEvent> {
    const body = this.toNativeRequest(request, true);

    const response = await fetch(`${this.config.baseURL}/chat/completions`, {
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
        Authorization: `Bearer ${this.config.apiKey}`,
        ...this.config.headers,
      },
      body: JSON.stringify(body),
    });

    if (!response.ok || !response.body) {
      throw new Error(`Agnes AI stream error: ${response.status}`);
    }

    yield* parseSSEStream(
      response.body,
      request.meta.requestId,
      request.meta.sessionId,
      request.meta.iteration,
    );
  }

  // ========== 私有方法 ==========

  /**
   * 构建 Agnes AI 原生请求体
   *
   * Agnes AI 特有参数：
   * - chat_template_kwargs: { enable_thinking: true } — 启用思考模式（非 thinking 字段）
   * - 默认 max_tokens: 65536（512K 上下文）
   */
  private toNativeRequest(request: MetonaRequest, stream: boolean): Record<string, unknown> {
    const messages = buildOpenAICompatibleMessages(request);
    const tools = buildOpenAICompatibleTools(request.tools);

    const body: Record<string, unknown> = {
      model: this.config.defaultModel,
      messages,
      temperature: request.params.temperature ?? 0,
      max_tokens: request.params.maxTokens ?? 65536,
      stream,
    };

    if (stream) {
      body.stream_options = { include_usage: true };
    }

    if (tools) {
      body.tools = tools;
    }

    // Thinking 模式：Agnes 使用 chat_template_kwargs 而非 thinking
    if (request.params.thinkingEnabled) {
      body.chat_template_kwargs = { enable_thinking: true };
    }

    // 停止序列
    if (request.params.stopSequences?.length) {
      body.stop = request.params.stopSequences;
    }

    return body;
  }
}