diff --git a/README.md b/README.md index 18210ed26..a25becfe3 100644 --- a/README.md +++ b/README.md @@ -199,6 +199,25 @@ Google Gemini Pro Api Key. Google Gemini Pro Api Url. +### `AZURE_URL` （可选） + +> 形如：https://{azure-resource-url}/openai/deployments +> +> ⚠️ 注意：这里与原项目配置不同，不需要指定 {deploy-name}，将模型名修改为 {deploy-name} 即可切换不同的模型 +> +> ⚠️ DALL-E 等需要 openai 密钥的插件暂不支持 Azure + +Azure 部署地址。 + +### `AZURE_API_KEY` （可选） + +Azure 密钥。 + +### `AZURE_API_VERSION` （可选） + +Azure Api 版本，你可以在这里找到：[Azure 文档](https://learn.microsoft.com/en-us/azure/ai-services/openai/reference#chat-completions)。 + + ## 部署 ### 容器部署 （推荐） diff --git a/README_CN.md b/README_CN.md index 422802a2b..570dd07a9 100644 --- a/README_CN.md +++ b/README_CN.md @@ -94,7 +94,7 @@ OpenAI 接口代理 URL，如果你手动配置了 openai 接口代理，请填 ### `AZURE_URL` （可选） -> 形如：https://{azure-resource-url}/openai/deployments/{deploy-name} +> 形如：https://{azure-resource-url}/openai/deployments Azure 部署地址。 diff --git a/app/api/common.ts b/app/api/common.ts index 9df835d17..c3bf87ec3 100644 --- a/app/api/common.ts +++ b/app/api/common.ts @@ -60,7 +60,6 @@ export async function requestOpenai(req: NextRequest) { path = makeAzurePath(path, serverConfig.azureApiVersion); } - const fetchUrl = `${baseUrl}/${path}`; const fetchOptions: RequestInit = { headers: { "Content-Type": "application/json", @@ -78,6 +77,12 @@ export async function requestOpenai(req: NextRequest) { duplex: "half", signal: controller.signal, }; + const clonedBody = await req.text(); + const jsonBody = JSON.parse(clonedBody) as { model?: string }; + if (serverConfig.isAzure) { + baseUrl = `${baseUrl}/${jsonBody.model}`; + } + const fetchUrl = `${baseUrl}/${path}`; // #1815 try to refuse gpt4 request if (serverConfig.customModels && req.body) { @@ -86,11 +91,10 @@ export async function requestOpenai(req: NextRequest) { DEFAULT_MODELS, serverConfig.customModels, ); - const clonedBody = await req.text(); + // const clonedBody = await req.text(); + // const jsonBody = JSON.parse(clonedBody) as { model?: string }; fetchOptions.body = clonedBody; - const jsonBody = JSON.parse(clonedBody) as { model?: string }; - // not undefined and is false if (modelTable[jsonBody?.model ?? ""].available === false) { return NextResponse.json( diff --git a/app/api/langchain/tool/agent/agentapi.ts b/app/api/langchain/tool/agent/agentapi.ts index e4e590f2e..d12009838 100644 --- a/app/api/langchain/tool/agent/agentapi.ts +++ b/app/api/langchain/tool/agent/agentapi.ts @@ -8,7 +8,7 @@ import { BaseCallbackHandler } from "langchain/callbacks"; import { AIMessage, HumanMessage, SystemMessage } from "langchain/schema"; import { BufferMemory, ChatMessageHistory } from "langchain/memory"; import { initializeAgentExecutorWithOptions } from "langchain/agents"; -import { ACCESS_CODE_PREFIX } from "@/app/constant"; +import { ACCESS_CODE_PREFIX, ServiceProvider } from "@/app/constant"; import * as langchainTools from "langchain/tools"; import { HttpGetTool } from "@/app/api/langchain-tools/http_get"; @@ -16,6 +16,7 @@ import { DuckDuckGo } from "@/app/api/langchain-tools/duckduckgo_search"; import { DynamicTool, Tool } from "langchain/tools"; import { BaiduSearch } from "@/app/api/langchain-tools/baidu_search"; import { GoogleSearch } from "@/app/api/langchain-tools/google_search"; +import { useAccessStore } from "@/app/store"; export interface RequestMessage { role: string; @@ -24,6 +25,8 @@ export interface RequestMessage { export interface RequestBody { messages: RequestMessage[]; + isAzure: boolean; + azureApiVersion?: string; model: string; stream?: boolean; temperature: number; @@ -152,10 +155,10 @@ export class AgentApi { async getOpenAIApiKey(token: string) { const serverConfig = getServerSideConfig(); - const isOpenAiKey = !token.startsWith(ACCESS_CODE_PREFIX); + const isApiKey = !token.startsWith(ACCESS_CODE_PREFIX); let apiKey = serverConfig.apiKey; - if (isOpenAiKey && token) { + if (isApiKey && token) { apiKey = token; } return apiKey; @@ -179,27 +182,31 @@ export class AgentApi { customTools: any[], ) { try { + let useTools = reqBody.useTools ?? []; const serverConfig = getServerSideConfig(); // const reqBody: RequestBody = await req.json(); - const authToken = req.headers.get("Authorization") ?? ""; + const isAzure = reqBody.isAzure || serverConfig.isAzure; + const authHeaderName = isAzure ? "api-key" : "Authorization"; + const authToken = req.headers.get(authHeaderName) ?? ""; const token = authToken.trim().replaceAll("Bearer ", "").trim(); - const isOpenAiKey = !token.startsWith(ACCESS_CODE_PREFIX); - let useTools = reqBody.useTools ?? []; - let apiKey = serverConfig.apiKey; - if (isOpenAiKey && token) { - apiKey = token; - } + let apiKey = await this.getOpenAIApiKey(token); + if (isAzure) apiKey = token; let baseUrl = "https://api.openai.com/v1"; if (serverConfig.baseUrl) baseUrl = serverConfig.baseUrl; if ( reqBody.baseUrl?.startsWith("http://") || reqBody.baseUrl?.startsWith("https://") - ) + ) { baseUrl = reqBody.baseUrl; - if (!baseUrl.endsWith("/v1")) + } + if (!isAzure && !baseUrl.endsWith("/v1")) { baseUrl = baseUrl.endsWith("/") ? `${baseUrl}v1` : `${baseUrl}/v1`; + } + if (!reqBody.isAzure && serverConfig.isAzure) { + baseUrl = serverConfig.azureUrl || baseUrl; + } console.log("[baseUrl]", baseUrl); var handler = await this.getHandler(reqBody); @@ -281,7 +288,7 @@ export class AgentApi { chatHistory: new ChatMessageHistory(pastMessages), }); - const llm = new ChatOpenAI( + let llm = new ChatOpenAI( { modelName: reqBody.model, openAIApiKey: apiKey, @@ -293,6 +300,23 @@ export class AgentApi { }, { basePath: baseUrl }, ); + + if (reqBody.isAzure || serverConfig.isAzure) { + llm = new ChatOpenAI({ + temperature: reqBody.temperature, + streaming: reqBody.stream, + topP: reqBody.top_p, + presencePenalty: reqBody.presence_penalty, + frequencyPenalty: reqBody.frequency_penalty, + azureOpenAIApiKey: apiKey, + azureOpenAIApiVersion: reqBody.isAzure + ? reqBody.azureApiVersion + : serverConfig.azureApiVersion, + azureOpenAIApiDeploymentName: reqBody.model, + azureOpenAIBasePath: baseUrl, + }); + } + const executor = await initializeAgentExecutorWithOptions(tools, llm, { agentType: "openai-functions", returnIntermediateSteps: reqBody.returnIntermediateSteps, diff --git a/app/client/platforms/openai.ts b/app/client/platforms/openai.ts index ea038b6a7..157cec90a 100644 --- a/app/client/platforms/openai.ts +++ b/app/client/platforms/openai.ts @@ -38,7 +38,7 @@ export interface OpenAIListModelResponse { export class ChatGPTApi implements LLMApi { private disableListModels = true; - path(path: string): string { + path(path: string, model?: string): string { const accessStore = useAccessStore.getState(); const isAzure = accessStore.provider === ServiceProvider.Azure; @@ -65,6 +65,7 @@ export class ChatGPTApi implements LLMApi { if (isAzure) { path = makeAzurePath(path, accessStore.azureApiVersion); + return [baseUrl, model, path].join("/"); } return [baseUrl, path].join("/"); @@ -136,7 +137,7 @@ export class ChatGPTApi implements LLMApi { options.onController?.(controller); try { - const chatPath = this.path(OpenaiPath.ChatPath); + const chatPath = this.path(OpenaiPath.ChatPath, modelConfig.model); const chatPayload = { method: "POST", body: JSON.stringify(requestPayload), @@ -284,16 +285,20 @@ export class ChatGPTApi implements LLMApi { model: options.config.model, }, }; - + const accessStore = useAccessStore.getState(); + const isAzure = accessStore.provider === ServiceProvider.Azure; + let baseUrl = isAzure ? accessStore.azureUrl : accessStore.openaiUrl; const requestPayload = { messages, + isAzure, + azureApiVersion: accessStore.azureApiVersion, stream: options.config.stream, model: modelConfig.model, temperature: modelConfig.temperature, presence_penalty: modelConfig.presence_penalty, frequency_penalty: modelConfig.frequency_penalty, top_p: modelConfig.top_p, - baseUrl: useAccessStore.getState().openaiUrl, + baseUrl: baseUrl, maxIterations: options.agentConfig.maxIterations, returnIntermediateSteps: options.agentConfig.returnIntermediateSteps, useTools: options.agentConfig.useTools, @@ -321,7 +326,7 @@ export class ChatGPTApi implements LLMApi { () => controller.abort(), REQUEST_TIMEOUT_MS, ); - console.log("shouldStream", shouldStream); + // console.log("shouldStream", shouldStream); if (shouldStream) { let responseText = ""; diff --git a/app/config/server.ts b/app/config/server.ts index 373c6d582..3bc5e5c67 100644 --- a/app/config/server.ts +++ b/app/config/server.ts @@ -23,7 +23,7 @@ declare global { CUSTOM_MODELS?: string; // to control custom models // azure only - AZURE_URL?: string; // https://{azure-url}/openai/deployments/{deploy-name} + AZURE_URL?: string; // https://{azure-url}/openai/deployments AZURE_API_KEY?: string; AZURE_API_VERSION?: string; diff --git a/app/constant.ts b/app/constant.ts index 92184d1df..ba21e4ef9 100644 --- a/app/constant.ts +++ b/app/constant.ts @@ -88,13 +88,8 @@ export const OpenaiPath = { ListModelPath: "v1/models", }; -export const GooglePath = { - ChatPath: "v1/models/{{model}}:streamGenerateContent", - ListModelPath: "v1/models", -}; - export const Azure = { - ExampleEndpoint: "https://{resource-url}/openai/deployments/{deploy-id}", + ExampleEndpoint: "https://{resource-url}/openai/deployments", }; export const Google = {