fix: gpt3计数器BUG

TBXark · Oct 7, 2023 · 004f788 · 004f788
1 parent 748eef6
commit 004f788
Show file tree

Hide file tree

Showing 12 changed files with 96 additions and 98 deletions.
diff --git a/dist/buildinfo.json b/dist/buildinfo.json
@@ -1 +1 @@
-{"sha": "682c32a", "timestamp": 1696664248}
+{"sha": "748eef6", "timestamp": 1696664893}
diff --git a/dist/index.js b/dist/index.js
@@ -26,6 +26,8 @@ var ENV = {
   MAX_TOKEN_LENGTH: 2048,
   // 使用GPT3的TOKEN计数
   GPT3_TOKENS_COUNT: false,
+  // GPT3计数器资源地址
+  GPT3_TOKENS_COUNT_REPO: "https://raw.githubusercontent.com/tbxark-arc/GPT-3-Encoder/master",
   // 全局默认初始化消息
   SYSTEM_INIT_MESSAGE: "You are a helpful assistant",
   // 全局默认初始化消息角色
@@ -39,9 +41,9 @@ var ENV = {
   // 检查更新的分支
   UPDATE_BRANCH: "master",
   // 当前版本
-  BUILD_TIMESTAMP: 1696664248,
+  BUILD_TIMESTAMP: 1696664893,
   // 当前版本 commit id
-  BUILD_VERSION: "682c32a",
+  BUILD_VERSION: "748eef6",
   I18N: null,
   LANGUAGE: "zh-cn",
   // 使用流模式
@@ -554,7 +556,7 @@ async function requestCompletionsFromOpenAI(message, history, context, onStream)
   const timeout = 1e3 * 60 * 5;
   setTimeout(() => controller.abort(), timeout);
   let url = `${ENV.OPENAI_API_DOMAIN}/v1/chat/completions`;
-  let header = {
+  const header = {
     "Content-Type": "application/json",
     "Authorization": `Bearer ${key}`
   };
@@ -660,33 +662,10 @@ async function updateBotUsage(usage, context) {
   await DATABASE.put(context.SHARE_CONTEXT.usageKey, JSON.stringify(dbValue));
 }
 
-// src/gpt3.js
-async function resourceLoader(key, url) {
-  try {
-    const raw = await DATABASE.get(key);
-    if (raw && raw !== "") {
-      return raw;
-    }
-  } catch (e) {
-    console.error(e);
-  }
-  try {
-    const bpe = await fetch(url, {
-      headers: {
-        "User-Agent": CONST.USER_AGENT
-      }
-    }).then((x) => x.text());
-    await DATABASE.put(key, bpe);
-    return bpe;
-  } catch (e) {
-    console.error(e);
-  }
-  return null;
-}
-async function gpt3TokensCounter() {
-  const repo = "https://raw.githubusercontent.com/tbxark-archive/GPT-3-Encoder/master";
-  const encoder = await resourceLoader("encoder_raw_file", `${repo}/encoder.json`).then((x) => JSON.parse(x));
-  const bpe_file = await resourceLoader("bpe_raw_file", `${repo}/vocab.bpe`);
+// src/vendors/gpt3.js
+async function gpt3TokensCounter(repo, loader) {
+  const encoder = await loader("encoder_raw_file", `${repo}/encoder.json`).then((x) => JSON.parse(x));
+  const bpe_file = await loader("bpe_raw_file", `${repo}/vocab.bpe`);
   const range = (x, y) => {
     const res = Array.from(Array(y).keys()).slice(x);
     return res;
@@ -919,7 +898,29 @@ async function tokensCounter() {
   let counter = (text) => Array.from(text).length;
   try {
     if (ENV.GPT3_TOKENS_COUNT) {
-      counter = await gpt3TokensCounter();
+      const loader = async (key, url) => {
+        try {
+          const raw = await DATABASE.get(key);
+          if (raw && raw !== "") {
+            return raw;
+          }
+        } catch (e) {
+          console.error(e);
+        }
+        try {
+          const bpe = await fetch(url, {
+            headers: {
+              "User-Agent": CONST.USER_AGENT
+            }
+          }).then((x) => x.text());
+          await DATABASE.put(key, bpe);
+          return bpe;
+        } catch (e) {
+          console.error(e);
+        }
+        return null;
+      };
+      counter = await gpt3TokensCounter(ENV.GPT3_TOKENS_COUNT_REPO, loader);
     }
   } catch (e) {
     console.error(e);
@@ -2259,7 +2260,7 @@ async function defaultIndexAction() {
 }
 async function gpt3TokenTest(request) {
   const text = new URL(request.url).searchParams.get("text") || "Hello World";
-  const counter = await gpt3TokensCounter();
+  const counter = await tokensCounter();
   const HTML = renderHTML(`
     <h1>ChatGPT-Telegram-Workers</h1>
     <br/>

diff --git a/dist/timestamp b/dist/timestamp
@@ -1 +1 @@
-1696664248
+1696664893
diff --git a/src/chat.js b/src/chat.js
@@ -1,15 +1,14 @@
 import {
   deleteMessageFromTelegramWithContext,
   sendChatActionToTelegramWithContext,
-  sendMessageToTelegramWithContext
+  sendMessageToTelegramWithContext,
 } from './telegram.js';
 import {DATABASE, ENV} from './env.js';
 // eslint-disable-next-line no-unused-vars
 import {Context} from './context.js';
-import {requestCompletionsFromOpenAI} from "./openai.js";
-import {tokensCounter} from "./utils.js";
-import {isWorkersAIEnable, requestCompletionsFromWorkersAI} from "./workers-ai.js";
-
+import {requestCompletionsFromOpenAI} from './openai.js';
+import {tokensCounter} from './utils.js';
+import {isWorkersAIEnable, requestCompletionsFromWorkersAI} from './workers-ai.js';
 
 
 /**
@@ -106,7 +105,6 @@ async function loadHistory(key, context) {
 }
 
 
-
 /**
  *
  * @param {string} text
@@ -171,7 +169,7 @@ export async function chatWithLLM(text, context, modifier) {
 
     let llm = requestCompletionsFromOpenAI;
     if (isWorkersAIEnable(context)) {
-      llm = requestCompletionsFromWorkersAI
+      llm = requestCompletionsFromWorkersAI;
     }
 
     const answer = await requestCompletionsFromLLM(text, context, llm, modifier, onStream);

diff --git a/src/context.js b/src/context.js
@@ -196,7 +196,7 @@ export class Context {
    */
   openAIKeyFromContext() {
     if (ENV.AZURE_COMPLETIONS_API) {
-      return ENV.AZURE_API_KEY
+      return ENV.AZURE_API_KEY;
     }
     if (this.USER_CONFIG.OPENAI_API_KEY) {
       return this.USER_CONFIG.OPENAI_API_KEY;

diff --git a/src/env.js b/src/env.js
@@ -13,6 +13,7 @@
  * @property {number} MAX_HISTORY_LENGTH
  * @property {number} MAX_TOKEN_LENGTH
  * @property {boolean} GPT3_TOKENS_COUNT
+ * @property {string} GPT3_TOKENS_COUNT_REPO
  * @property {string} SYSTEM_INIT_MESSAGE
  * @property {string} SYSTEM_INIT_MESSAGE_ROLE
  * @property {boolean} ENABLE_USAGE_STATISTICS
@@ -67,6 +68,8 @@ export const ENV = {
   MAX_TOKEN_LENGTH: 2048,
   // 使用GPT3的TOKEN计数
   GPT3_TOKENS_COUNT: false,
+  // GPT3计数器资源地址
+  GPT3_TOKENS_COUNT_REPO: 'https://raw.githubusercontent.com/tbxark-arc/GPT-3-Encoder/master',
   // 全局默认初始化消息
   SYSTEM_INIT_MESSAGE: 'You are a helpful assistant',
   // 全局默认初始化消息角色
@@ -106,11 +109,11 @@ export const ENV = {
 
   // Azure API Key
   AZURE_API_KEY: null,
-    // Azure Completions API
+  // Azure Completions API
   AZURE_COMPLETIONS_API: null,
 
   // workers ai模型
-  WORKERS_AI_MODEL: null
+  WORKERS_AI_MODEL: null,
 
 };
 

diff --git a/src/message.js b/src/message.js
@@ -4,8 +4,8 @@ import {sendMessageToTelegramWithContext} from './telegram.js';
 import {handleCommandMessage} from './command.js';
 import {errorToString} from './utils.js';
 import {chatWithLLM} from './chat.js';
-import {isOpenAIEnable} from "./openai.js";
-import {isWorkersAIEnable} from "./workers-ai.js";
+import {isOpenAIEnable} from './openai.js';
+import {isWorkersAIEnable} from './workers-ai.js';
 // import {TelegramMessage, TelegramWebhookRequest} from './type.d.ts';
 
 
@@ -79,7 +79,7 @@ async function msgIgnoreOldMessage(message, context) {
  * @return {Promise<Response>}
  */
 async function msgCheckEnvIsReady(message, context) {
-  const llmEnable = isOpenAIEnable(context) || isWorkersAIEnable(context)
+  const llmEnable = isOpenAIEnable(context) || isWorkersAIEnable(context);
   if (!llmEnable) {
     return sendMessageToTelegramWithContext(context)('LLM Not Set');
   }

diff --git a/src/openai.js b/src/openai.js
@@ -63,15 +63,15 @@ export async function requestCompletionsFromOpenAI(message, history, context, on
   setTimeout(() => controller.abort(), timeout);
 
   let url = `${ENV.OPENAI_API_DOMAIN}/v1/chat/completions`;
-  let header = {
+  const header = {
     'Content-Type': 'application/json',
     'Authorization': `Bearer ${key}`,
-  }
+  };
   if (ENV.AZURE_COMPLETIONS_API) {
     url = ENV.AZURE_COMPLETIONS_API;
-    header['api-key'] = key
-    delete header['Authorization']
-    delete body.model
+    header['api-key'] = key;
+    delete header['Authorization'];
+    delete body.model;
   }
   const resp = await fetch(url, {
     method: 'POST',

diff --git a/src/router.js b/src/router.js
@@ -2,8 +2,7 @@ import {handleMessage} from './message.js';
 import {API_GUARD, DATABASE, ENV} from './env.js';
 import {bindCommandForTelegram, commandsDocument} from './command.js';
 import {bindTelegramWebHook, getBot} from './telegram.js';
-import {errorToString, historyPassword, makeResponse200, renderHTML} from './utils.js';
-import {gpt3TokensCounter} from './gpt3.js';
+import {errorToString, historyPassword, makeResponse200, renderHTML, tokensCounter} from './utils.js';
 
 
 const helpLink = 'https://github.com/TBXark/ChatGPT-Telegram-Workers/blob/master/doc/DEPLOY.md';
@@ -160,7 +159,7 @@ async function defaultIndexAction() {
  */
 async function gpt3TokenTest(request) {
   const text = new URL(request.url).searchParams.get('text') || 'Hello World';
-  const counter = await gpt3TokensCounter();
+  const counter = await tokensCounter();
   const HTML = renderHTML(`
     <h1>ChatGPT-Telegram-Workers</h1>
     <br/>

diff --git a/src/utils.js b/src/utils.js
@@ -1,5 +1,5 @@
 import {CONST, DATABASE, ENV} from './env.js';
-import {gpt3TokensCounter} from './gpt3.js';
+import {gpt3TokensCounter} from './vendors/gpt3.js';
 
 /**
  * @param {number} length
@@ -127,7 +127,29 @@ export async function tokensCounter() {
   let counter = (text) => Array.from(text).length;
   try {
     if (ENV.GPT3_TOKENS_COUNT) {
-      counter = await gpt3TokensCounter();
+      const loader = async (key, url) => {
+        try {
+          const raw = await DATABASE.get(key);
+          if (raw && raw !== '') {
+            return raw;
+          }
+        } catch (e) {
+          console.error(e);
+        }
+        try {
+          const bpe = await fetch(url, {
+            headers: {
+              'User-Agent': CONST.USER_AGENT,
+            },
+          }).then((x) => x.text());
+          await DATABASE.put(key, bpe);
+          return bpe;
+        } catch (e) {
+          console.error(e);
+        }
+        return null;
+      };
+      counter = await gpt3TokensCounter( ENV.GPT3_TOKENS_COUNT_REPO, loader);
     }
   } catch (e) {
     console.error(e);

diff --git a/src/gpt3.js → src/vendors/gpt3.js b/src/gpt3.js → src/vendors/gpt3.js
@@ -1,35 +1,10 @@
-/* eslint-disable camelcase, require-jsdoc */
+/* eslint-disable */
 // https://github.com/latitudegames/GPT-3-Encoder
 
-import {CONST, DATABASE} from './env.js';
 
-async function resourceLoader(key, url) {
-  try {
-    const raw = await DATABASE.get(key);
-    if (raw && raw !== '') {
-      return raw;
-    }
-  } catch (e) {
-    console.error(e);
-  }
-  try {
-    const bpe = await fetch(url, {
-      headers: {
-        'User-Agent': CONST.USER_AGENT,
-      },
-    }).then((x) => x.text());
-    await DATABASE.put(key, bpe);
-    return bpe;
-  } catch (e) {
-    console.error(e);
-  }
-  return null;
-}
-
-export async function gpt3TokensCounter() {
-  const repo = 'https://raw.githubusercontent.com/tbxark-archive/GPT-3-Encoder/master';
-  const encoder = await resourceLoader('encoder_raw_file', `${repo}/encoder.json`).then((x) => JSON.parse(x));
-  const bpe_file = await resourceLoader('bpe_raw_file', `${repo}/vocab.bpe`);
+export async function gpt3TokensCounter(repo, loader) {
+  const encoder = await loader('encoder_raw_file', `${repo}/encoder.json`).then((x) => JSON.parse(x));
+  const bpe_file = await loader('bpe_raw_file', `${repo}/vocab.bpe`);
 
   const range = (x, y) => {
     const res = Array.from(Array(y).keys()).slice(x);

diff --git a/src/workers-ai.js b/src/workers-ai.js
@@ -1,13 +1,14 @@
-import {ENV, AI_LLM} from "./env.js";
-import {Ai} from "./vendors/cloudflare-ai.js";
+import {ENV, AI_LLM} from './env.js';
+import {Ai} from './vendors/cloudflare-ai.js';
 
 
 /**
+ * @param {Context} context
  * @return {boolean}
  */
 export function isWorkersAIEnable(context) {
-    return AI_LLM !== null;
-    // return ENV.WORKERS_AI_MODEL !== null;
+  return AI_LLM !== null;
+  // return ENV.WORKERS_AI_MODEL !== null;
 }
 
 
@@ -21,12 +22,11 @@ export function isWorkersAIEnable(context) {
  * @return {Promise<string>}
  */
 export async function requestCompletionsFromWorkersAI(message, history, context, onStream) {
-
-    const ai = new Ai(AI_LLM);
-    const model = ENV.WORKERS_AI_MODEL || '@cf/meta/llama-2-7b-chat-int8'
-    const request = {
-        messages: [...history || [], { role: "user", content: message }]
-    };
-    const response = await ai.run(model, request);
-    return response.response;
+  const ai = new Ai(AI_LLM);
+  const model = ENV.WORKERS_AI_MODEL || '@cf/meta/llama-2-7b-chat-int8';
+  const request = {
+    messages: [...history || [], {role: 'user', content: message}],
+  };
+  const response = await ai.run(model, request);
+  return response.response;
 }
Original file line number	Diff line number	Diff line change
		@@ -1 +1 @@
		{"sha": "682c32a", "timestamp": 1696664248}
		{"sha": "748eef6", "timestamp": 1696664893}