mirror of
https://github.com/Wx-2025/ST-Amily2-Chat-Optimisation.git
synced 2026-06-06 22:05:50 +00:00
275 lines
8.4 KiB
JavaScript
275 lines
8.4 KiB
JavaScript
// 所有Google专用域名后缀 - 检测是否需要激活适配器
|
||
const GOOGLE_DOMAINS = [
|
||
"generativelanguage.googleapis.com",
|
||
"ai.google.dev",
|
||
"us-central1-aiplatform.googleapis.com" // 添加Vertex AI专用域名
|
||
];
|
||
|
||
// 工具函数：检测URL是否是Google API端点
|
||
export function isGoogleEndpoint(url) {
|
||
try {
|
||
if (!url || typeof url !== 'string') return false;
|
||
const parsedUrl = new URL(url);
|
||
const hostname = parsedUrl.hostname.toLowerCase();
|
||
const pathname = parsedUrl.pathname.toLowerCase();
|
||
|
||
// 如果是OpenAI兼容层，则不视为Google原生API
|
||
if (pathname.includes('/openai')) {
|
||
return false;
|
||
}
|
||
|
||
return GOOGLE_DOMAINS.some(domain => hostname.includes(domain));
|
||
} catch (error) {
|
||
console.error('[GoogleAdapter] URL解析错误:', url, error);
|
||
return false;
|
||
}
|
||
}
|
||
|
||
// 转换OpenAI格式请求为Google API所需格式
|
||
export function convertToGoogleRequest(openaiRequest) {
|
||
const { model: _, ...rest } = openaiRequest; // 忽略model参数(Google放在URL中)
|
||
|
||
// 将messages数组转换为Google的内容结构
|
||
const contents = openaiRequest.messages.map(message => {
|
||
let role;
|
||
// Google API只支持 'user' 和 'model' 角色
|
||
if (message.role === 'assistant') {
|
||
role = 'model';
|
||
} else {
|
||
// 'system' 和 'user' 都转换为 'user'
|
||
role = 'user';
|
||
}
|
||
|
||
return {
|
||
role: role,
|
||
parts: [{ text: message.content }]
|
||
};
|
||
});
|
||
|
||
return {
|
||
contents,
|
||
generationConfig: {
|
||
maxOutputTokens: openaiRequest.max_tokens,
|
||
temperature: openaiRequest.temperature || 0.7,
|
||
topP: 0.95,
|
||
},
|
||
safetySettings: [
|
||
{
|
||
category: "HARM_CATEGORY_HARASSMENT",
|
||
threshold: "BLOCK_NONE"
|
||
},
|
||
{
|
||
category: "HARM_CATEGORY_HATE_SPEECH",
|
||
threshold: "BLOCK_NONE"
|
||
},
|
||
{
|
||
category: "HARM_CATEGORY_SEXUALLY_EXPLICIT",
|
||
threshold: "BLOCK_NONE"
|
||
},
|
||
{
|
||
category: "HARM_CATEGORY_DANGEROUS_CONTENT",
|
||
threshold: "BLOCK_NONE"
|
||
}
|
||
]
|
||
};
|
||
}
|
||
|
||
// 将Google API响应转换为标准OpenAI格式
|
||
export function parseGoogleResponse(googleResponse) {
|
||
try {
|
||
// 处理Google的错误响应格式
|
||
if (googleResponse.error) {
|
||
throw new Error(
|
||
`Google API错误: ${googleResponse.error.message || '未知错误'}\n代码: ${googleResponse.error.code}`
|
||
);
|
||
}
|
||
|
||
// 处理候选数据
|
||
const candidate = googleResponse.candidates?.[0];
|
||
if (!candidate || !candidate.content) {
|
||
throw new Error('无效的Google API响应: 未找到候选内容');
|
||
}
|
||
|
||
// 合并多部分内容
|
||
const content = candidate.content.parts
|
||
.map(part => part.text || '')
|
||
.join('\n')
|
||
.trim();
|
||
|
||
// 转换为OpenAI兼容格式
|
||
return {
|
||
choices: [
|
||
{
|
||
index: 0,
|
||
message: {
|
||
role: 'assistant',
|
||
content
|
||
},
|
||
finish_reason: candidate.finishReason || 'stop'
|
||
}
|
||
]
|
||
};
|
||
} catch (error) {
|
||
console.error('[GoogleAdapter] 响应解析错误:', error);
|
||
console.log('原始Google响应:', googleResponse);
|
||
throw error;
|
||
}
|
||
}
|
||
|
||
// 构建GoogleAPI的完整端点URL
|
||
export function buildGoogleApiUrl(baseUrl, modelName) {
|
||
try {
|
||
const url = new URL(baseUrl);
|
||
const pathname = url.pathname.replace(/\/$/, ''); // 移除末尾的斜杠
|
||
|
||
// 检查是否为v1或v1beta的基础路径
|
||
if (pathname.endsWith('/v1beta') || pathname.endsWith('/v1')) {
|
||
if (!modelName) {
|
||
throw new Error('Google API需要模型名称');
|
||
}
|
||
|
||
const apiVersion = pathname.endsWith('/v1beta') ? 'v1beta' : 'v1';
|
||
|
||
// 构建完整的模型调用URL
|
||
url.pathname = `/${apiVersion}/models/${modelName}:generateContent`;
|
||
return url.href;
|
||
}
|
||
|
||
// 如果已经是完整API路径或自定义路径，则直接返回
|
||
return url.href;
|
||
} catch (error) {
|
||
console.error('[GoogleAdapter] URL构建错误:', baseUrl, modelName, error);
|
||
throw new Error(`无效的API地址: ${baseUrl}`);
|
||
}
|
||
}
|
||
|
||
export function buildGoogleEmbeddingRequest(texts, modelName) {
|
||
const requests = texts.map(text => ({
|
||
model: `models/${modelName}`,
|
||
content: {
|
||
parts: [{ text: text }]
|
||
}
|
||
}));
|
||
|
||
return { requests };
|
||
}
|
||
|
||
export function parseGoogleEmbeddingResponse(googleResponse, originalTexts) {
|
||
if (!googleResponse || !Array.isArray(googleResponse.embeddings)) {
|
||
console.error('[GoogleAdapter] Google向量API响应格式无效，缺少 "embeddings" 数组:', googleResponse);
|
||
throw new Error('Google API返回的向量数据格式不正确。');
|
||
}
|
||
|
||
// Google API 返回的字段名是 "values" 而不是 "value"
|
||
return googleResponse.embeddings.map(emb => emb.values);
|
||
}
|
||
|
||
export function buildGoogleEmbeddingApiUrl(baseUrl, modelName) {
|
||
const url = new URL(baseUrl);
|
||
const apiVersion = modelName.includes('gemini') ? 'v1beta' : 'v1';
|
||
url.pathname = `/${apiVersion}/models/${modelName}:batchEmbedContents`;
|
||
return url.href;
|
||
}
|
||
|
||
function convertOaiToGoogleForPlotOptimization(messages) {
|
||
const contents = [];
|
||
let system_instruction = null;
|
||
let lastRole = '';
|
||
|
||
for (const message of messages) {
|
||
if (message.role === 'user' && lastRole === 'user') {
|
||
const lastContent = contents[contents.length - 1];
|
||
lastContent.parts.push({ text: `\n\n${message.content}` });
|
||
continue;
|
||
}
|
||
|
||
if (message.role === 'assistant') {
|
||
contents.push({
|
||
role: 'model',
|
||
parts: [{ text: message.content }],
|
||
});
|
||
lastRole = 'model';
|
||
} else {
|
||
contents.push({
|
||
role: 'user',
|
||
parts: [{ text: message.content }],
|
||
});
|
||
lastRole = 'user';
|
||
}
|
||
}
|
||
|
||
return { contents, system_instruction };
|
||
}
|
||
|
||
export function buildPlotOptimizationGoogleRequest(messages, apiSettings) {
|
||
const { contents } = convertOaiToGoogleForPlotOptimization(messages);
|
||
|
||
const generationConfig = {
|
||
temperature: apiSettings.temperature,
|
||
topP: apiSettings.top_p,
|
||
topK: apiSettings.top_k,
|
||
maxOutputTokens: apiSettings.max_tokens,
|
||
};
|
||
|
||
if(generationConfig.topK) generationConfig.topK = Math.round(generationConfig.topK);
|
||
|
||
Object.keys(generationConfig).forEach(key => {
|
||
if (generationConfig[key] === undefined || generationConfig[key] === null) {
|
||
delete generationConfig[key];
|
||
}
|
||
});
|
||
|
||
if (contents.length === 0) {
|
||
contents.push({ role: 'user', parts: [{ text: 'Hi' }] });
|
||
}
|
||
|
||
const safetySettings = [
|
||
{ category: 'HARM_CATEGORY_HARASSMENT', threshold: 'BLOCK_NONE' },
|
||
{ category: 'HARM_CATEGORY_HATE_SPEECH', threshold: 'BLOCK_NONE' },
|
||
{ category: 'HARM_CATEGORY_SEXUALLY_EXPLICIT', threshold: 'BLOCK_NONE' },
|
||
{ category: 'HARM_CATEGORY_DANGEROUS_CONTENT', threshold: 'BLOCK_NONE' },
|
||
];
|
||
|
||
const payload = {
|
||
contents,
|
||
generationConfig,
|
||
safetySettings,
|
||
};
|
||
|
||
return payload;
|
||
}
|
||
|
||
export function parsePlotOptimizationGoogleResponse(googleResponse) {
|
||
try {
|
||
const candidates = googleResponse?.candidates;
|
||
if (!candidates || candidates.length === 0) {
|
||
let message = `Google API returned no candidates.`;
|
||
if (googleResponse?.promptFeedback?.blockReason) {
|
||
message += `\nPrompt was blocked due to: ${googleResponse.promptFeedback.blockReason}`;
|
||
console.error(message, googleResponse.promptFeedback.safetyRatings);
|
||
}
|
||
return { choices: [{ message: { content: `Error: ${message}` } }] };
|
||
}
|
||
|
||
const responseContent = candidates[0].content;
|
||
const responseText = responseContent?.parts?.map(part => part.text).join('') || '';
|
||
|
||
if (!responseText) {
|
||
let message = `Google API response text is empty.`;
|
||
console.warn(message, googleResponse);
|
||
return { choices: [{ message: { content: 'Error: Received an empty response from the API.' } }] };
|
||
}
|
||
|
||
return {
|
||
choices: [{
|
||
message: {
|
||
content: responseText
|
||
}
|
||
}]
|
||
};
|
||
} catch (error) {
|
||
console.error(`Error parsing Google response:`, error, googleResponse);
|
||
return { choices: [{ message: { content: `Error: Failed to parse Google API response. Details: ${error.message}` } }] };
|
||
}
|
||
}
|