GeneralBots/botserver
2
1
Fork 0
Code Issues Pull requests 1 Projects Releases 323 Packages Wiki Activity Actions

new(basic.gblib): GPT replacing ALLEN NLP Reading Comp.

Browse source
This commit is contained in:
Rodrigo Rodriguez 2024-01-30 19:21:04 -03:00
parent 3ddeabdf19
commit 2a3e0f5782
4 changed files with 48 additions and 36 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
package.json
View file

@ -120,6 +120,7 @@
"ffmpeg-static": "5.1.0",
"google-libphonenumber": "3.2.31",
"googleapis": "126.0.1",
"hnswlib-node": "1.4.2",
"ibm-watson": "7.1.2",
"join-images-updated": "1.1.4",
"js-md5": "0.8.3",
@ -148,6 +149,7 @@
"open-docxtemplater-image-module": "1.0.3",
"openai": "4.6.0",
"pdf-extraction": "1.0.2",
"pdf-parse": "1.1.1",
"pdf-to-png-converter": "3.1.0",
"pdfkit": "0.13.0",
"phone": "3.1.30",

2
packages/core.gbapp/services/GBDeployer.ts
View file

@ -56,7 +56,7 @@ import MicrosoftGraph from '@microsoft/microsoft-graph-client';
import { GBLogEx } from './GBLogEx.js';
import { DialogKeywords } from '../../basic.gblib/services/DialogKeywords.js';
import { GBUtil } from '../../../src/util.js';
import { HNSWLib } from 'langchain/vectorstores/hnswlib.js';
import { HNSWLib } from 'langchain/vectorstores/hnswlib';
import { OpenAIEmbeddings } from 'langchain/embeddings/openai';
/**

2
packages/kb.gbapp/dialogs/AskDialog.ts
View file

@ -340,7 +340,7 @@ export class AskDialog extends IGBDialog {
}
private static async handleAnswer(service: KBService, min: GBMinInstance, step: any, user, answer: GuaribasAnswer) {
const text = answer.content;
const text = typeof(answer)==='string'? answer:answer.content;
if (text.endsWith('.docx')) {
const mainName = GBVMService.getMethodNameFromVBSFilename(text);
await step.endDialog();

78
packages/kb.gbapp/services/KBService.ts
View file

@ -357,22 +357,20 @@ export class KBService implements IGBKBService {
return { answer: undefined, questionId: 0 };
}
} else {
GBLog.info(
`SEARCH called but returned LOW level score,
returnedScore: ${returnedScore} < required (searchScore): ${searchScore}`
);
return await this.answerByGPT(min,
query,
searchScore,
subjects
);
}
}
return { answer: undefined, questionId: 0 };
GBLog.info(
`SEARCH returned LOW level score, calling GPT
returnedScore: ${returnedScore} < required (searchScore): ${searchScore}`
);
return await this.answerByGPT(min,
query,
searchScore,
subjects
);
}
}
@ -395,7 +393,7 @@ export class KBService implements IGBKBService {
searchScore: number,
subjects: GuaribasSubject[]
) {
const contextVectorStore = min['VectorStore'];
const contextVectorStore = min['vectorStore'];
const question = query.trim().replaceAll('\n', ' ');
const context = await this.getRelevantContext(contextVectorStore, question, 1);
@ -404,7 +402,18 @@ export class KBService implements IGBKBService {
const chatPrompt = ChatPromptTemplate.fromPromptMessages([
systemPrompt,
HumanMessagePromptTemplate.fromTemplate('QUESTION: """{input}"""'),
HumanMessagePromptTemplate.fromTemplate(`Answer in pt-br.
You have access to the context (RELEVANTDOCS) provided by the user.
When answering think about whether the question in RELEVANTDOCS, but never mention
to user about the source.
Don’t justify your answers. Don't refer to yourself in any of the created content.
Don´t prefix RESPONSE: when answering the user.
RELEVANTDOCS: {context}
QUESTION: """{input}"""
`),
]);
const windowMemory = new BufferWindowMemory({
returnMessages: false,
@ -440,7 +449,7 @@ export class KBService implements IGBKBService {
});
if (response) {
return { answer: response.response, questionId: 0 };
return { answer: response.text, questionId: 0 };
}
return { answer: undefined, questionId: 0 };
@ -686,32 +695,33 @@ export class KBService implements IGBKBService {
return await GuaribasQuestion.bulkCreate(questions);
}
public async sendAnswer(min: GBMinInstance, channel: string, step: GBDialogStep, answer: GuaribasAnswer) {
if (answer.content.endsWith('.mp4')) {
public async sendAnswer(min: GBMinInstance, channel: string, step: GBDialogStep, answer) {
answer = typeof (answer) === 'string' ? answer : answer.content;
if (answer.endsWith('.mp4')) {
await this.playVideo(min, min.conversationalService, step, answer, channel);
} else if (
answer.content.endsWith('.ppt') ||
answer.content.endsWith('.pptx') ||
answer.content.endsWith('.doc') ||
answer.content.endsWith('.docx') ||
answer.content.endsWith('.xls') ||
answer.content.endsWith('.xlsx')
answer.endsWith('.ppt') ||
answer.endsWith('.pptx') ||
answer.endsWith('.doc') ||
answer.endsWith('.docx') ||
answer.endsWith('.xls') ||
answer.endsWith('.xlsx')
) {
const path = DialogKeywords.getGBAIPath(min.botId, `gbkb`);
const doc = urlJoin(GBServer.globals.publicAddress, 'kb', path, 'assets', answer.content);
const doc = urlJoin(GBServer.globals.publicAddress, 'kb', path, 'assets', answer);
const url = `http://view.officeapps.live.com/op/view.aspx?src=${doc}`;
await this.playUrl(min, min.conversationalService, step, url, channel);
} else if (answer.content.endsWith('.pdf')) {
} else if (answer.endsWith('.pdf')) {
const path = DialogKeywords.getGBAIPath(min.botId, `gbkb`);
const url = urlJoin('kb', path, 'assets', answer.content);
const url = urlJoin('kb', path, 'assets', answer);
await this.playUrl(min, min.conversationalService, step, url, channel);
} else if (answer.format === '.md') {
await min.conversationalService['playMarkdown'](min, answer.content, channel, step,
await min.conversationalService['playMarkdown'](min, answer, channel, step,
GBMinService.userMobile(step));
} else if (answer.content.endsWith('.ogg') && process.env.AUDIO_DISABLED !== 'true') {
} else if (answer.endsWith('.ogg') && process.env.AUDIO_DISABLED !== 'true') {
await this.playAudio(min, answer, channel, step, min.conversationalService);
} else {
await min.conversationalService.sendText(min, step, answer.content);
await min.conversationalService.sendText(min, step, answer);
await min.conversationalService.sendEvent(min, step, 'stop', undefined);
}
}
@ -934,13 +944,13 @@ export class KBService implements IGBKBService {
}
defaultRecursiveCharacterTextSplitter = new RecursiveCharacterTextSplitter({
chunkSize: 1024,
chunkOverlap: 1024,
chunkSize: 700,
chunkOverlap: 50,
});
markdownRecursiveCharacterTextSplitter = RecursiveCharacterTextSplitter.fromLanguage('markdown', {
chunkSize: 1024,
chunkOverlap: 1024,
chunkSize: 700,
chunkOverlap: 50,
});