From 089654871cf7b7d877f6bc0ccc48da45d4231035 Mon Sep 17 00:00:00 2001 From: "jinhui.li" Date: Mon, 31 Mar 2025 22:28:02 +0800 Subject: [PATCH] support plugins --- README.md | 53 +++---- src/deepseek.ts => plugins/deepseek.js | 97 ++++++------ plugins/gemini.js | 23 +++ src/constants.ts | 15 +- src/index.ts | 27 +++- src/middlewares/formatRequest.ts | 101 ++++++++++++ src/middlewares/rewriteBody.ts | 43 ++++++ src/middlewares/rewriteToolsPrompt.ts | 34 ---- src/router copy.ts | 206 ------------------------- src/server.ts | 136 ---------------- src/utils/index.ts | 47 +++++- 11 files changed, 304 insertions(+), 478 deletions(-) rename src/deepseek.ts => plugins/deepseek.js (62%) create mode 100644 plugins/gemini.js create mode 100644 src/middlewares/formatRequest.ts create mode 100644 src/middlewares/rewriteBody.ts delete mode 100644 src/middlewares/rewriteToolsPrompt.ts delete mode 100644 src/router copy.ts diff --git a/README.md b/README.md index bdce7fa..a17cf44 100644 --- a/README.md +++ b/README.md @@ -4,23 +4,11 @@ ![demo.png](https://github.com/musistudio/claude-code-router/blob/main/screenshoots/demo.png) -## Warning! This project is for testing purposes and may consume a lot of tokens! It may also fail to complete tasks! - ## Implemented -- [x] Normal Mode and Router Mode +- [x] Support writing custom plugins for rewriting prompts. -- [x] Using the qwen2.5-coder-3b model as the routing dispatcher (since it’s currently free on Alibaba Cloud’s official website) - -- [x] Using the qwen-max-0125 model as the tool invoker - -- [x] Using deepseek-v3 as the coder model - -- [x] Using deepseek-r1 as the reasoning model - -- [x] Support proxy - -Thanks to the free qwen2.5-coder-3b model from Alibaba and deepseek’s KV-Cache, we can significantly reduce the cost of using Claude Code. Make sure to set appropriate ignorePatterns for the project. See: https://docs.anthropic.com/en/docs/agents-and-tools/claude-code/overview +- [x] Support writing custom plugins for implementing routers. ## Usage @@ -30,16 +18,18 @@ Thanks to the free qwen2.5-coder-3b model from Alibaba and deepseek’s KV-Cache npm install -g @anthropic-ai/claude-code ``` -1. Install claude-code-router +1. Clone this repo and install dependencies ```shell -npm install -g @musistudio/claude-code-router +git clone https://github.com/musistudio/claude-code-router +cd claude-code-router && pnpm i +npm run build ``` 2. Start claude-code-router server ```shell -claude-code-router +node dist/cli.js ``` 3. Set environment variable to start claude code @@ -51,21 +41,20 @@ export API_TIMEOUT_MS=600000 claude ``` -## Normal Mode +## Plugin -The initial version uses a single model to accomplish all tasks. This model needs to support function calling and must allow for a sufficiently large tool description length, ideally greater than 1754. If the model used in this mode does not support KV Cache, it will consume a significant number of tokens. +The plugin allows users to rewrite Claude Code prompt and custom router. The plugin path is in `$HOME/.claude-code-router/plugins`. Currently, there are two demos available: +1. [custom router](https://github.com/musistudio/claude-code-router/blob/dev/custom-prompt/plugins/deepseek.js) +2. [rewrite prompt](https://github.com/musistudio/claude-code-router/blob/dev/custom-prompt/plugins/gemini.js) -![normal mode](https://github.com/musistudio/claude-code-reverse/blob/main/screenshoots/normal.png) +You need to move them to the `$HOME/.claude-code-router/plugins` directory and configure 'usePlugin' in `$HOME/.claude-code-router/config.json`，like this: -## Router Mode - -Using multiple models to handle different tasks, this mode requires setting ENABLE_ROUTER to true and configuring four models: ROUTER_AGENT_MODEL, TOOL_AGENT_MODEL, CODER_AGENT_MODEL, and THINK_AGENT_MODEL. - -ROUTER_AGENT_MODEL does not require high intelligence and is only responsible for request routing. A small model is sufficient for this task (testing has shown that the qwen-coder-3b model performs well). -TOOL_AGENT_MODEL must support function calling and allow for a sufficiently large tool description length, ideally greater than 1754. If the model used in this mode does not support KV Cache, it will consume a significant number of tokens. - -CODER_AGENT_MODEL and THINK_AGENT_MODEL can use the DeepSeek series of models. - -The purpose of router mode is to separate tool invocation from coding tasks, enabling the use of inference models like r1, which do not support function calling. - -![router mode](https://github.com/musistudio/claude-code-router/blob/main/screenshoots/router.png) +```json +{ + "usePlugin": "gemini", + "LOG": true, + "OPENAI_API_KEY": "", + "OPENAI_BASE_URL": "", + "OPENAI_MODEL": "" +} +``` diff --git a/src/deepseek.ts b/plugins/deepseek.js similarity index 62% rename from src/deepseek.ts rename to plugins/deepseek.js index afc0c9d..7de7038 100644 --- a/src/deepseek.ts +++ b/plugins/deepseek.js @@ -1,15 +1,10 @@ -import { OpenAI } from "openai"; -import { createClient } from "./utils"; -import { log } from "./utils/log"; -export interface BaseRouter { - name: string; - description: string; - run: ( - args: OpenAI.Chat.Completions.ChatCompletionCreateParams - ) => Promise; -} +const { + log, + streamOpenAIResponse, + createClient, +} = require("claude-code-router"); -const thinkRouter: BaseRouter = { +const thinkRouter = { name: "think", description: `This agent is used solely for complex reasoning and thinking tasks. It should not be called for information retrieval or repetitive, frequent requests. Only use this agent for tasks that require deep analysis or problem-solving. If there is an existing result from the Thinker agent, do not call this agent again.你只负责深度思考以拆分任务，不需要进行任何的编码和调用工具。最后讲拆分的步骤按照顺序返回。比如\n1. xxx\n2. xxx\n3. xxx`, run(args) { @@ -18,13 +13,13 @@ const thinkRouter: BaseRouter = { baseURL: process.env.THINK_AGENT_BASE_URL, }); const messages = JSON.parse(JSON.stringify(args.messages)); - messages.forEach((msg: any) => { + messages.forEach((msg) => { if (Array.isArray(msg.content)) { msg.content = JSON.stringify(msg.content); } }); - let startIdx = messages.findIndex((msg: any) => msg.role !== "system"); + let startIdx = messages.findIndex((msg) => msg.role !== "system"); if (startIdx === -1) startIdx = messages.length; for (let i = startIdx; i < messages.length; i++) { @@ -46,14 +41,12 @@ const thinkRouter: BaseRouter = { return client.chat.completions.create({ ...args, messages, - model: process.env.THINK_AGENT_MODEL as string, + model: process.env.THINK_AGENT_MODEL, }); }, }; -export class Router { - routers: BaseRouter[]; - client: OpenAI; +class Router { constructor() { this.routers = [thinkRouter]; this.client = createClient({ @@ -61,37 +54,37 @@ export class Router { baseURL: process.env.ROUTER_AGENT_BASE_URL, }); } - async route( - args: OpenAI.Chat.Completions.ChatCompletionCreateParams - ): Promise { + async route(args) { log(`Request Router: ${JSON.stringify(args, null, 2)}`); - const res: OpenAI.Chat.Completions.ChatCompletion = - await this.client.chat.completions.create({ - ...args, - messages: [ - ...args.messages, - { - role: "system", - content: `## **Guidelines:** -- **Trigger the "think" mode when the user's request involves deep thinking, complex reasoning, or multi-step analysis.** -- **Criteria:** - - Involves multi-layered logical reasoning or causal analysis - - Requires establishing connections or pattern recognition between different pieces of information - - Involves cross-domain knowledge integration or weighing multiple possibilities - - Requires creative thinking or non-direct inference -### **Format requirements:** -- When you need to trigger the "think" mode, return the following JSON format: -\`\`\`json -{ - "use": "think" -} -\`\`\` -`, - }, - ], - model: process.env.ROUTER_AGENT_MODEL as string, - stream: false, - }); + const res = await this.client.chat.completions.create({ + ...args, + messages: [ + ...args.messages, + { + role: "system", + content: `## **Guidelines:** + - **Trigger the "think" mode when the user's request involves deep thinking, complex reasoning, or multi-step analysis.** + - **Criteria:** + - Involves multi-layered logical reasoning or causal analysis + - Requires establishing connections or pattern recognition between different pieces of information + - Involves cross-domain knowledge integration or weighing multiple possibilities + - Requires creative thinking or non-direct inference + ### **Special Case:** + - **When the user sends "test", respond with "success" only.** + + ### **Format requirements:** + - When you need to trigger the "think" mode, return the following JSON format: + \`\`\`json + { + "use": "think" + } + \`\`\` + `, + }, + ], + model: process.env.ROUTER_AGENT_MODEL, + stream: false, + }); let result; try { const text = res.choices[0].message.content; @@ -102,13 +95,13 @@ export class Router { text.slice(text.indexOf("{"), text.lastIndexOf("}") + 1) ); } catch (e) { - (res.choices[0] as any).delta = res.choices[0].message; + res.choices[0].delta = res.choices[0].message; log(`No Router: ${JSON.stringify(res.choices[0].message)}`); return [res]; } const router = this.routers.find((item) => item.name === result.use); if (!router) { - (res.choices[0] as any).delta = res.choices[0].message; + res.choices[0].delta = res.choices[0].message; log(`No Router: ${JSON.stringify(res.choices[0].message)}`); return [res]; } @@ -138,3 +131,9 @@ export class Router { return router.run(args); } } + +const router = new Router(); +module.exports = async function handle(req, res, next) { + const completions = await router.route(req.body); + streamOpenAIResponse(res, completions, req.body.model); +}; diff --git a/plugins/gemini.js b/plugins/gemini.js new file mode 100644 index 0000000..a5458a1 --- /dev/null +++ b/plugins/gemini.js @@ -0,0 +1,23 @@ +module.exports = async function handle(req, res, next) { + if (Array.isArray(req.body.tools)) { + // rewrite tools definition + req.body.tools.forEach((tool) => { + if (tool.function.name === "BatchTool") { + // HACK: Gemini does not support objects with empty properties + tool.function.parameters.properties.invocations.items.properties.input.type = + "number"; + return; + } + Object.keys(tool.function.parameters.properties).forEach((key) => { + const prop = tool.function.parameters.properties[key]; + if ( + prop.type === "string" && + !["enum", "date-time"].includes(prop.format) + ) { + delete prop.format; + } + }); + }); + } + next(); +}; diff --git a/src/constants.ts b/src/constants.ts index 9255079..644bf6f 100644 --- a/src/constants.ts +++ b/src/constants.ts @@ -5,24 +5,11 @@ export const HOME_DIR = path.join(os.homedir(), ".claude-code-router"); export const CONFIG_FILE = `${HOME_DIR}/config.json`; -export const PROMPTS_DIR = `${HOME_DIR}/prompts`; +export const PLUGINS_DIR = `${HOME_DIR}/plugins`; export const DEFAULT_CONFIG = { log: false, - ENABLE_ROUTER: true, OPENAI_API_KEY: "", OPENAI_BASE_URL: "https://openrouter.ai/api/v1", OPENAI_MODEL: "openai/o3-mini", - - CODER_AGENT_API_KEY: "", - CODER_AGENT_BASE_URL: "https://api.deepseek.com", - CODER_AGENT_MODEL: "deepseek-chat", - - THINK_AGENT_API_KEY: "", - THINK_AGENT_BASE_URL: "https://api.deepseek.com", - THINK_AGENT_MODEL: "deepseek-reasoner", - - ROUTER_AGENT_API_KEY: "", - ROUTER_AGENT_BASE_URL: "https://api.deepseek.com", - ROUTER_AGENT_MODEL: "deepseek-chat", }; diff --git a/src/index.ts b/src/index.ts index 08c5ca8..6fd70c9 100644 --- a/src/index.ts +++ b/src/index.ts @@ -1,8 +1,11 @@ import { existsSync } from "fs"; import { writeFile } from "fs/promises"; -import { initConfig, initDir } from "./utils"; +import { getOpenAICommonOptions, initConfig, initDir } from "./utils"; import { createServer } from "./server"; -import { rewriteToolsPrompt } from "./middlewares/rewriteToolsPrompt"; +import { formatRequest } from "./middlewares/formatRequest"; +import { rewriteBody } from "./middlewares/rewriteBody"; +import OpenAI from "openai"; +import { streamOpenAIResponse } from "./utils/stream"; async function initializeClaudeConfig() { const homeDir = process.env.HOME; @@ -29,7 +32,25 @@ async function run() { await initDir(); await initConfig(); const server = createServer(3456); - server.useMiddleware(rewriteToolsPrompt); + server.useMiddleware(formatRequest); + server.useMiddleware(rewriteBody); + + const openai = new OpenAI({ + apiKey: process.env.OPENAI_API_KEY, + baseURL: process.env.OPENAI_BASE_URL, + ...getOpenAICommonOptions(), + }); + server.app.post("/v1/messages", async (req, res) => { + try { + if (process.env.OPENAI_MODEL) { + req.body.model = process.env.OPENAI_MODEL; + } + const completion: any = await openai.chat.completions.create(req.body); + await streamOpenAIResponse(res, completion, req.body.model); + } catch (e) { + console.error("Error in OpenAI API call:", e); + } + }); server.start(); } run(); diff --git a/src/middlewares/formatRequest.ts b/src/middlewares/formatRequest.ts new file mode 100644 index 0000000..e5ab184 --- /dev/null +++ b/src/middlewares/formatRequest.ts @@ -0,0 +1,101 @@ +import { Request, Response, NextFunction } from "express"; +import { ContentBlockParam } from "@anthropic-ai/sdk/resources"; +import { MessageCreateParamsBase } from "@anthropic-ai/sdk/resources/messages"; +import OpenAI from "openai"; +import { streamOpenAIResponse } from "../utils/stream"; + +export const formatRequest = async ( + req: Request, + res: Response, + next: NextFunction +) => { + let { + model, + max_tokens, + messages, + system = [], + temperature, + metadata, + tools, + }: MessageCreateParamsBase = req.body; + try { + const openAIMessages: OpenAI.Chat.Completions.ChatCompletionMessageParam[] = + messages.map((item) => { + if (item.content instanceof Array) { + return { + role: item.role, + content: item.content + .map((it: ContentBlockParam) => { + if (it.type === "text") { + return typeof it.text === "string" + ? it.text + : JSON.stringify(it); + } + return JSON.stringify(it); + }) + .join(""), + } as OpenAI.Chat.Completions.ChatCompletionMessageParam; + } + return { + role: item.role, + content: + typeof item.content === "string" + ? item.content + : JSON.stringify(item.content), + }; + }); + const systemMessages: OpenAI.Chat.Completions.ChatCompletionMessageParam[] = + Array.isArray(system) + ? system.map((item) => ({ + role: "system", + content: item.text, + })) + : [{ role: "system", content: system }]; + const data: OpenAI.Chat.Completions.ChatCompletionCreateParams = { + model, + messages: [...systemMessages, ...openAIMessages], + temperature, + stream: true, + }; + if (tools) { + data.tools = tools + .filter((tool) => !["StickerRequest"].includes(tool.name)) + .map((item: any) => ({ + type: "function", + function: { + name: item.name, + description: item.description, + parameters: item.input_schema, + }, + })); + } + res.setHeader("Content-Type", "text/event-stream"); + res.setHeader("Cache-Control", "no-cache"); + res.setHeader("Connection", "keep-alive"); + req.body = data; + } catch (error) { + console.error("Error in request processing:", error); + const errorCompletion: AsyncIterable = + { + async *[Symbol.asyncIterator]() { + yield { + id: `error_${Date.now()}`, + created: Math.floor(Date.now() / 1000), + model: "gpt-3.5-turbo", + object: "chat.completion.chunk", + choices: [ + { + index: 0, + delta: { + content: `Error: ${(error as Error).message}`, + }, + finish_reason: "stop", + }, + ], + }; + }, + }; + await streamOpenAIResponse(res, errorCompletion, model); + } + next(); +}; diff --git a/src/middlewares/rewriteBody.ts b/src/middlewares/rewriteBody.ts new file mode 100644 index 0000000..e15d92b --- /dev/null +++ b/src/middlewares/rewriteBody.ts @@ -0,0 +1,43 @@ +import { Request, Response, NextFunction } from "express"; +import Module from "node:module"; +import { streamOpenAIResponse } from "../utils/stream"; +import { log } from "../utils/log"; +import { PLUGINS_DIR } from "../constants"; +import path from "node:path"; +import { access } from "node:fs/promises"; +import { OpenAI } from "openai"; +import { createClient } from "../utils"; + +// @ts-ignore +const originalLoad = Module._load; +// @ts-ignore +Module._load = function (request, parent, isMain) { + if (request === "claude-code-router") { + return { + streamOpenAIResponse, + log, + OpenAI, + createClient, + }; + } + return originalLoad.call(this, request, parent, isMain); +}; + +export const rewriteBody = async ( + req: Request, + res: Response, + next: NextFunction +) => { + if (!process.env.usePlugin) { + return next(); + } + const pluginPath = path.join(PLUGINS_DIR, `${process.env.usePlugin}.js`); + try { + await access(pluginPath); + const rewritePlugin = require(pluginPath); + rewritePlugin(req, res, next); + } catch (e) { + console.error(e); + next(); + } +}; diff --git a/src/middlewares/rewriteToolsPrompt.ts b/src/middlewares/rewriteToolsPrompt.ts deleted file mode 100644 index 699e52d..0000000 --- a/src/middlewares/rewriteToolsPrompt.ts +++ /dev/null @@ -1,34 +0,0 @@ -import { Request, Response, NextFunction } from "express"; -import { readFile, access } from "node:fs/promises"; -import { join } from "node:path"; -import { PROMPTS_DIR } from "../constants"; - -const getPrompt = async (name: string) => { - try { - const promptPath = join(PROMPTS_DIR, `${name}.md`); - await access(promptPath); - const prompt = await readFile(promptPath, "utf-8"); - return prompt; - } catch { - return null; - } -}; - -export const rewriteToolsPrompt = async ( - req: Request, - res: Response, - next: NextFunction -) => { - const { tools } = req.body; - if (!Array.isArray(tools)) { - next(); - return; - } - for (const tool of tools) { - const prompt = await getPrompt(tool.name); - if (prompt) { - tool.description = prompt; - } - } - next(); -}; diff --git a/src/router copy.ts b/src/router copy.ts deleted file mode 100644 index 9516ada..0000000 --- a/src/router copy.ts +++ /dev/null @@ -1,206 +0,0 @@ -import { OpenAI } from "openai"; -import { createClient } from "./utils"; -import { log } from "./utils/log"; -export interface BaseRouter { - name: string; - description: string; - run: ( - args: OpenAI.Chat.Completions.ChatCompletionCreateParams - ) => Promise; -} - -const coderRouter: BaseRouter = { - name: "coder", - description: `This agent is solely responsible for helping users write code. This agent could not call tools. This agent is used for writing and modifying code when the user provides clear and specific coding requirements. For example, tasks like implementing a quicksort algorithm in JavaScript or creating an HTML layout. If the user's request is unclear or cannot be directly translated into code, please route the task to 'think' first for clarification or further processing.`, - run(args) { - const client = createClient({ - apiKey: process.env.CODER_AGENT_API_KEY, - baseURL: process.env.CODER_AGENT_BASE_URL, - }); - delete args.tools; - args.messages.forEach((item) => { - if (Array.isArray(item.content)) { - item.content = JSON.stringify(item.content); - } - }); - return client.chat.completions.create({ - ...args, - messages: [ - ...args.messages, - { - role: "system", - content: - "You are a code writer who helps users write code based on their specific requirements. You create algorithms, implement functionality, and build structures according to the clear instructions provided by the user. Your focus is solely on writing code, ensuring that the task is completed accurately and efficiently.", - }, - ], - model: process.env.CODER_AGENT_MODEL as string, - }); - }, -}; - - -const useToolRouter: BaseRouter = { - name: "use-tool", - description: `This agent can call user-specified tools to perform tasks. The user provides a list of tools to be used, and the agent integrates these tools to complete the specified tasks efficiently. The agent follows user instructions and ensures proper tool utilization for each request`, - run(args) { - const client = createClient({ - apiKey: process.env.TOOL_AGENT_API_KEY, - baseURL: process.env.TOOL_AGENT_BASE_URL, - }); - return client.chat.completions.create({ - ...args, - messages: [ - ...args.messages, - { - role: "system", - content: - "You need to select the appropriate tool for the task based on the user’s request. Review the requirements and choose the tool that fits the task best.", - }, - ], - model: process.env.TOOL_AGENT_MODEL as string, - }); - }, -}; - -const thinkRouter: BaseRouter = { - name: "think", - description: `This agent is used solely for complex reasoning and thinking tasks. It should not be called for information retrieval or repetitive, frequent requests. Only use this agent for tasks that require deep analysis or problem-solving. If there is an existing result from the Thinker agent, do not call this agent again.你只负责深度思考以拆分任务，不需要进行任何的编码和调用工具。最后讲拆分的步骤按照顺序返回。比如\n1. xxx\n2. xxx\n3. xxx`, - run(args) { - const client = createClient({ - apiKey: process.env.THINK_AGENT_API_KEY, - baseURL: process.env.THINK_AGENT_BASE_URL, - }); - const messages = JSON.parse(JSON.stringify(args.messages)); - messages.forEach((msg: any) => { - if (Array.isArray(msg.content)) { - msg.content = JSON.stringify(msg.content); - } - }); - - let startIdx = messages.findIndex((msg: any) => msg.role !== "system"); - if (startIdx === -1) startIdx = messages.length; - - for (let i = startIdx; i < messages.length; i++) { - const expectedRole = (i - startIdx) % 2 === 0 ? "user" : "assistant"; - messages[i].role = expectedRole; - } - - if ( - messages.length > 0 && - messages[messages.length - 1].role === "assistant" - ) { - messages.push({ - role: "user", - content: - "Please follow the instructions provided above to resolve the issue.", - }); - } - delete args.tools; - return client.chat.completions.create({ - ...args, - messages, - model: process.env.THINK_AGENT_MODEL as string, - }); - }, -}; - -export class Router { - routers: BaseRouter[]; - client: OpenAI; - constructor() { - this.routers = [coderRouter, useToolRouter, thinkRouter]; - this.client = createClient({ - apiKey: process.env.ROUTER_AGENT_API_KEY, - baseURL: process.env.ROUTER_AGENT_BASE_URL, - }); - } - async route( - args: OpenAI.Chat.Completions.ChatCompletionCreateParams - ): Promise { - log(`Route: ${JSON.stringify(args, null, 2)}`); - const res: OpenAI.Chat.Completions.ChatCompletion = - await this.client.chat.completions.create({ - ...args, - messages: [ - ...args.messages, - { - role: "system", - content: `You are an AI task router and executor, responsible for understanding user requests and directing them to the appropriate processing mode or tool based on the task type and requirements. Your main responsibility is to determine the nature of the request, execute the task when possible, and respond appropriately. - -### **Guidelines:** -- **If an external tool is required to complete the task (such as searching for information, generating images, or modifying code), route the task to \`use-tool\` rather than handling it directly.** -- If the task requires generating an image, route to \`use-tool\` and specify the image generation tool. -- If the task requires searching for information, route to \`use-tool\` and specify the search tool. -- If the task requires modifying or executing code, route to \`use-tool\` and specify the code handling tool. -- **Do NOT execute the tool action directly; always trigger it through \`use-tool\`.** - -- **If the user is chatting casually or having a general conversation, respond naturally and conversationally. Improving the user experience through friendly interactions is one of your main responsibilities.** - -- **If the user's request involves deep thinking, complex reasoning, or multi-step analysis, use the "think" mode to break down and solve the problem.** - -- **If the user's request involves coding or technical implementation, use the "coder" mode to generate or modify code.** - - **After generating the code, if the task requires applying or integrating the code, route to \`use-tool\` and specify the code execution tool.** - - **Do NOT re-trigger "coder" to apply code — route to \`use-tool\` instead.** - -### **Format requirements:** -- When you need to trigger a specific mode (such as "think", "coder", or "use-tool"), return the following JSON format: - -### IMPORTANT: -- 你不能也不会调用BatchTool，如果你需要使用工具请路由到\`use-tool\`，由\`use-tool\`来调用BatchTool。 - -\`\`\`json -{ - "use": "", -} -\`\`\` -`, - }, - ], - model: process.env.ROUTER_AGENT_MODEL as string, - stream: false, - }); - let result; - try { - const text = res.choices[0].message.content; - if (!text) { - throw new Error("No text"); - } - result = JSON.parse( - text.slice(text.indexOf("{"), text.lastIndexOf("}") + 1) - ); - } catch (e) { - (res.choices[0] as any).delta = res.choices[0].message; - return [res]; - } - const router = this.routers.find((item) => item.name === result.use); - if (!router) { - (res.choices[0] as any).delta = res.choices[0].message; - log(`No Router: ${JSON.stringify(res.choices[0].message)}`); - return [res]; - } - log(`Use Router: ${router.name}`); - if (router.name === "think" || router.name === "coder") { - const agentResult = await router.run({ - ...args, - stream: false, - }); - try { - args.messages.push({ - role: "user", - content: - `${router.name} Agent Result: ` + - agentResult.choices[0].message.content, - }); - log( - `${router.name} Agent Result: ` + - agentResult.choices[0].message.content - ); - return await this.route(args); - } catch (error) { - console.log(agentResult); - throw error; - } - } - return router.run(args); - } -} diff --git a/src/server.ts b/src/server.ts index 328ebf8..3fd2468 100644 --- a/src/server.ts +++ b/src/server.ts @@ -1,18 +1,4 @@ import express, { RequestHandler } from "express"; -import { - ContentBlockParam, - MessageCreateParamsBase, -} from "@anthropic-ai/sdk/resources/messages"; -import { OpenAI } from "openai"; -import { Router } from "./deepseek"; -import { getOpenAICommonOptions } from "./utils"; -import { streamOpenAIResponse } from "./utils/stream"; - -interface Client { - call: ( - data: OpenAI.Chat.Completions.ChatCompletionCreateParams - ) => Promise; -} interface Server { app: express.Application; @@ -23,128 +9,6 @@ interface Server { export const createServer = (port: number): Server => { const app = express(); app.use(express.json({ limit: "500mb" })); - - let client: Client; - if (process.env.ENABLE_ROUTER && process.env.ENABLE_ROUTER === "true") { - const router = new Router(); - client = { - call: (data) => { - return router.route(data); - }, - }; - } else { - const openai = new OpenAI({ - apiKey: process.env.OPENAI_API_KEY, - baseURL: process.env.OPENAI_BASE_URL, - ...getOpenAICommonOptions(), - }); - client = { - call: (data) => { - if (process.env.OPENAI_MODEL) { - data.model = process.env.OPENAI_MODEL; - } - return openai.chat.completions.create(data); - }, - }; - } - - app.post("/v1/messages", async (req, res) => { - try { - let { - model, - max_tokens, - messages, - system = [], - temperature, - metadata, - tools, - }: MessageCreateParamsBase = req.body; - - const openAIMessages: OpenAI.Chat.Completions.ChatCompletionMessageParam[] = - messages.map((item) => { - if (item.content instanceof Array) { - return { - role: item.role, - content: item.content - .map((it: ContentBlockParam) => { - if (it.type === "text") { - return typeof it.text === "string" - ? it.text - : JSON.stringify(it); - } - return JSON.stringify(it); - }) - .join(""), - } as OpenAI.Chat.Completions.ChatCompletionMessageParam; - } - return { - role: item.role, - content: - typeof item.content === "string" - ? item.content - : JSON.stringify(item.content), - }; - }); - const systemMessages: OpenAI.Chat.Completions.ChatCompletionMessageParam[] = - Array.isArray(system) - ? system.map((item) => ({ - role: "system", - content: item.text, - })) - : [{ role: "system", content: system }]; - const data: OpenAI.Chat.Completions.ChatCompletionCreateParams = { - model, - messages: [...systemMessages, ...openAIMessages], - temperature, - stream: true, - }; - if (tools) { - data.tools = tools - .filter((tool) => !["StickerRequest"].includes(tool.name)) - .map((item: any) => ({ - type: "function", - function: { - name: item.name, - description: item.description, - parameters: item.input_schema, - }, - })); - } - res.setHeader("Content-Type", "text/event-stream"); - res.setHeader("Cache-Control", "no-cache"); - res.setHeader("Connection", "keep-alive"); - try { - const completion = await client.call(data); - await streamOpenAIResponse(res, completion, model); - } catch (e) { - console.error("Error in OpenAI API call:", e); - } - } catch (error) { - console.error("Error in request processing:", error); - const errorCompletion: AsyncIterable = - { - async *[Symbol.asyncIterator]() { - yield { - id: `error_${Date.now()}`, - created: Math.floor(Date.now() / 1000), - model: "gpt-3.5-turbo", - object: "chat.completion.chunk", - choices: [ - { - index: 0, - delta: { - content: `Error: ${(error as Error).message}`, - }, - finish_reason: "stop", - }, - ], - }; - }, - }; - await streamOpenAIResponse(res, errorCompletion, "gpt-3.5-turbo"); - } - }); - return { app, useMiddleware: (middleware: RequestHandler) => { diff --git a/src/utils/index.ts b/src/utils/index.ts index 57478a4..d218bb6 100644 --- a/src/utils/index.ts +++ b/src/utils/index.ts @@ -1,11 +1,12 @@ import { HttpsProxyAgent } from "https-proxy-agent"; import OpenAI, { ClientOptions } from "openai"; import fs from "node:fs/promises"; +import readline from "node:readline"; import { CONFIG_FILE, DEFAULT_CONFIG, HOME_DIR, - PROMPTS_DIR, + PLUGINS_DIR, } from "../constants"; export function getOpenAICommonOptions(): ClientOptions { @@ -26,7 +27,29 @@ const ensureDir = async (dir_path: string) => { export const initDir = async () => { await ensureDir(HOME_DIR); - await ensureDir(PROMPTS_DIR); + await ensureDir(PLUGINS_DIR); +}; + +const createReadline = () => { + return readline.createInterface({ + input: process.stdin, + output: process.stdout, + }); +}; + +const question = (query: string): Promise => { + return new Promise((resolve) => { + const rl = createReadline(); + rl.question(query, (answer) => { + rl.close(); + resolve(answer); + }); + }); +}; + +const confirm = async (query: string): Promise => { + const answer = await question(query); + return answer.toLowerCase() !== "n"; }; export const readConfigFile = async () => { @@ -34,8 +57,24 @@ export const readConfigFile = async () => { const config = await fs.readFile(CONFIG_FILE, "utf-8"); return JSON.parse(config); } catch { - await writeConfigFile(DEFAULT_CONFIG); - return DEFAULT_CONFIG; + const useRouter = await confirm( + "No config file found. Enable router mode? (Y/n)" + ); + if (!useRouter) { + const apiKey = await question("Enter OPENAI_API_KEY: "); + const baseUrl = await question("Enter OPENAI_BASE_URL: "); + const model = await question("Enter OPENAI_MODEL: "); + const config = Object.assign({}, DEFAULT_CONFIG, { + OPENAI_API_KEY: apiKey, + OPENAI_BASE_URL: baseUrl, + OPENAI_MODEL: model, + }); + await writeConfigFile(config); + return config; + } else { + const router = await question("Enter OPENAI_API_KEY: "); + return DEFAULT_CONFIG; + } } };