Files
DocsGPT/frontend/src/conversation/Conversation.tsx
2023-08-05 21:58:54 +01:00

147 lines
5.1 KiB
TypeScript

import { Fragment, useEffect, useRef } from 'react';
import { useDispatch, useSelector } from 'react-redux';
import Hero from '../Hero';
import { AppDispatch } from '../store';
import ConversationBubble from './ConversationBubble';
import {
addQuery,
fetchAnswer,
selectQueries,
selectStatus,
updateQuery,
} from './conversationSlice';
import Send from './../assets/send.svg';
import Spinner from './../assets/spinner.svg';
import { FEEDBACK, Query } from './conversationModels';
import { sendFeedback } from './conversationApi';
export default function Conversation() {
const queries = useSelector(selectQueries);
const status = useSelector(selectStatus);
const dispatch = useDispatch<AppDispatch>();
const endMessageRef = useRef<HTMLDivElement>(null);
const inputRef = useRef<HTMLDivElement>(null);
useEffect(
() => endMessageRef?.current?.scrollIntoView({ behavior: 'smooth' }),
[queries.length, queries[queries.length - 1]],
);
const handleQuestion = (question: string) => {
dispatch(addQuery({ prompt: question }));
dispatch(fetchAnswer({ question }));
};
const handleFeedback = (query: Query, feedback: FEEDBACK, index: number) => {
const prevFeedback = query.feedback;
dispatch(updateQuery({ index, query: { feedback } }));
sendFeedback(query.prompt, query.response!, feedback).catch(() =>
dispatch(updateQuery({ index, query: { feedback: prevFeedback } })),
);
};
const prepResponseView = (query: Query, index: number) => {
let responseView;
if (query.error) {
responseView = (
<ConversationBubble
ref={endMessageRef}
className={`${index === queries.length - 1 ? 'mb-24' : 'mb-7'}`}
key={`${index}ERROR`}
message={query.error}
type="ERROR"
></ConversationBubble>
);
} else if (query.response) {
responseView = (
<ConversationBubble
ref={endMessageRef}
className={`${index === queries.length - 1 ? 'mb-24' : 'mb-7'}`}
key={`${index}ANSWER`}
message={query.response}
type={'ANSWER'}
sources={query.sources}
feedback={query.feedback}
handleFeedback={(feedback: FEEDBACK) =>
handleFeedback(query, feedback, index)
}
></ConversationBubble>
);
}
return responseView;
};
const handlePaste = (e: React.ClipboardEvent) => {
e.preventDefault();
const text = e.clipboardData.getData('text/plain');
document.execCommand('insertText', false, text);
};
return (
<div className="flex justify-center p-4">
{queries.length > 0 && (
<div className="mt-20 flex flex-col transition-all md:w-3/4">
{queries.map((query, index) => {
return (
<Fragment key={index}>
<ConversationBubble
ref={endMessageRef}
className={'mb-7'}
key={`${index}QUESTION`}
message={query.prompt}
type="QUESTION"
sources={query.sources}
></ConversationBubble>
{prepResponseView(query, index)}
</Fragment>
);
})}
</div>
)}
{queries.length === 0 && <Hero className="mt-24 md:mt-52"></Hero>}
<div className="fixed bottom-0 flex w-10/12 flex-col items-end self-center md:w-[50%]">
<div className="flex w-full">
<div
ref={inputRef}
contentEditable
onPaste={handlePaste}
className={`border-000000 overflow-x-hidden; max-h-24 min-h-[2.6rem] w-full overflow-y-auto whitespace-pre-wrap rounded-xl border bg-white py-2 pl-4 pr-9 leading-7 opacity-100 focus:outline-none`}
onKeyDown={(e) => {
if (e.key === 'Enter' && !e.shiftKey) {
e.preventDefault();
if (inputRef.current?.textContent && status !== 'loading') {
handleQuestion(inputRef.current.textContent);
inputRef.current.textContent = '';
}
}
}}
></div>
{status === 'loading' ? (
<img
src={Spinner}
className="relative right-[38px] bottom-[7px] -mr-[30px] animate-spin cursor-pointer self-end"
></img>
) : (
<div className="relative right-[43px] bottom-[7px] -mr-[35px] h-[35px] w-[35px] cursor-pointer self-end rounded-full hover:bg-gray-3000">
<img
className="ml-[9px] mt-[9px]"
onClick={() => {
if (inputRef.current?.textContent) {
handleQuestion(inputRef.current.textContent);
inputRef.current.textContent = '';
}
}}
src={Send}
></img>
</div>
)}
</div>
<p className="w-[100vw] self-center bg-white p-5 text-center text-xs text-gray-2000">
This is a chatbot that uses the GPT-3, Faiss and LangChain to answer
questions.
</p>
</div>
</div>
);
}