101 lines
3.9 KiB
JavaScript
101 lines
3.9 KiB
JavaScript
// Gemini 原生 API & OpenAI 兼容 API — 文本向量化 (Text Embeddings)
|
|
|
|
export const runtime = 'edge';
|
|
|
|
export async function POST(request) {
|
|
try {
|
|
const { text, apiConfig } = await request.json();
|
|
|
|
const isCustomEmbed = apiConfig?.useCustomEmbed;
|
|
const provider = isCustomEmbed ? apiConfig.embedProvider : (apiConfig?.provider || 'zhipu');
|
|
const apiKey = isCustomEmbed ? (apiConfig.embedApiKey || apiConfig?.apiKey) : apiConfig?.apiKey;
|
|
|
|
// 自动识别默认填写或遗留的智谱URL并矫正为对应官方URL
|
|
let defaultBaseUrl = provider === 'gemini-native' ? 'https://generativelanguage.googleapis.com/v1beta' : 'https://open.bigmodel.cn/api/paas/v4';
|
|
|
|
let rawBaseUrl;
|
|
if (isCustomEmbed) {
|
|
rawBaseUrl = apiConfig.embedBaseUrl;
|
|
} else {
|
|
// 如果是自定义提供商且没开独立Embed，默认继承对聊的baseUrl
|
|
rawBaseUrl = apiConfig?.baseUrl || defaultBaseUrl;
|
|
}
|
|
|
|
if (!rawBaseUrl || (provider === 'gemini-native' && rawBaseUrl.includes('open.bigmodel.cn'))) {
|
|
rawBaseUrl = defaultBaseUrl;
|
|
}
|
|
|
|
const baseUrl = rawBaseUrl.replace(/\/$/, '');
|
|
|
|
let embedModelName;
|
|
if (isCustomEmbed) {
|
|
embedModelName = apiConfig.embedModel || 'embedding-3';
|
|
} else if (provider === 'custom') {
|
|
// 如果没开独立embed，但选了custom，默认用text-embedding-v3-small或用户主模型
|
|
embedModelName = 'text-embedding-v3-small';
|
|
} else {
|
|
embedModelName = provider === 'zhipu' ? 'embedding-3' : 'text-embedding-v3-small';
|
|
}
|
|
|
|
if (!apiKey) {
|
|
return new Response(JSON.stringify({ error: isCustomEmbed ? '请在API配置中填写独立的 Embedding API Key' : '请先配置 API Key' }), { status: 400 });
|
|
}
|
|
|
|
if (!text || typeof text !== 'string') {
|
|
return new Response(JSON.stringify({ error: '无效的文本输入' }), { status: 400 });
|
|
}
|
|
|
|
let embeddings = [];
|
|
|
|
if (provider === 'gemini-native') {
|
|
const geminiModel = embedModelName || 'text-embedding-004';
|
|
const url = `${baseUrl}/models/${geminiModel}:embedContent?key=${apiKey}`;
|
|
console.log('Fetching Gemini Embeddings:', url);
|
|
const res = await fetch(url, {
|
|
method: 'POST',
|
|
headers: { 'Content-Type': 'application/json' },
|
|
body: JSON.stringify({
|
|
model: `models/${geminiModel}`,
|
|
content: { parts: [{ text }] }
|
|
})
|
|
});
|
|
|
|
if (!res.ok) {
|
|
const errText = await res.text();
|
|
throw new Error(`Gemini Embedding Error: ${errText}`);
|
|
}
|
|
const data = await res.json();
|
|
embeddings = data.embedding.values;
|
|
} else {
|
|
// OpenAI 兼容格式
|
|
const url = `${baseUrl}/embeddings`;
|
|
|
|
const res = await fetch(url, {
|
|
method: 'POST',
|
|
headers: {
|
|
'Content-Type': 'application/json',
|
|
'Authorization': `Bearer ${apiKey}`
|
|
},
|
|
body: JSON.stringify({
|
|
input: text,
|
|
model: embedModelName
|
|
})
|
|
});
|
|
|
|
if (!res.ok) {
|
|
const errText = await res.text();
|
|
throw new Error(`Embedding API Error: ${errText}`);
|
|
}
|
|
const data = await res.json();
|
|
embeddings = data.data[0].embedding;
|
|
}
|
|
|
|
return new Response(JSON.stringify({ embedding: embeddings }), {
|
|
headers: { 'Content-Type': 'application/json' }
|
|
});
|
|
} catch (err) {
|
|
console.error('Embedding API Error:', err);
|
|
return new Response(JSON.stringify({ error: err.message }), { status: 500 });
|
|
}
|
|
}
|