feat: add google chat model

package.json

@@ -33,6 +33,7 @@
     "node": ">=16"
   },
   "dependencies": {
+    "@langchain/google-genai": "^0.0.10",
     "ai": "^2.2.10",
     "ioredis": "^5.3.2",
     "langchain": "^0.0.178",

src/interface/agent.interface.ts

@@ -1,6 +1,6 @@
 import { DataSource } from 'typeorm';
 
-export type LLM_TYPE = 'azure' | 'gpt' | 'aws';
+export type LLM_TYPE = 'azure' | 'gpt' | 'aws' | 'google';
 export type DATABASE_TYPE = 'cosmos' | 'redis' | 'postgres';
 
 export const SYSTEM_MESSAGE_DEFAULT = `
@@ -58,9 +58,9 @@ export interface IChatConfig {
 export interface ILLMConfig {
   type: LLM_TYPE;
   model: string;
-  instance: string;
+  instance?: string;
   apiKey: string;
-  apiVersion: string;
+  apiVersion?: string;
 }
 
 export interface IVectorStoreConfig {

src/services/chain/index.ts

@@ -13,6 +13,7 @@ import {
   SystemMessagePromptTemplate,
 } from 'langchain/prompts';
 
+import { AIMessage } from 'langchain/schema';
 import {
   IAgentConfig,
   SYSTEM_MESSAGE_DEFAULT,
@@ -90,11 +91,12 @@ class ChainService {
         this.buildSystemMessages(systemMessages)
       ),
       new MessagesPlaceholder('chat_history'),
+      new AIMessage('Olá! Em que posso ajudar?'),
       HumanMessagePromptTemplate.fromTemplate('{question}'),
     ];
 
     const CHAT_COMBINE_PROMPT =
-      ChatPromptTemplate.fromPromptMessages(combine_messages);
+      ChatPromptTemplate.fromMessages(combine_messages);
 
     return CHAT_COMBINE_PROMPT;
   }

src/services/chain/sql-database-chain.ts

@@ -12,7 +12,7 @@ import {
   MessagesPlaceholder,
   SystemMessagePromptTemplate,
 } from 'langchain/prompts';
-import { ChainValues } from 'langchain/schema';
+import { AIMessage, ChainValues } from 'langchain/schema';
 import { StringOutputParser } from 'langchain/schema/output_parser';
 import { RunnableSequence } from 'langchain/schema/runnable';
 import { SqlDatabase } from 'langchain/sql_db';
@@ -88,6 +88,7 @@ export default class SqlDatabaseChain extends BaseChain {
   getSQLPrompt(): string {
     return `
       Based on the SQL table schema provided below, write an SQL query that answers the user's question.\n
+      Remeber to put double quotes around database table names\n
       Your response must only be a valid SQL query, based on the schema provided.\n
       -------------------------------------------\n
       Here are some important observations for generating the query:\n
@@ -152,6 +153,7 @@ export default class SqlDatabaseChain extends BaseChain {
     const combine_messages = [
       SystemMessagePromptTemplate.fromTemplate(systemMessages),
       new MessagesPlaceholder('chat_history'),
+      new AIMessage('Aguarde! Estamos pesquisando em nossa base de dados.'),
       HumanMessagePromptTemplate.fromTemplate('{question}'),
     ];
 

src/services/llm/google-llm-service.ts

@@ -0,0 +1,24 @@
+import { ChatGoogleGenerativeAI } from '@langchain/google-genai';
+import { IChatConfig, ILLMConfig } from '../../interface/agent.interface';
+
+class GoogleLLMService {
+  private _chatSettings: IChatConfig;
+  private _llmSettings: ILLMConfig;
+
+  constructor(chatSettings: IChatConfig, llmSettings: ILLMConfig) {
+    this._chatSettings = chatSettings;
+    this._llmSettings = llmSettings;
+  }
+
+  public build(): ChatGoogleGenerativeAI {
+    return new ChatGoogleGenerativeAI({
+      temperature: this._chatSettings.temperature,
+      modelName: this._llmSettings.model,
+      apiKey: this._llmSettings.apiKey,
+      maxOutputTokens: 2048,
+      streaming: true,
+    });
+  }
+}
+
+export default GoogleLLMService;

src/services/llm/index.ts

@@ -1,18 +1,20 @@
-import { IChatConfig, ILLMConfig } from '../../interface/agent.interface';
 import { BaseChatModel } from 'langchain/chat_models/base';
+import { IChatConfig, ILLMConfig } from '../../interface/agent.interface';
 
 import AzureLLMService from './azure-llm-service';
+import GoogleLLMService from './google-llm-service';
 
 const ServiceLLM = {
   azure: AzureLLMService,
+  google: GoogleLLMService,
 } as any;
 
 class LLMFactory {
-  public static create(chatSettings: IChatConfig, llmSettings: ILLMConfig): BaseChatModel {
-    return new ServiceLLM[llmSettings.type](
-      chatSettings,
-      llmSettings,
-    ).build();
+  public static create(
+    chatSettings: IChatConfig,
+    llmSettings: ILLMConfig
+  ): BaseChatModel {
+    return new ServiceLLM[llmSettings.type](chatSettings, llmSettings).build();
   }
 }