WuYanBo
/
CodeReview
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

增加了自定义请求头功能； 

修复了API调用的相关功能.
This commit is contained in:
lintsinghua 2025-10-26 15:32:39 +08:00
parent ecfee09ad4
commit ea97d67165
16 changed files with 196 additions and 58 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
README.md
View File

@ -189,11 +189,18 @@ VITE_LLM_GAP_MS=500 # 请求间隔(ms)
```env
VITE_LLM_TIMEOUT=300000 # 增加超时时间
VITE_LLM_BASE_URL=https://your-proxy.com # 使用代理或中转服务
VITE_LLM_BASE_URL=https://your-proxy.com/v1 # 使用代理或中转服务
VITE_LLM_CONCURRENCY=1 # 降低并发数
VITE_LLM_GAP_MS=1000 # 增加请求间隔
```
**自定义请求头示例**（针对特殊中转站）：
```env
# JSON 格式字符串
VITE_LLM_CUSTOM_HEADERS='{"X-API-Version":"v1","X-Custom-Auth":"token123"}'
```
### 常见问题
<details>
@ -581,19 +588,27 @@ pnpm lint
#### 核心LLM配置
| 变量名 | 必需 | 默认值 | 说明 |
|--------|------|--------|------|
| `VITE_LLM_PROVIDER` | ✅ | `gemini` | LLM提供商：`gemini`\|`openai`\|`claude`\|`qwen`\|`deepseek`\|`zhipu`\|`moonshot`\|`baidu`\|`minimax`\|`doubao` |
| `VITE_LLM_PROVIDER` | ✅ | `gemini` | LLM提供商：`gemini`\|`openai`\|`claude`\|`qwen`\|`deepseek`\|`zhipu`\|`moonshot`\|`baidu`\|`minimax`\|`doubao`\|`ollama` |
| `VITE_LLM_API_KEY` | ✅ | - | 通用API Key（优先级高于平台专用配置） |
| `VITE_LLM_MODEL` | ❌ | 自动 | 模型名称（不指定则使用各平台默认模型） |
| `VITE_LLM_BASE_URL` | ❌ | - | 自定义API端点（用于代理、中转或私有部署） |
| `VITE_LLM_BASE_URL` | ❌ | - | 自定义API端点（**支持所有平台的中转站**、代理或私有部署） |
| `VITE_LLM_TIMEOUT` | ❌ | `150000` | 请求超时时间（毫秒） |
| `VITE_LLM_TEMPERATURE` | ❌ | `0.2` | 温度参数（0.0-2.0），控制输出随机性 |
| `VITE_LLM_MAX_TOKENS` | ❌ | `4096` | 最大输出token数 |
| `VITE_LLM_CUSTOM_HEADERS` | ❌ | - | 自定义HTTP请求头（JSON格式字符串），用于特殊中转站或自建服务 |
> 💡 **API 格式支持**：XCodeReviewer 支持三种主流 API 格式：
> - **OpenAI 兼容格式**（最常见）：适用于大多数中转站和 OpenRouter
> - **Gemini 格式**：Google Gemini 官方及兼容服务
> - **Claude 格式**：Anthropic Claude 官方及兼容服务
>
> 配置时只需选择对应的 LLM 提供商，填入中转站地址和 Key 即可。自定义请求头功能可满足特殊中转站的额外要求。
#### 平台专用API Key配置（可选）
| 变量名 | 说明 | 特殊要求 |
|--------|------|---------|
| `VITE_GEMINI_API_KEY` | Google Gemini API Key | - |
| `VITE_GEMINI_MODEL` | Gemini模型 (默认: gemini-2.5-flash) | - |
| `VITE_GEMINI_MODEL` | Gemini模型 (默认: gemini-1.5-flash) | - |
| `VITE_OPENAI_API_KEY` | OpenAI API Key | - |
| `VITE_OPENAI_MODEL` | OpenAI模型 (默认: gpt-4o-mini) | - |
| `VITE_OPENAI_BASE_URL` | OpenAI自定义端点 | 用于中转服务 |

23
README_EN.md
View File

@ -189,11 +189,18 @@ For timeout or connection issues, adjust these parameters:
```env
VITE_LLM_TIMEOUT=300000 # Increase timeout
VITE_LLM_BASE_URL=https://your-proxy.com # Use proxy or relay service
VITE_LLM_BASE_URL=https://your-proxy.com/v1 # Use proxy or relay service
VITE_LLM_CONCURRENCY=1 # Reduce concurrency
VITE_LLM_GAP_MS=1000 # Increase request interval
```
**Custom Headers Example** (for special relay services):
```env
# JSON format string
VITE_LLM_CUSTOM_HEADERS='{"X-API-Version":"v1","X-Custom-Auth":"token123"}'
```
### FAQ
<details>
@ -582,19 +589,27 @@ pnpm lint
#### Core LLM Configuration
| Variable | Required | Default | Description |
|----------|----------|---------|-------------|
| `VITE_LLM_PROVIDER` | ✅ | `gemini` | LLM provider: `gemini`\|`openai`\|`claude`\|`qwen`\|`deepseek`\|`zhipu`\|`moonshot`\|`baidu`\|`minimax`\|`doubao` |
| `VITE_LLM_PROVIDER` | ✅ | `gemini` | LLM provider: `gemini`\|`openai`\|`claude`\|`qwen`\|`deepseek`\|`zhipu`\|`moonshot`\|`baidu`\|`minimax`\|`doubao`\|`ollama` |
| `VITE_LLM_API_KEY` | ✅ | - | Universal API Key (higher priority than platform-specific config) |
| `VITE_LLM_MODEL` | ❌ | Auto | Model name (uses platform default if not specified) |
| `VITE_LLM_BASE_URL` | ❌ | - | Custom API endpoint (for proxy, relay, or private deployment) |
| `VITE_LLM_BASE_URL` | ❌ | - | Custom API endpoint (**supports relay services for all platforms**, proxy, or private deployment) |
| `VITE_LLM_TIMEOUT` | ❌ | `150000` | Request timeout (milliseconds) |
| `VITE_LLM_TEMPERATURE` | ❌ | `0.2` | Temperature parameter (0.0-2.0), controls output randomness |
| `VITE_LLM_MAX_TOKENS` | ❌ | `4096` | Maximum output tokens |
| `VITE_LLM_CUSTOM_HEADERS` | ❌ | - | Custom HTTP headers (JSON string format), for special relay services or self-hosted instances |
> 💡 **API Format Support**: XCodeReviewer supports 3 mainstream API formats:
> - **OpenAI-Compatible Format** (Most Common): Works with most relay services and OpenRouter
> - **Gemini Format**: Google Gemini official and compatible services
> - **Claude Format**: Anthropic Claude official and compatible services
>
> Simply select the corresponding LLM provider, enter the relay service address and Key. The custom headers feature can meet additional requirements of special relay services.
#### Platform-Specific API Key Configuration (Optional)
| Variable | Description | Special Requirements |
|----------|-------------|---------------------|
| `VITE_GEMINI_API_KEY` | Google Gemini API Key | - |
| `VITE_GEMINI_MODEL` | Gemini model (default: gemini-2.5-flash) | - |
| `VITE_GEMINI_MODEL` | Gemini model (default: gemini-1.5-flash) | - |
| `VITE_OPENAI_API_KEY` | OpenAI API Key | - |
| `VITE_OPENAI_MODEL` | OpenAI model (default: gpt-4o-mini) | - |
| `VITE_OPENAI_BASE_URL` | OpenAI custom endpoint | For relay services |

17
src/components/system/SystemConfig.tsx
View File

@ -40,7 +40,7 @@ const LLM_PROVIDERS = [
// 默认模型配置
const DEFAULT_MODELS = {
gemini: 'gemini-2.5-flash',
gemini: 'gemini-1.5-flash',
openai: 'gpt-4o-mini',
claude: 'claude-3-5-sonnet-20241022',
qwen: 'qwen-turbo',
@ -62,6 +62,7 @@ interface SystemConfigData {
llmTimeout: number;
llmTemperature: number;
llmMaxTokens: number;
llmCustomHeaders: string;
// 平台专用配置
geminiApiKey: string;
@ -97,6 +98,7 @@ export function SystemConfig() {
llmTimeout: 150000,
llmTemperature: 0.2,
llmMaxTokens: 4096,
llmCustomHeaders: '',
geminiApiKey: '',
openaiApiKey: '',
claudeApiKey: '',
@ -155,6 +157,7 @@ export function SystemConfig() {
llmTimeout: Number(import.meta.env.VITE_LLM_TIMEOUT) || 150000,
llmTemperature: Number(import.meta.env.VITE_LLM_TEMPERATURE) || 0.2,
llmMaxTokens: Number(import.meta.env.VITE_LLM_MAX_TOKENS) || 4096,
llmCustomHeaders: import.meta.env.VITE_LLM_CUSTOM_HEADERS || '',
geminiApiKey: import.meta.env.VITE_GEMINI_API_KEY || '',
openaiApiKey: import.meta.env.VITE_OPENAI_API_KEY || '',
claudeApiKey: import.meta.env.VITE_CLAUDE_API_KEY || '',
@ -423,6 +426,18 @@ export function SystemConfig() {
/>
</div>
</div>
<div className="space-y-2 pt-4 border-t">
<Label>自定义请求头（高级，可选）</Label>
<Input
value={config.llmCustomHeaders}
onChange={(e) => updateConfig('llmCustomHeaders', e.target.value)}
placeholder='{"X-Custom-Header": "value", "Another-Header": "value2"}'
/>
<p className="text-xs text-muted-foreground">
JSON 格式，用于某些中转站或自建服务的特殊要求。例如：<code className="bg-muted px-1 py-0.5 rounded">&#123;"X-API-Version": "v1"&#125;</code>
</p>
</div>
</CardContent>
</Card>

4
src/shared/config/env.ts
View File

@ -28,10 +28,12 @@ export const env = {
LLM_TEMPERATURE: runtimeConfig?.llmTemperature !== undefined ? runtimeConfig.llmTemperature : (Number(import.meta.env.VITE_LLM_TEMPERATURE) || 0.2),
// LLM 最大token数
LLM_MAX_TOKENS: runtimeConfig?.llmMaxTokens || Number(import.meta.env.VITE_LLM_MAX_TOKENS) || 4096,
// LLM 自定义请求头 (JSON字符串格式)
LLM_CUSTOM_HEADERS: runtimeConfig?.llmCustomHeaders || import.meta.env.VITE_LLM_CUSTOM_HEADERS || '',
// ==================== Gemini AI 配置 (兼容旧配置) ====================
GEMINI_API_KEY: runtimeConfig?.geminiApiKey || import.meta.env.VITE_GEMINI_API_KEY || '',
GEMINI_MODEL: import.meta.env.VITE_GEMINI_MODEL || 'gemini-2.5-flash',
GEMINI_MODEL: import.meta.env.VITE_GEMINI_MODEL || 'gemini-1.5-flash',
GEMINI_TIMEOUT_MS: Number(import.meta.env.VITE_GEMINI_TIMEOUT_MS) || 25000,
// ==================== OpenAI 配置 ====================

16
src/shared/services/llm/adapters/claude-adapter.ts
View File

@ -47,12 +47,20 @@ export class ClaudeAdapter extends BaseLLMAdapter {
requestBody.system = systemMessage.content;
}
// 构建请求头
const headers: Record<string, string> = {
'x-api-key': this.config.apiKey,
'anthropic-version': '2023-06-01',
};
// 合并自定义请求头
if (this.config.customHeaders) {
Object.assign(headers, this.config.customHeaders);
}
const response = await fetch(`${this.baseUrl}/messages`, {
method: 'POST',
headers: this.buildHeaders({
'x-api-key': this.config.apiKey,
'anthropic-version': '2023-06-01',
}),
headers: this.buildHeaders(headers),
body: JSON.stringify(requestBody),
});

14
src/shared/services/llm/adapters/deepseek-adapter.ts
View File

@ -27,11 +27,19 @@ export class DeepSeekAdapter extends BaseLLMAdapter {
private async _sendRequest(request: LLMRequest): Promise<LLMResponse> {
// DeepSeek API兼容OpenAI格式
// 构建请求头
const headers: Record<string, string> = {
'Authorization': `Bearer ${this.config.apiKey}`,
};
// 合并自定义请求头
if (this.config.customHeaders) {
Object.assign(headers, this.config.customHeaders);
}
const response = await fetch(`${this.baseUrl}/v1/chat/completions`, {
method: 'POST',
headers: this.buildHeaders({
'Authorization': `Bearer ${this.config.apiKey}`,
}),
headers: this.buildHeaders(headers),
body: JSON.stringify({
model: this.config.model,
messages: request.messages,

9
src/shared/services/llm/adapters/doubao-adapter.ts
View File

@ -27,11 +27,14 @@ export class DoubaoAdapter extends BaseLLMAdapter {
private async _sendRequest(request: LLMRequest): Promise<LLMResponse> {
// 豆包API兼容OpenAI格式
const headers: Record<string, string> = {
'Authorization': `Bearer ${this.config.apiKey}`,
};
if (this.config.customHeaders) Object.assign(headers, this.config.customHeaders);
const response = await fetch(`${this.baseUrl}/chat/completions`, {
method: 'POST',
headers: this.buildHeaders({
'Authorization': `Bearer ${this.config.apiKey}`,
}),
headers: this.buildHeaders(headers),
body: JSON.stringify({
model: this.config.model,
messages: request.messages,

77
src/shared/services/llm/adapters/gemini-adapter.ts
View File

@ -1,17 +1,17 @@
/**
* Google Gemini适配器
* Google Gemini适配器 - 支持官方API和中转站
*/
import { GoogleGenerativeAI } from '@google/generative-ai';
import { BaseLLMAdapter } from '../base-adapter';
import type { LLMRequest, LLMResponse } from '../types';
export class GeminiAdapter extends BaseLLMAdapter {
private client: GoogleGenerativeAI;
private baseUrl: string;
constructor(config: any) {
super(config);
this.client = new GoogleGenerativeAI(this.config.apiKey);
// 支持自定义baseUrl（中转站）或使用官方API
this.baseUrl = this.config.baseUrl || 'https://generativelanguage.googleapis.com/v1beta';
}
async complete(request: LLMRequest): Promise<LLMResponse> {
@ -27,16 +27,7 @@ export class GeminiAdapter extends BaseLLMAdapter {
}
private async _generateContent(request: LLMRequest): Promise<LLMResponse> {
const model = this.client.getGenerativeModel({
model: this.config.model,
generationConfig: {
temperature: request.temperature ?? this.config.temperature,
maxOutputTokens: request.maxTokens ?? this.config.maxTokens,
topP: request.topP ?? this.config.topP,
}
});
// 将消息转换为Gemini格式
// 转换消息格式为 Gemini 格式
const contents = request.messages
.filter(msg => msg.role !== 'system')
.map(msg => ({
@ -44,24 +35,69 @@ export class GeminiAdapter extends BaseLLMAdapter {
parts: [{ text: msg.content }],
}));
// 系统消息作为第一条用户消息的前缀
// 将系统消息合并到第一条用户消息
const systemMessage = request.messages.find(msg => msg.role === 'system');
if (systemMessage && contents.length > 0) {
contents[0].parts[0].text = `${systemMessage.content}\n\n${contents[0].parts[0].text}`;
}
const result = await model.generateContent({
// 构建请求体
const requestBody = {
contents,
safetySettings: [],
generationConfig: {
temperature: request.temperature ?? this.config.temperature,
maxOutputTokens: request.maxTokens ?? this.config.maxTokens,
topP: request.topP ?? this.config.topP,
}
};
// 构建请求头
const headers: Record<string, string> = {
'Content-Type': 'application/json',
};
// 如果有自定义请求头，合并进去
if (this.config.customHeaders) {
Object.assign(headers, this.config.customHeaders);
}
// API Key 可能在 URL 参数或请求头中
const url = `${this.baseUrl}/models/${this.config.model}:generateContent?key=${this.config.apiKey}`;
const response = await fetch(url, {
method: 'POST',
headers: this.buildHeaders(headers),
body: JSON.stringify(requestBody),
});
const response = result.response;
const text = response.text();
if (!response.ok) {
const error = await response.json().catch(() => ({}));
throw {
statusCode: response.status,
message: error.error?.message || `HTTP ${response.status}: ${response.statusText}`,
details: error,
};
}
const data = await response.json();
// 解析 Gemini 响应格式
const candidate = data.candidates?.[0];
if (!candidate || !candidate.content) {
throw new Error('API响应格式异常: 缺少candidates或content字段');
}
const text = candidate.content.parts?.map((part: any) => part.text).join('') || '';
return {
content: text,
model: this.config.model,
finishReason: 'stop',
usage: data.usageMetadata ? {
promptTokens: data.usageMetadata.promptTokenCount || 0,
completionTokens: data.usageMetadata.candidatesTokenCount || 0,
totalTokens: data.usageMetadata.totalTokenCount || 0,
} : undefined,
finishReason: candidate.finishReason || 'stop',
};
}
@ -75,4 +111,3 @@ export class GeminiAdapter extends BaseLLMAdapter {
return true;
}
}

9
src/shared/services/llm/adapters/minimax-adapter.ts
View File

@ -27,11 +27,14 @@ export class MinimaxAdapter extends BaseLLMAdapter {
private async _sendRequest(request: LLMRequest): Promise<LLMResponse> {
// MiniMax API兼容OpenAI格式
const headers: Record<string, string> = {
'Authorization': `Bearer ${this.config.apiKey}`,
};
if (this.config.customHeaders) Object.assign(headers, this.config.customHeaders);
const response = await fetch(`${this.baseUrl}/text/chatcompletion_v2`, {
method: 'POST',
headers: this.buildHeaders({
'Authorization': `Bearer ${this.config.apiKey}`,
}),
headers: this.buildHeaders(headers),
body: JSON.stringify({
model: this.config.model,
messages: request.messages,

9
src/shared/services/llm/adapters/moonshot-adapter.ts
View File

@ -27,11 +27,14 @@ export class MoonshotAdapter extends BaseLLMAdapter {
private async _sendRequest(request: LLMRequest): Promise<LLMResponse> {
// Moonshot API兼容OpenAI格式
const headers: Record<string, string> = {
'Authorization': `Bearer ${this.config.apiKey}`,
};
if (this.config.customHeaders) Object.assign(headers, this.config.customHeaders);
const response = await fetch(`${this.baseUrl}/chat/completions`, {
method: 'POST',
headers: this.buildHeaders({
'Authorization': `Bearer ${this.config.apiKey}`,
}),
headers: this.buildHeaders(headers),
body: JSON.stringify({
model: this.config.model,
messages: request.messages,

7
src/shared/services/llm/adapters/ollama-adapter.ts
View File

@ -37,9 +37,14 @@ export class OllamaAdapter extends BaseLLMAdapter {
headers['Authorization'] = `Bearer ${this.config.apiKey}`;
}
// 合并自定义请求头
if (this.config.customHeaders) {
Object.assign(headers, this.config.customHeaders);
}
const response = await fetch(`${this.baseUrl}/chat/completions`, {
method: 'POST',
headers,
headers: this.buildHeaders(headers),
body: JSON.stringify({
model: this.config.model,
messages: request.messages,

14
src/shared/services/llm/adapters/openai-adapter.ts
View File

@ -26,11 +26,19 @@ export class OpenAIAdapter extends BaseLLMAdapter {
}
private async _sendRequest(request: LLMRequest): Promise<LLMResponse> {
// 构建请求头
const headers: Record<string, string> = {
'Authorization': `Bearer ${this.config.apiKey}`,
};
// 合并自定义请求头
if (this.config.customHeaders) {
Object.assign(headers, this.config.customHeaders);
}
const response = await fetch(`${this.baseUrl}/chat/completions`, {
method: 'POST',
headers: this.buildHeaders({
'Authorization': `Bearer ${this.config.apiKey}`,
}),
headers: this.buildHeaders(headers),
body: JSON.stringify({
model: this.config.model,
messages: request.messages,

11
src/shared/services/llm/adapters/qwen-adapter.ts
View File

@ -26,12 +26,15 @@ export class QwenAdapter extends BaseLLMAdapter {
}
private async _sendRequest(request: LLMRequest): Promise<LLMResponse> {
const headers: Record<string, string> = {
'Authorization': `Bearer ${this.config.apiKey}`,
'X-DashScope-SSE': 'disable',
};
if (this.config.customHeaders) Object.assign(headers, this.config.customHeaders);
const response = await fetch(`${this.baseUrl}/services/aigc/text-generation/generation`, {
method: 'POST',
headers: this.buildHeaders({
'Authorization': `Bearer ${this.config.apiKey}`,
'X-DashScope-SSE': 'disable',
}),
headers: this.buildHeaders(headers),
body: JSON.stringify({
model: this.config.model,
input: {

9
src/shared/services/llm/adapters/zhipu-adapter.ts
View File

@ -27,11 +27,14 @@ export class ZhipuAdapter extends BaseLLMAdapter {
private async _sendRequest(request: LLMRequest): Promise<LLMResponse> {
// 智谱AI API兼容OpenAI格式
const headers: Record<string, string> = {
'Authorization': `Bearer ${this.config.apiKey}`,
};
if (this.config.customHeaders) Object.assign(headers, this.config.customHeaders);
const response = await fetch(`${this.baseUrl}/chat/completions`, {
method: 'POST',
headers: this.buildHeaders({
'Authorization': `Bearer ${this.config.apiKey}`,
}),
headers: this.buildHeaders(headers),
body: JSON.stringify({
model: this.config.model,
messages: request.messages,

11
src/shared/services/llm/llm-service.ts
View File

@ -80,6 +80,16 @@ export class LLMService {
baseUrl = env.OLLAMA_BASE_URL;
}
// 解析自定义请求头
let customHeaders: Record<string, string> | undefined;
if (env.LLM_CUSTOM_HEADERS) {
try {
customHeaders = JSON.parse(env.LLM_CUSTOM_HEADERS);
} catch (e) {
console.warn('Invalid LLM_CUSTOM_HEADERS format, should be JSON string');
}
}
const config: LLMConfig = {
provider,
apiKey,
@ -88,6 +98,7 @@ export class LLMService {
timeout: env.LLM_TIMEOUT || env.GEMINI_TIMEOUT_MS,
temperature: env.LLM_TEMPERATURE,
maxTokens: env.LLM_MAX_TOKENS,
customHeaders,
};
return new LLMService(config);

1
src/shared/services/llm/types.ts
View File

@ -28,6 +28,7 @@ export interface LLMConfig {
topP?: number; // Top-p采样
frequencyPenalty?: number; // 频率惩罚
presencePenalty?: number; // 存在惩罚
customHeaders?: Record<string, string>; // 自定义请求头
}
// LLM请求消息