【根因(main.log 实证)】 19:04 / 19:05 / 19:06 三次会话终止均为同一报错: DeepSeek 400 "Messages with role 'tool' must be a response to a preceding message with 'tool_calls'" 缺陷链:engine 主循环仅在 step.thought 存在(该轮有文本或思考内容)时才 将 assistant 消息加入请求历史。当模型发起纯工具调用(零文本零思考 — DeepSeek 高频行为)时: - assistant(tool_calls) 消息不进 messages - 但 tool 结果消息照常 push → 下一轮请求出现孤立 tool 消息 → 协议 400(不可重试)→ 会话 ERROR 终止 "不稳定" = 模型每轮是否附带文本是概率性行为:带文本正常,纯调用必崩。 DB 持久化侧同源缺陷(if (!step.thought) continue)导致这些步骤的 assistant 与 tool 结果全部不落库 — 重启后工具上下文丢失,模型重复调用。 【修复】 - engine.ts: 有 toolCalls 的轮次必 push assistant(content=null,C-6 规范) - agent.ts: 持久化条件同步修复(无 thought 但有 toolCalls 的步骤落库) - 回归测试: 纯 tool_calls 轮后第二次请求中 tool 消息前必须是带 tool_calls 的 assistant(请求契约断言,engine-toolchain.test.ts) 【纵深防御 — 孤立 tool 消息过滤】 - openai-format.ts(DeepSeek/Agnes/MiMo/OpenAI 四家共享): 构建请求时 按 tool_call_id 配对过滤孤立 tool 消息(任何来源的历史污染不再 400 死锁) - anthropic.adapter.ts: tool_use/tool_result 同策略配对过滤 - 单测 ×6: 正常配对保留 / 孤立丢弃 / id 不匹配丢弃 / 多轮配对 / includeImages 原位转换 / 非 vision 静默丢弃 【多模态索引对齐收敛】 4 家 adapter 的 images 处理循环原按未过滤的 nonSystemMsgs[i-1] 对齐索引, 孤立 tool 过滤引入后会错位 — 统一收进 buildOpenAICompatibleMessages (includeImages 参数,基于 sanitized 序列原位转换),4 家 adapter 删除 各自的索引对齐循环(DeepSeek vision 判断 / OpenAI 推理模型拒绝保留在 adapter)。 【终止原因可见化】 MAX_ITERATIONS / TIMEOUT 终止此前无任何提示(用户感知"会话直接停止")— 前端 DONE 事件非 completed 终止原因显示为 system 消息。 【v0.6.1 回归缓解】 web_fetch timeoutMs 120s → 240s:浏览器回退串行化后并发 3 个排队最坏 ~127.5s,旧值让排队末位抓取被工具超时杀掉(表现为抓取不稳定)。 【验证】 lint 0/0;typecheck 双工程 0 错误;test:electron 259/259(+7); electron-vite build 成功
201 lines
7.2 KiB
TypeScript
201 lines
7.2 KiB
TypeScript
/**
|
||
* Agnes AI Provider Adapter
|
||
*
|
||
* OpenAI 兼容 API。支持 Tool Calling、Thinking 模式、多模态(图片 — URL + Base64)。
|
||
*
|
||
* 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 type { MetonaModelInfo } from '../types/metona-adapter';
|
||
import { buildOpenAICompatibleMessages, buildOpenAICompatibleTools } from './shared/openai-format';
|
||
import { parseSSEStream, parseOpenAICompatibleResponse } from './shared/sse-stream';
|
||
import log from 'electron-log';
|
||
|
||
export class AgnesAdapter extends BaseAdapter {
|
||
// H-2 修复: provider → providerId(规范要求)
|
||
override readonly providerId: string = 'agnes';
|
||
readonly supportedModels = ['agnes-2.0-flash'];
|
||
readonly supportsToolCalling = true;
|
||
readonly supportsThinking = true;
|
||
|
||
// H-2 修复: Agnes 模型元信息(1M 上下文,65.5K 最大输出)
|
||
private static readonly MODEL_INFO: Record<string, MetonaModelInfo> = {
|
||
'agnes-2.0-flash': {
|
||
id: 'agnes-2.0-flash',
|
||
name: 'Agnes 2.0 Flash',
|
||
contextWindow: 1_000_000,
|
||
maxOutputTokens: 65_536,
|
||
supportsToolCalling: true,
|
||
supportsThinking: true,
|
||
description: 'Agnes AI 快速版,1M 上下文,支持多模态图片(URL + Base64)与思考模式',
|
||
},
|
||
};
|
||
|
||
// ===== POST /chat/completions (非流式) =====
|
||
|
||
// H-2 修复: chat → send(规范要求)
|
||
async send(request: MetonaRequest): Promise<MetonaResponse> {
|
||
const body = this.toNativeRequest(request, false);
|
||
|
||
// #24 修复: 使用 fetchWithTimeout 替代 getFetchSignal + fetch,确保 timer 清理
|
||
const response = await this.fetchWithTimeout(
|
||
`${this.config.baseURL}/chat/completions`,
|
||
{
|
||
method: 'POST',
|
||
headers: {
|
||
'Content-Type': 'application/json',
|
||
Authorization: `Bearer ${this.config.apiKey}`,
|
||
...this.config.headers,
|
||
},
|
||
body: JSON.stringify(body),
|
||
},
|
||
this.config.timeoutMs ?? 300_000,
|
||
);
|
||
|
||
if (!response.ok) {
|
||
await this.throwHttpError(response, 'Agnes AI API error');
|
||
}
|
||
|
||
const data = (await response.json()) as Record<string, unknown>;
|
||
const parsed = parseOpenAICompatibleResponse(data);
|
||
|
||
return {
|
||
meta: {
|
||
requestId: request.meta.requestId,
|
||
provider: this.providerId,
|
||
model: (data.model as string) ?? this.config.defaultModel,
|
||
latencyMs: 0,
|
||
timestamp: Date.now(),
|
||
},
|
||
content: parsed.content,
|
||
reasoningContent: parsed.reasoningContent,
|
||
toolCalls: parsed.toolCalls,
|
||
usage: parsed.usage,
|
||
finishReason: parsed.finishReason as MetonaFinishReason,
|
||
};
|
||
}
|
||
|
||
// ===== POST /chat/completions (流式) =====
|
||
|
||
// H-2 修复: chatStream → sendStream(规范要求)
|
||
async *sendStream(request: MetonaRequest): AsyncIterable<MetonaStreamEvent> {
|
||
const body = this.toNativeRequest(request, true);
|
||
|
||
// #24 修复: 使用 fetchWithTimeout 替代 getFetchSignal + fetch,确保 timer 清理
|
||
const response = await this.fetchWithTimeout(
|
||
`${this.config.baseURL}/chat/completions`,
|
||
{
|
||
method: 'POST',
|
||
headers: {
|
||
'Content-Type': 'application/json',
|
||
Authorization: `Bearer ${this.config.apiKey}`,
|
||
...this.config.headers,
|
||
},
|
||
body: JSON.stringify(body),
|
||
},
|
||
this.config.timeoutMs ?? 300_000,
|
||
);
|
||
|
||
if (!response.ok || !response.body) {
|
||
await this.throwHttpError(response, 'Agnes AI stream error');
|
||
}
|
||
|
||
yield* parseSSEStream(
|
||
// 非空断言:上方 if 已确保 response.body 不为 null
|
||
response.body!,
|
||
request.meta.requestId,
|
||
request.meta.sessionId,
|
||
request.meta.iteration,
|
||
);
|
||
}
|
||
|
||
/**
|
||
* H-2 修复: 获取上下文窗口大小(规范要求)
|
||
*
|
||
* v0.3.1: 优先使用配置注入的 contextWindow,回退到 MODEL_INFO 默认值。
|
||
* Agnes OpenAI 兼容 API 不支持 context_window 参数,此值仅用于
|
||
* Engine 压缩判断和前端 UI 显示。
|
||
* 注意:Agnes API 未提供 /models 端点,listModels 使用基类默认实现。
|
||
*/
|
||
override getContextWindow(): number {
|
||
// v0.3.1: 优先使用配置注入的 contextWindow
|
||
if (typeof this.config.contextWindow === 'number' && this.config.contextWindow > 0) {
|
||
return this.config.contextWindow;
|
||
}
|
||
// 回退到 MODEL_INFO
|
||
const modelInfo = AgnesAdapter.MODEL_INFO[this.config.defaultModel];
|
||
return modelInfo?.contextWindow ?? 1_000_000;
|
||
}
|
||
|
||
// ========== 私有方法 ==========
|
||
|
||
/**
|
||
* 构建 Agnes AI 原生请求体
|
||
*
|
||
* Agnes AI 特有参数:
|
||
* - 多模态图片:user 消息的 images[] → OpenAI content 数组 [{type:"text"}, {type:"image_url"}]
|
||
* 支持 HTTPS URL 或 base64 Data URI(与 MiMo 一致)
|
||
* - chat_template_kwargs: { enable_thinking: true } — 启用思考模式(非 thinking 字段)
|
||
* - 默认 max_tokens: 65536(1M 上下文,65.5K 最大输出)
|
||
*/
|
||
private toNativeRequest(request: MetonaRequest, stream: boolean): Record<string, unknown> {
|
||
// v0.6.2: images 处理收敛至共享层(原索引对齐循环在孤立 tool 过滤后会错位)
|
||
const messages = buildOpenAICompatibleMessages(request, true);
|
||
const tools = buildOpenAICompatibleTools(request.tools);
|
||
|
||
const imageCount = request.messages.reduce((n, m) => n + (m.images?.length ?? 0), 0);
|
||
if (imageCount > 0) {
|
||
const firstUrl = request.messages.find((m) => m.images?.length)?.images?.[0]?.url ?? '';
|
||
log.info(
|
||
`[Agnes] Processing ${imageCount} image(s), first URL prefix: ${firstUrl.slice(0, 50)}`,
|
||
);
|
||
}
|
||
|
||
// v0.5.3: max_tokens 按模型上限钳制(agnes-2.0-flash 上限 65536)
|
||
const modelInfo = AgnesAdapter.MODEL_INFO[this.config.defaultModel];
|
||
const maxOutput = modelInfo?.maxOutputTokens ?? 65_536;
|
||
const maxTokens = Math.min(request.params.maxTokens ?? maxOutput, maxOutput);
|
||
|
||
const body: Record<string, unknown> = {
|
||
model: this.config.defaultModel,
|
||
messages,
|
||
temperature: request.params.temperature,
|
||
max_tokens: maxTokens,
|
||
stream,
|
||
};
|
||
|
||
if (stream) {
|
||
body.stream_options = { include_usage: true };
|
||
}
|
||
|
||
if (tools) {
|
||
body.tools = tools;
|
||
}
|
||
|
||
// C-3 修复: Thinking 模式 — Agnes 使用 chat_template_kwargs 而非 thinking
|
||
// Agnes API 仅支持 enable_thinking: true/false,不支持 effort 级别
|
||
// thinkingEffort === 'low' 时映射为 false(不启用深度思考),其他级别映射为 true
|
||
if (request.params.thinkingEnabled) {
|
||
const effort = request.params.thinkingEffort ?? 'high';
|
||
body.chat_template_kwargs = { enable_thinking: effort !== 'low' };
|
||
}
|
||
|
||
// 停止序列
|
||
if (request.params.stopSequences?.length) {
|
||
body.stop = request.params.stopSequences;
|
||
}
|
||
|
||
return body;
|
||
}
|
||
}
|