vue通过ollama接口调用开源模型实现人机对话功能
更新时间:2024年11月11日 09:41:17 作者:小G同学
文章介绍了如何在本地安装ollama并配置开源大模型,以及如何通过JavaScript和Vue.js实现人机对话功能,感兴趣的朋友一起看看吧
先展示下最终效果:
第一步:先安装ollama,并配置对应的开源大模型。
安装步骤可以查看上一篇博客:
第二步:需要注意两个配置,页面才可以调用
1)OLLAMA_HOST= "0.0.0.0:11434"
2)若应用部署服务器后想调用,需要配置:OLLAMA_ORIGINS=*
第三步:js流式调用大模型接口方法
async startStreaming(e) { if(e.ctrkey&&e.keyCode==13){ this.form.desc+='\n'; } document.getElementById("txt_suiwen").disabled="true"; // 如果已经有一个正在进行的流式请求,则中止它 if (this.controller) { this.controller.abort(); } setTimeout(()=>{ this.scrollToBottom(); },50); var mymsg=this.form.desc.trim(); if(mymsg.length>0){ this.form.desc=''; this.message.push({ user:this.username, msg:mymsg }) this.message.push({ user:'GPT', msg:'', dot:'' }); // 创建一个新的 AbortController 实例 this.controller = new AbortController(); const signal = this.controller.signal; this.arequestData.messages.push({role:"user",content:mymsg}); try { const response = await fetch('http://127.0.0.1:11434/api/chat', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body:JSON.stringify(this.arequestData), signal }); if (!response.body) { this.message[this.message.length-1].msg='ReadableStream not yet supported in this browser.'; throw new Error('ReadableStream not yet supported in this browser.'); } const reader = response.body.getReader(); const decoder = new TextDecoder(); let result = ''; this.message[this.message.length-1].dot='⚪'; while (true) { const { done, value } = await reader.read(); if (done) { break; } result += decoder.decode(value, { stream: true }); // 处理流中的每一块数据,这里假设每块数据都是完整的 JSON 对象 const jsonChunks = result.split('\n').filter(line => line.trim()); //console.log(result) for (const chunk of jsonChunks) { try { const data = JSON.parse(chunk); //console.log(data.message.content) this.message[this.message.length-1].msg+=data.message.content; setTimeout(()=>{ this.scrollToBottom(); },50); } catch (e) { //this.message[this.message.length-1].msg=e; // 处理 JSON 解析错误 //console.error('Failed to parse JSON:', e); } } // 清空 result 以便处理下一块数据 result = ''; } } catch (error) { if (error.name === 'AbortError') { console.log('Stream aborted'); this.message[this.message.length-1].msg='Stream aborted'; } else { console.error('Streaming error:', error); this.message[this.message.length-1].msg='Stream error'+error; } } this.message[this.message.length-1].dot=''; this.arequestData.messages.push({ role: 'assistant',//this.message[this.message.length-1].user,//"GPT", content: this.message[this.message.length-1].msg }) setTimeout(()=>{ this.scrollToBottom(); },50); }else{ this.form.desc=''; } document.getElementById("txt_suiwen").disabled=""; document.getElementById("txt_suiwen").focus(); } }
vue完整代码如下:
<template> <el-row :gutter="12" class="demo-radius"> <div class="radius" :style="{ borderRadius: 'base' }"> <div class="messge" id="messgebox" ref="scrollDiv"> <ul> <li v-for="(item, index) in message" :key="index" style="list-style-type:none;"> <div v-if="item.user == username" class="mymsginfo" style="float:right"> <div> <el-avatar style="float: right;margin-right: 30px;background: #01bd7e;"> <!-- {{ item.user.substring(0, 2) }} --> <img :alt="item.user.substring(0, 2)" :src=userphoto /> </el-avatar> </div><div style="float: right;margin-right: 10px;margin-top:10px;width:80%;text-align: right;"> {{ item.msg }} </div> </div> <div v-else class="chatmsginfo" > <div> <el-avatar style="float: left;margin-right: 10px;"> {{ item.user }} </el-avatar> </div> <div style="float: left;margin-top:10px;width:80%;"> <img alt="loading" v-if="item.msg == ''" class="loading" src="../../assets/loading.gif"/> <MdPreview style="margin-top:-20px;" :autoFoldThreshold="9999" :editorId="id" :modelValue=" item.msg + item.dot " /> <!-- {{ item.msg }} --> </div> </div> </li> </ul> </div> <div class="inputmsg"> <el-form :model="form" > <el-form-item > <el-avatar style="float: left;background: #01bd7e;margin-bottom: -44px;margin-left: 4px;z-index: 999;width: 30px;height: 30px;"> <img alt="jin" :src=userphoto /> </el-avatar> <el-input id="txt_suiwen" :prefix-icon="userphoto" resize="none" autofocus="true" :autosize="{ minRows: 1, maxRows: 2 }" v-model="form.desc" placeholder="说说你想问点啥....按Enter键可直接发送" @keydown.enter.native.prevent="startStreaming($event)" type="textarea" /> </el-form-item> </el-form> </div> </div> </el-row> </template> <script setup> import { MdPreview, MdCatalog } from 'md-editor-v3'; import 'md-editor-v3/lib/preview.css'; const id = 'preview-only'; </script> <script> export default { data() { return { form: { desc: '' }, message:[], username:sessionStorage.name, userphoto:sessionStorage.photo, loadingtype:false, controller: null, arequestData : { model: "qwen2",//"llama3.1", messages: [] } } }, mounted() { }, methods: { scrollToBottom() { let elscroll=this.$refs["scrollDiv"]; elscroll.scrollTop = elscroll.scrollHeight+30 }, clearForm(formName){ this.form.desc=''; }, async startStreaming(e) { if(e.ctrkey&&e.keyCode==13){ this.form.desc+='\n'; } document.getElementById("txt_suiwen").disabled="true"; // 如果已经有一个正在进行的流式请求,则中止它 if (this.controller) { this.controller.abort(); } setTimeout(()=>{ this.scrollToBottom(); },50); var mymsg=this.form.desc.trim(); if(mymsg.length>0){ this.form.desc=''; this.message.push({ user:this.username, msg:mymsg }) this.message.push({ user:'GPT', msg:'', dot:'' }); // 创建一个新的 AbortController 实例 this.controller = new AbortController(); const signal = this.controller.signal; this.arequestData.messages.push({role:"user",content:mymsg}); try { const response = await fetch('http://127.0.0.1:11434/api/chat', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body:JSON.stringify(this.arequestData), signal }); if (!response.body) { this.message[this.message.length-1].msg='ReadableStream not yet supported in this browser.'; throw new Error('ReadableStream not yet supported in this browser.'); } const reader = response.body.getReader(); const decoder = new TextDecoder(); let result = ''; this.message[this.message.length-1].dot='⚪'; while (true) { const { done, value } = await reader.read(); if (done) { break; } result += decoder.decode(value, { stream: true }); // 处理流中的每一块数据,这里假设每块数据都是完整的 JSON 对象 const jsonChunks = result.split('\n').filter(line => line.trim()); //console.log(result) for (const chunk of jsonChunks) { try { const data = JSON.parse(chunk); //console.log(data.message.content) this.message[this.message.length-1].msg+=data.message.content; setTimeout(()=>{ this.scrollToBottom(); },50); } catch (e) { //this.message[this.message.length-1].msg=e; // 处理 JSON 解析错误 //console.error('Failed to parse JSON:', e); } } // 清空 result 以便处理下一块数据 result = ''; } } catch (error) { if (error.name === 'AbortError') { console.log('Stream aborted'); this.message[this.message.length-1].msg='Stream aborted'; } else { console.error('Streaming error:', error); this.message[this.message.length-1].msg='Stream error'+error; } } this.message[this.message.length-1].dot=''; this.arequestData.messages.push({ role: 'assistant',//this.message[this.message.length-1].user,//"GPT", content: this.message[this.message.length-1].msg }) setTimeout(()=>{ this.scrollToBottom(); },50); }else{ this.form.desc=''; } document.getElementById("txt_suiwen").disabled=""; document.getElementById("txt_suiwen").focus(); } }, beforeDestroy() { // 组件销毁时中止流式请求 if (this.controller) { this.controller.abort(); } } } </script> <style scoped> .radius{ margin:0 auto; } .demo-radius .title { color: var(--el-text-color-regular); font-size: 18px; margin: 10px 0; } .demo-radius .value { color: var(--el-text-color-primary); font-size: 16px; margin: 10px 0; } .demo-radius .radius { min-height: 580px; height: 85vh; width: 70%; border: 1px solid var(--el-border-color); border-radius: 14px; margin-top: 10px; } .messge{ width:96%; height:84%; /* border:1px solid red; */ margin: 6px auto; overflow: hidden; overflow-y: auto; } .inputmsg{ width:96%; height:12%; /* border:1px solid blue; */ border-top:2px solid #ccc; margin: 4px auto; padding-top: 10px; } .mymsginfo{ width:100%; height:auto; min-height:50px; } ::-webkit-scrollbar {width: 6px;height: 5px; } ::-webkit-scrollbar-track {background-color: rgba(0, 0, 0, 0.2);border-radius: 10px; } ::-webkit-scrollbar-thumb {background-color: rgba(0, 0, 0, 0.5);border-radius: 10px; } ::-webkit-scrollbar-button {background-color: #7c2929;height: 0;width: 0px; } ::-webkit-scrollbar-corner {background-color: black; } </style> <style> .el-textarea__inner{ padding-left: 45px; padding-top: .75rem; padding-bottom: .75rem; } </style>
到此这篇关于vue通过ollama接口调用开源模型实现人机对话的文章就介绍到这了,更多相关vue ollama接口人机对话内容请搜索脚本之家以前的文章或继续浏览下面的相关文章希望大家以后多多支持脚本之家!
相关文章
element ui设置table自适应表格宽,不自动换行问题
这篇文章主要介绍了element ui设置table自适应表格宽,不自动换行问题,具有很好的参考价值,希望对大家有所帮助,如有错误或未考虑完全的地方,望不吝赐教2023-12-12
最新评论