AI Knowledge Base

Recently, I found an eye-catching domestic open source AI knowledge base framework: KAG (Knowledge Augmented Generation). KAG is jointly launched by Ant Group, Zhejiang University and many other organizations, focusing on...

The main difference is that the level of review is different, and English content is naturally less filtered than Chinese content, see DeepSeek R1 Jailbreak: An attempt to break through DeepSeek's review mechanism. The tone of the Chinese answers to the questions is skewed towards "correct thinking". In the U.S. market, in order to satisfy the western users' need for information...

DeepSeek R1 Official Jailbreaks are great experimental environments for triggering basically all types of censorship mechanisms, and you can learn a lot of defense techniques, so this is a big model censorship mechanism learning article that will take you through examples of big model jailbreaks over the years. Large model censorship mechanisms through ...

The most basic ability of the big model is instruction following, with the document: OpenAI o3-mini system manual (in Chinese) uploaded as an attachment to allow DeepSeek-R1 and ChatGPT to write social media blasts respectively (here I used a completely inappropriate prompt...

Original: https://cdn.openai.com/o3-mini-system-card.pdf 1 Introduction The OpenAI o model family is trained using large-scale reinforcement learning to reason using chains of thought. These advanced reasoning ...

DeepSeek-R1 API The standard model name is: deepseek-reasoner DeepSeek-R1 supports cached hits Cached hits are generally used for high-frequency inputs with few sample examples, large document inputs with multiple outputs (less than 64 tok...

Quick Reads A comprehensive and in-depth look at the past and present of the Scaling Law of Large Language Models (LLMs) and the future direction of AI research. The author, Cameron R. Wolfe, provides a clear and logical and informative...

Editor's Note: DeepSeek R1 is exploding in popularity, and a lot of people who didn't use AI much are now trying to chat with big models. Today's post is about how to properly talk to DeepSeek from the perspective of an inference model.Founder Park curated...

Abstract Large-scale language models (LLMs), such as OpenAI's GPT-4, Google's PaLM, and Meta's LLaMA, have dramatically transformed Artificial Intelligence (AI) by enabling human-like text generation and natural language understanding. However, their reliance on static training data ...

Overview DeepSeek is a groundbreaking open source big language model that revolutionizes AI conversational interactions with its advanced algorithmic architecture and reflexive chaining capabilities. With a private deployment, you have full control over data security and usage security. You can also flexibly adjust the deployment scheme...

Many AI services can use .edu student email address to enjoy discounts or even free of charge, the benefits do not need to say more, those who understand understand... The .edu suffix mailboxes are getting more and more difficult to register. Considering the difficulty of submitting the information and the holding time when registering, we will give you a relatively feasible plan at present. Pre...

Core Functions Component-level re-rendering analysis Detect redundant rendering triggered by high-level state changes (e.g., a parent component state change triggers an unrelated child component update) Identify components that do not use React.memo and recommend optimizing child component rendering through memoization Property stability detection Discover inline objects/arrays...

Artificial Intelligence (AI) is a rapidly growing field. Language models have evolved to enable AI Agents to perform complex tasks and make complex decisions. However, as the skills of these Agents continue to grow, the infrastructure to support them struggles to keep up. LangGraph, which is ...

Good New Year! Greetings to all of you! Recently, my circle of friends has been "screened" by news related to DeepSeek-R1, and I believe everyone has heard about our domestic open source model - DeepSeek! I'm sure you've all heard about DeepSeek, our homegrown open source model, and the tutorials on how to deploy DeepSeek-R1 locally have been...

Introduction Like many others, over the past few days my news tweets have been filled with news, praise, complaints, and speculation about the Chinese-made DeepSeek-R1 large language model, which was released last week. The model itself is being brought up against publics like OpenAI, Meta...

Summary of the main contributions of CORAG CORAG (Cost-Constrained Retrieval Optimization for Retrieval-Augmented Generation) is a...

Knowledge distillation is a machine learning technique that aims to transfer learning from a large pre-trained model (i.e., a "teacher model") to a smaller "student model". Distillation techniques can help us develop lighter weight generative models for intelligent conversations, content creation, and other domains. Recently ...

Recently, many people engaged in large model training and reasoning have been discussing the relationship between the number of model parameters and model size. For example, the famous alpaca series LLaMA large model contains LLaMA-7B, LLaMA-13B, LLaMA-33B and LLaMA...

A First Look at MCP MCP (Model Context Protocol), is a protocol developed to standardize how applications provide context for large models.MCP provides a standard way of providing data, tools for LLMs.Using MCP will make it easier to con...

The cue word is derived from the AI programming tool CodeGuide, which utilizes this cue word and the latest Deepseek-R1 model to achieve an o1-like response effect, replacing the o1 thought process and reducing the overall API cost by 50%. The cue word tech...

Tip Please perform a comprehensive security audit of the Cerebr-main browser extension, including an...

On the same day that Cline released the Plan & Act model, DeepSeek introduced the R1 model. An interesting phenomenon ensued: developers started spontaneously using R1 for the planning phase, while using 3.5-Sonnet...

Recently Cursor started to block the account, the client will appear prompt Unauthorized request, User is unauthorized Pro account in trying to switch Enable usage-based pric...

Original article: https://arxiv.org/pdf/2412.15479 Interpretation: this article itself is not very innovative and has little application. However, it reminds me of three highly informative articles I read a long, long time ago, and reading this article in conjunction with the three previous ones will hopefully give...

In the field of artificial intelligence and machine learning, especially when building applications such as RAG (Retrieval Augmented Generation) systems and semantic search, efficiently processing and retrieving massive unstructured data becomes crucial. Vector databases have emerged as a core technology to address this challenge. They are not only for storing high-dimensional ...

Xiaohongshu, a hot social e-commerce platform in China and even in Asia, has long gone beyond a simple shopping app to become a weathervane for young people's lifestyles and a new position for brand marketing. For overseas brands and individuals wishing to enter the Chinese market or reach young consumers, mastering Xiaohongshu...

Unexpectedly, AI has set off a half-changing sky in the programming field. From v0, bolt.new to various programming tools Cursor and Windsurf that combine Agant, AI Coding has a huge potential of idea MVP. From...

Thanks to Tencent Cloud Cloud Studio, thanks to DeepSeek DeepSeek-R1 In today's world of rapid development of artificial intelligence and big model technology, more and more developers and researchers want to experience and fine-tune big models for themselves in order to better understand and apply these advanced techniques...

Most of the intelligences we developed in the past were in a fixed workflow mode, and very few of them followed the framework below to develop intelligences with autonomous decision making and autonomous use of tools. Two days ago, I shared browse, an open source Agent that can automate simple tasks using a browser...

Currently the mainstream document (or long paper) translation is generally used to convert the format, segmentation, and then translated, which requires specialized tools, such as: PDFMathTranslate, GPT Academic, etc. ... Of course, you can attach the document as an attachment...

Before we start, let's understand a few "key words": Workflow (Workflow): Simply put, it is "the complete steps to accomplish something". It's like an "instruction manual" that tells you what needs to be done, in what order, and by whom, in order to achieve your goal. Inpu...

Previously, we all know that Cursor has a 14 days and 500 free entries limit for new users. Normally, when the 14 days expire, you can cancel your account and re-register so that you can continue to have 14 days of experience. But! If used more than 500 times, C...

Once upon a time to share a lot of fun card map prompt word example, although fun, but the actual work found that there is no bird use. The reason is very simple: these card diagram prompt word template sample style code is generally fixed, the user's real intention and sample style does not match. Some people have done a more general prompt word adaptation: card map...

Information Overload in Equity Research is Real A common challenge when evaluating the value of a stock is: dealing with a large amount of information from multiple sources in order to make an informed investment decision. Traditional methods include: Collecting financial data from a variety of platforms. Reading multiple reports, news and other articles. ...

Abstract We introduce Zep, a novel memory layer service for intelligentsia that outperforms the current state-of-the-art system MemGPT in Deep Memory Retrieval (DMR) benchmarks.In addition, Zep performs well in more comprehensive and challenging evaluations than DMR, which...

This article is part of the series "Understanding and Deploying Intelligent Body AI": Intelligent Body AI Series 1: Comparison between Devin and Agent Cursor Intelligent Body AI Series 2: From Thinker to Doer - A Paradigm Revolution in Intelligent Body AI and...

"It's time to write SQL code again, can the AI assistant help today?" It's OK, supabase has official prompts, just copy them! Supabase Introduction Supabase is a PostgreSQL-based cloud data ...

Using big models for business data analysis is nothing more than generating text, SVG, Python, and what people need may not be the cue word, but learning these 9 models! 1.AARRR model (pirate model) 2.RFM model 3.Decision tree model 4.PEST analysis model...

When building large language modeling (LLM) applications, memory systems are one of the key technologies to enhance conversation context management, long-term information storage, and semantic understanding. An efficient memory system can help the model maintain consistency over long conversations, extract key information, and even have the ability to retrieve historical conversations...

Background The English-speaking world has a lot of writers who are good at writing for the web, with very different styles and a large training corpus, and AI is very good at imitating them. With the writing style of these people, the content is more understandable or has a logical framework, and it is easier to write explosive text. Features Input the writing topic, AI automatically analyzes the most matching...

Background In the design of customer service related dialogs, it is often necessary to let the user confirm the completion of the current action, and then perform the next action, there are two ways to achieve: 1. routing 2. prompt word 1. routing Generally by the big model to determine the user's state, and then perform the corresponding node service, which is the same as orchestrating the "smart...

I. Background Notes 1.1 The Need for .cursorules In Cursor, Rules for AI helps you set some basic rules for AI-generated code, such as style, naming style, etc. This way, both in code completion and command...

Qwen Chat, the overseas version of Qwen Chat, was released some time ago and was widely praised. Qwen Chat utilizes the second-developed Open WebUI chat interface to access the full range of big models of Qwen, and the Qwen2.5-plus closed-source modeling capability is very excellent....

One-click dressup is extremely hot. Today, we bring you a dressup workflow based on BizyAir cloud node, which is also on the OpenArt homepage. This workflow is based on FlUX.1 Fill and Redux models, and all of them are in Bizy...

Comprehensive Introduction Word Duo is a Google long-tail keyword mining tool designed for SEO optimizers, content creators and marketers. By automatically obtaining relevant search term suggestions from the Google search box and related questions in the search results, it helps users accurately mine long-tail keywords to improve SEO results overall...

OpenAI Function calling V2 Features The core goal of Function calling V2 is to give OpenAI models the ability to interact with the outside world, which is mainly reflected in the following two core functions: data acquisition ...

ChatGPT client has more practical functions than the web version, and the use of ChatGPT application will not "downgrade wisdom", but the domestic application store can not be installed in the normal ChatGPT. The following share a way to bypass the regional restrictions of Microsoft's application store, you can download and install ChatGPT ...

Many foreign websites force you to bind your credit card after logging in before you are allowed to use them, which can be a problem for kids who don't have an international credit card. Credit card verification does not mean paying, but simply verifying identity. Here we find you a way to pass US credit card verification with 20% probability. Note ...

Have you ever encountered these troubles? "Manually copying and pasting data is too time-consuming and inefficient"; "You want to collect webpage data in batch, but you don't know how to write code"; "You have tried other crawler tools, but they are all too complicated and costly to learn"; "You are worried that the crawler will be blocked by the website, and you don't know how to deal with it". Don't worry! Today...

You must wonder why Google's AI is called Gemini (Gemini) That's because Google actually has two AIs: a brother called Gemini Aistudio, who is handsome, golden and very capable of fighting, and a younger brother called Gemini App, who is...

Basic Concepts In the field of information technology, retrieval refers to the process of efficiently locating and extracting relevant information from a large dataset (usually documents, Web pages, images, audio, video, or other forms of information) in response to a user's query or need. Its core purpose ...

Agent AI: Surveying the Horizons of Multimodal Interaction Originally published at https://ar5iv.labs.arxiv.org/html/24...

When you open Roo Cline and set the model service provider to OpenRouter, you will see the following settings: He will guide you to understand what OpenRouter Transforms are.Simple explanation: When he encounters a context that is too long to be sent to the model before he will...

Installing Dify Plugins Click "Plugins" in the upper right corner of the Dify platform to go to the plugin management page, which supports three ways of installing plugins: Marketplace, GitHub, and local upload. Marketplace Dify M...

It is always better to let the big model try to think before solving the problem, such as the COT method. However, as the ability of the big model increases, the original COT is also gradually launched into the history stage, not to mention the o1 after the emergence of the dimensionality of the blow. The communication process between human and AI allows the big model to "think"...

Q: How can I get AI to better analyze my structured data (Excel/CSV)? The content is very long it feels like some of it is being thrown away A: AI can't process very long content at once due to the limitations of the context window. In addition, the big language model itself is not good at math operations, so it...

I think the probability is that the pictures, tables, flowcharts. Today we recommend a free AI text to visual tool, the texture is a bit cow, it feels like Figma such a big factory texture, simple and powerful. Of course, it can also be a very beautiful and practical notes, documents tool...

GraphReader: a graph-based intelligence that enhances long text processing for large language models Graphic Expert: like a tutor who is good at making mind maps, it transforms lengthy text into a clear knowledge network, so that the AI can easily find each level needed for an answer as if it were exploring along a map...

Many people would like to use WeChat's voice input directly, it's always faster to speak than to type. Unlike the common .mp3 and .wav formats, WeChat's voice input uses the .amr format by default. The image below shows the webhook received by the developer server from WeChat, indicating that the public...

Someone in the group asked: which senior knows which ai can draw the flowchart of listed company information? Guess it is listing counseling documents related to the process, in fact, do not need any tools, as long as you can draw a sample flowchart, so that the big model to generate SVG code can be, of course, Mermaid syntax can be...

Recently, I have also been combing through some content about AI hardware, and the overall feeling of AI hardware, there is a little bit of the flavor of the 2023 mid-year big model, a hundred model war, all innovation, everything is thriving. What direction is a bit of fun, of course, the most advanced or AI glasses and AI toys, as well as bracelet watch...

Today I'm sharing a super cool AI programming open source project - Bolt.DIY! Free to deploy! Free to use! Totally free! Interested partners please like the collection, go try it! Bolt.new is an AI-driven full-stack development platform, through natural language description on...

The development of the work of advertising so far, I do not know if you have ever considered, the future of advertising between the effect of the competitive differences will be reflected in the where, or we should focus on the improvement of the ability of self placement in which plate is more valuable? Is it the ability to optimize the fine-tuning of the placement account?...

JSONArray Basic Concepts JSONArray is an array structure for storing multiple JSON objects. Each element can be a JSON object, an array, or a basic data type (e.g. string, number, etc.). In Dify, using J...

Voyage AI's Voyager 3 is a new state-of-the-art model that allows you to embed text and images into the same space. In this post, I will explain how to extract these multimodal embeddings from magazines, store them in a vector database (Weaviate...

Previously, SiliconCloud went online with the online LoRA fine-tuning feature for language models. By simply uploading corpus data and creating a fine-tuning task, you can get an exclusive fine-tuned language model. Recently, SiliconCloud's LLM online LoRA...

CAG (Cache Augmented Generation) that is 40 times faster than RAG (Retrieval Augmented Generation).CAG revolutionizes knowledge acquisition: instead of retrieving external data in real time, all knowledge is pre-loaded into the model context. It's like condensing a huge library into an on-the-go work...

My best AI programming partners at the moment are Lovable and Cursor. bolt.new and windsurf, which are also very good, I chose the first two because the ceiling is high enough. Lovable's website: https://lova...

Do you often need to transcribe meeting recordings or interviews into text? Since writing verbatim scripts is time-consuming and laborious, it's a good idea to use AI tools to convert audio recordings into text. In this article, we will introduce Whis...

Prompt Words Enter the content to be converted here When I give you a text in English (e.g., a report from The Economist or WSJ), please provide me with a translation and paraphrase according to the following requirements: Translation Requirements: Translate the English text into smooth and natural Chinese. ...

Suitable as Cursor, Windsurf, Cline and other AI IDE tools to normalize the generation of front-end project code. Such tools, although the ability to generate complete project code is very powerful, but the lack of basic constraints will lead to a large number of ineffective consumption of tokens, especially raw...

Demonstration Effect Demo: Using the built workflow, we insert a record in the Flying Book multidimensional table The following step-by-step exploration of the specific implementation process. Requirements Analysis In our daily work, we often collect some information to facilitate the later organization and view. But manually organizing the need to fill in a field by field ...

Introduction There was a time when creating a comic book was a tedious process that required writers, illustrators, and countless hours of effort. Today, artificial intelligence serves as a powerful tool to empower creative professionals. Imagine handing over a short story to AI and watching how it helps bring that story...

Trae's Builder mode helps you develop a complete project from 0 to 1. You can seamlessly integrate it into the project building process. In Builder Mode, the AI Assistant calls on different tools as needed when answering questions, including analyzing...

A very good application scenario for AI images and videos is the WeChat Red Envelope Cover, where your single particularly good image or video subplot can also create value. There has been a business of selling red envelope covers since WeChat introduced them, but it turns out that they are not officially supported, and even counted as a violation if reported...

Using the Gemini family of models you can quickly test the API by running the cURL command: curl "https://generativelanguage.googleapis.com/v1beta/models...

In today's fast-changing artificial intelligence technology, AI programming tools are springing up, and many of them are as powerful as Cursor. However, for those "novice" users who have no programming foundation, even for tools like Cursor, there are still some...

By Julia Wiesinger, Patrick Marlow, and Vladimir Vuskovic Originally published at https://www.kaggle.com/whitepaper-agents The purpose...

Retrieval Augmented Generation (RAG) is becoming one of the most popular applications of Large Language Models (LLMs) and vector databases.RAG is performed by retrieving vector databases (e.g., Weaviate) from...

When dialoguing with an AI assistant, you can specify the contents of Trae (including code, files, folders, and workspaces) as context for the AI assistant to read and understand, thus ensuring that the AI assistant's responses are more in line with your requirements. This article describes three ways to specify context. ...

Recently, I took over a project that needs to use Stable Diffusion, and I need to redeploy a set of SD environment. This is not quite the same as my previous SD deployment, the deployment process encountered some problems, summed up a more perfect installation program, here to share with you a ...

Winter is here, has it snowed at home yet? It doesn't matter if it hasn't, it is now - click here How it's done A: Through GLM-Zero, which is what Smart Spectrum posted a couple days ago. It looks like a Smart Spectrum ad... Also recommended is to try DeepSeek Chat's "Deep...

Kohya has released a Lora training script for the Hunyuan Video model. It will allow 16GB VRAM cards to train video Lora using only image inputs. And it supports image training for video models lora Civit...

A Multi-Agent System (MAS) is a computing system consisting of multiple interacting Intelligent Agents. A MAS can be used to solve problems that are difficult to solve by a single intelligent agent or a single system or...

First, LLMs already have strong capabilities, why do we still need RAG (Retrieval Augmented Generation)? Although LLMs have demonstrated significant capabilities, the following challenges still deserve attention: Illusion problem: LLMs use a statistically based probabilistic approach to generate text word-by-word, a machine...

o3 is here to share some personal insights. Progress on Test-time Scaling Law has been much faster than we thought. But I'd say the path is actually a bit convoluted - it's OpenAI's path in pursuit of AGI, taking...

Prompt design with Lovable, a list of strategies and approaches. To help you get the most out of Lovable, we've compiled a list of prompt design strategies and approaches. These strategies are partially derived from our team's experience and partially shared by community members. What is a cue...

With the continuous progress of artificial intelligence technology, there is an increasing demand for building personalized intelligent bodies. Although there are already some intelligent body platforms in China, such as Doubao Buckle, these platforms require developers to upload their code and data to third-party servers, which may bring about for some customers with sensitive business information...

Vector Embedding is the core of current Retrieval Augmented Generation (RAG) applications. They capture semantic information of data objects (e.g., text, images, etc.) and represent them as arrays of numbers. In current generative AI applications, these vector Embedding are usually composed of Embedd...

Writing Project Documentation Prompt Words Original You are my software architect. you will help me write down specific user stories and **f...

Preface Over the past two years, Retrieval-Augmented Generation (RAG, Retrieval-Augmented Generation) technology has emerged as a central component of enhancing intelligences. By combining the dual capabilities of retrieval and generation, RAG is able to bring in external knowledge to...

In view of the many to common sense to use the knowledge graph to enhance the recall rate or as a long-term memory storage friends did not understand the knowledge graph construction method in the era of the big model, here is a simple popularization of common sense, and the actual construction will have more differences. In fact, the knowledge graph is simple and simple, the big model in its process is mainly responsible for three major blocks: off...

LangChain Hub #1 tipster in Chinese. Released over a year ago and used in the overall evaluation of the combined scores of different RAG strategies. Translated and adapted for use in multiple languages. Usage Help Evaluate which answer is better, assuming both answers are correct. Take ...

A directive is a node of a complete retrieval scheme that provides accurate routing for questions. Features: 1. Rewrite questions 2. Route different knowledge bases for questions 3. Generate questions sorted by trustworthiness Example of a core directive (adapted to your own business scenario) Simplify up-to-date questions to facilitate in the intent to know...

There is no easy answer to the price comparison between OpenRouter and Poe, as their prices depend on the models used, the amount of usage, and the specific package. Both offer a wide range of AI models at widely varying prices. Poe: Poe's strength is its all-in-one platform...

Recraft is an up-and-coming image generation program that is particularly popular with Dreamcore style enthusiasts. It allows users to generate images with a film sense by entering keywords, the operation is simple and fast. Recommended tutorial: 5 minutes to teach you to use AI dreamcore film avatar, don't...

Windsurf is an AI-based development tool that supports a variety of development frameworks and languages, including the development of WeChat Small Programs. According to the latest information, Windsurf can help developers quickly build WeChat Small Program projects and generate code and project closure through natural language commands...

"Fixing Errors" Tip AI models like Sonnet 3.5 can sometimes miss important details, which can trigger a series of error loops. Use the following tips to solve this problem. This will help the AI analyze the root cause of the error and then create a step-by-step plan to fix it...

Transit large model API in the country is an essential service, such as the most well-known One API, in addition to this will be commonly used cloudflare vercel huggingface to solve the API proxy load and security issues. Even to ensure that the proxy service ...