feat: llama using cf worker

commands/llama.ts

@@ -1,21 +1,15 @@
 // Import necessary modules and dependencies
 import { Client, Message } from "whatsapp-web.js";
 import config from "../config.js";
-import { LlamaModel, LlamaContext, LlamaChatSession } from "node-llama-cpp";
-
-const modelPath = config.llama_model_path; // Retrieve model path from environment variable
-
-const model = new LlamaModel({ modelPath });
-const context = new LlamaContext({ model });
-const session = new LlamaChatSession({ context });
+import axios from "../helpers/axios.js";
 
 const execute = async (client: Client, msg: Message, args: string[]) => {
   const chatId = (await msg.getChat()).id._serialized;
 
-  if (!modelPath) {
+  if (!config.cf_worker.url) {
     return client.sendMessage(
       chatId,
-      "Sorry, llama model path not specified in the environment variable."
+      "Sorry, cf worker url not specified in the environment variable."
     );
   }
 
@@ -28,11 +22,30 @@ const execute = async (client: Client, msg: Message, args: string[]) => {
 
   const text = args.join(" ") || quotedMsg.body;
 
+  const username = config.cf_worker.username;
+  const password = config.cf_worker.password;
+
+  const encodedCredentials = Buffer.from(`${username}:${password}`).toString(
+    "base64"
+  );
+  const authHeader = `Basic ${encodedCredentials}`;
+
   // Call Llama model with the obtained text
-  const response = await session.prompt(text);
+  const response = await axios.get<{ response: string }>(config.cf_worker.url, {
+    params: {
+      prompt: text,
+    },
+    headers: {
+      Authorization: authHeader,
+    },
+  });
 
-  // Send the response back to the user
-  await client.sendMessage(chatId, `Llama: ${response}`);
+  try {
+    // Send the response back to the user
+    await client.sendMessage(chatId, `Llama: ${response.data.response}`);
+  } catch {
+    await client.sendMessage(chatId, "LLaMA generation failed.");
+  }
 
   // Optionally, you can handle conversation history or context here
 

config.ts

@@ -10,7 +10,11 @@ const config = {
   default_tr_lang: process.env.DEFAULT_TR_LANG || "en",
   enable_delete_alert: process.env.ENABLE_DELETE_ALERT || "true",
   ocr_space_api_key: process.env.OCR_SPACE_API_KEY || "",
-  llama_model_path: process.env.LLAMA_MODEL_PATH || "",
+  cf_worker: {
+    url: process.env.CF_WORKER_URL,
+    username: process.env.CF_WORKER_USERNAME,
+    password: process.env.CF_WORKER_PASSWORD,
+  },
 };
 
 export default config;

docker/docker-compose.yml

@@ -33,7 +33,9 @@ services:
       BITLY_API_KEY: ${BITLY_API_KEY}
       OPENAI_API_KEY: ${OPENAI_API_KEY}
       TZ: ${TZ:-Asia/Hong_Kong}
-      LLAMA_MODEL_PATH: ${LLAMA_MODEL_PATH}
+      CF_WORKER_URL: ${CF_WORKER_URL}
+      CF_WORKER_USERNAME: ${CF_WORKER_USERNAME}
+      CF_WORKER_PASSWORD: ${CF_WORKER_PASSWORD}
     networks:
       - whatsbot-network
   wtstodis-mongo:

example.env

@@ -19,4 +19,6 @@ DISCORD_OWNER_ID=""
 BITLY_API_KEY=""
 OPENAI_API_KEY=""
 TZ="Asia/Hong_Kong"
-LLAMA_MODEL_PATH="llama/openllama-3b-v2-q4_0.gguf"
+CF_WORKER_URL=
+CF_WORKER_USERNAME=
+CF_WORKER_PASSWORD=