Tag: ai
141 articles
LLM Application Engineering: Key Practices from Demo to P...
Core experience moving LLM applications from prototype to production: context management, error handling, cost control, observability. No basics, just real pitfalls.
Real-time Voice Interaction Pipeline Latency Optimization...
Documenting the process of building ASR→LLM→TTS real-time voice pipeline: why latency is high, how pipeline concurrency reduces first-byte latency, VAD endpoint detection pitfalls, and practical co...
AI Research #135: Gemini 3 Pro Back on Top - MoE, Million...
Explains Gemini 3 Pro's advantages through sparse MoE architecture, million-token context, native multimodal (text/image/video/PDF), thinking depth control (thinking_level), and Deep Think mode. St...
AI Research #134: Java 2025 - Will It Decline?
In 2025, Java remains the enterprise backend and critical industry workhorse. Key keywords: Java 25 LTS, Java 21 LTS, Spring Boot, MicroProfile, Kubernetes, Serverless, Project Loom, GraalVM Native...
AI Research #131: Java 17/21/25 Complete Comparison
Java 17 (2021), Java 21 (2023), Java 25 (2025) language and JVM changes, covering Virtual Threads (Project Loom), Records/Pattern Matching (Project Amber),...
AI Research #130: Qwen2.5-Omni Practical Applications
Office assistant, education and training, programming and operations, search-enhanced RAG, device control/plugin agents, and companion entertainment. Covers...
AI Research #129: Qwen2.5-Omni-7B Key Specs - VRAM, Conte...
Runs stably at FP16 ~14GB VRAM, with INT8/INT4 quantization (<4GB) enabling deployment on consumer GPUs or edge devices. Combined with FlashAttention 2 and...
AI Research #128: Qwen2.5-Omni Training Pipeline - Three-...
Complete training pipeline breakdown for Qwen2.5-Omni: Thinker based on Qwen2.5, vision initialized from Qwen2.5-VL, audio from Whisper-large-v3. Uses...
AI Research #127: Qwen2.5-Omni Deep Dive - Thinker-Talker...
Engineering breakdown of Qwen2.5-Omni (2024-2025) Thinker-Talker dual-core architecture: unified Transformer decoder for text/image/video/audio fusion, TMRoPE...
AI Research #123: FSD V14 Deep Analysis - Vision-Only SDF...
FSD V14 (2025) technical evolution compared to V12 (2023), focusing on vision-only approach, SDF (Signed Distance Field) occupancy reconstruction, end-to-end...
AI Research #121: DeepSeek-OCR Research Directions
Frontier approaches and engineering implementation for DeepSeek-OCR (2025, including 3B parameter direction). Summarizes research directions including...
AI Research #119: DeepSeek-OCR PyTorch FlashAttn 2.7.3 In...
Comprehensive guide for DeepSeek-OCR local/private deployment based on Python 3.12, PyTorch 2.6.0, Transformers 4.46.3 and FlashAttention 2.7.3. Includes ~3B parameter model inference, deployment o...
AI Research #120: DeepSeek-OCR from 0 to 1 - Getting Star...
Complete getting started path and engineering essentials for DeepSeek-OCR (as of 2025), covering environment setup (Python/PyTorch 2.x, Transformers 4.x), model loading, output parsing, parameter e...
AI Investigation #107: RL and Robot Training Data Format ...
Data formats and development processes in robot and reinforcement learning systems, including time series trajectories, state-action pairs, offline RL data,...
AI Investigation #106: Robot Learning Data Collection Too...
Core data collection methods and application scenarios, covering over ten methods from manual entry, sensor collection, web crawlers, API calls, log collection...
AI Investigation #105: Robot Learning Data Collection - F...
Data collection is a critical step in robot learning development, covering demonstration video collection, trajectory recording, state-action pair generation, and data quality control strategies.
AI Investigation #104: From Model Training to Robot Deplo...
AI model deployment optimization guide: ONNX conversion, TensorRT/OpenVINO inference engines, quantization (FP16/INT8), and real-time robotics applications.
AI Investigation #103: Embodied AI Technology Landscape
Comprehensive overview of embodied AI tech stack: hardware (GPU, sensors, actuators), software (ROS, simulation), and algorithms (deep learning, RL, VLA models).
AI Investigation #102: Intelligent Robotic Arms, Autonomo...
Different types of robots have huge differences in structure, tasks and control methods, so AI algorithm adaptation strategies also need to be tailored.
AI Investigation #101: Modern AI Methods - VLA, RT-1, RT-...
Modern AI robot control methods are undergoing a major transition from reinforcement learning and imitation learning to multimodal agents driven by large models. The combination of Vision-Language-...
AI Investigation #99: Sensor Fusion Technology - Camera, ...
Sensor Fusion is a core technology in autonomous driving, robotics and smart security. Through multi-sensor data fusion of cameras, LiDAR, radar, IMU,...
AI Investigation #98: Visual SLAM - ORB-SLAM, RTAB-Map, V...
Visual SLAM is a technology that achieves autonomous positioning and environment mapping without relying on LiDAR, using only cameras. By extracting environmental features (corners, edges, textures...
AI Investigation #96: Robot Scenario Testing - From Extre...
Complete guide to robot scenario testing, covering three dimensions: environment testing, load testing, and anomaly testing. Traditional manual testing has...
AI Investigation #95: Robot Scenario Testing - From Extre...
Before robots enter practical applications, systematic scenario testing must be conducted, covering boundary conditions like extreme weather, complex terrain,...
AI Investigation #93: Robot Simulation Tools - Comprehens...
Simulation tools are an important part of robot R&D, enabling algorithm verification and system debugging in risk-free environments, accelerating iteration.
AI Investigation #92: Robot Motion Control - From Traditi...
Robot motion control can be divided into two categories: traditional model-based methods and deep learning-based intelligent control. The former emphasizes kinematics/dynamics modeling, trajectory ...
AI Investigation #91: Multi-modal Data Annotation Tools -...
In robot vision and perception model training, high-quality multi-modal data annotation tools are crucial. Current mainstream solutions cover 2D images, video, text, audio and 3D point cloud multi-...
AI Investigation #90: Robot Data Collection and Communica...
Modern robot systems require efficient data collection and communication middleware to connect sensors, controllers and computing units, achieving collaborative perception, control and decision-mak...
AI Investigation #89: Multi-dimensional Autonomic Nervous...
Individual differences in autonomic nervous system (ANS) profoundly affect disease susceptibility. Sympathetic-dominant people are prone to hypertension and...
AI Investigation #88: HRV and Autonomic Nervous System - ...
Heart rate variability (HRV) is an important indicator for assessing autonomic nervous system balance, reflecting the dynamic interaction between sympathetic...
AI Investigation #87: HRV Time Domain and Frequency Domai...
HRV calculation methods mainly include time domain and frequency domain. Time domain indicators assess autonomic nervous function by statistically analyzing...
AI Investigation #85: Autonomic Nervous System Assessment...
The autonomic nervous system consists of sympathetic and parasympathetic nerves, its functional state directly affects cardiovascular, respiratory and...
AI Investigation #84: Fat Loss Science - Low Body Fat Mai...
Maintaining low body fat is more challenging than losing it. The human body is naturally inclined to store energy. Long-term maintenance of around 10% body fat triggers metabolic adaptation: decrea...
AI Investigation #83: Fat Loss Science - Complete Diet Guide
During fat loss, intake is recommended at 1900-2100 kcal, with daily deficit of 300-500 kcal. Protein intake 1.6-2.2g per kg body weight, carbohydrates divided into fast carbs before/after training...
AI Investigation #82: Fat Loss Science - Fitness Guide fo...
For lean people, the key to fat loss and body shaping is 'muscle building primary, cardio auxiliary'. Simple fat loss can make the body appear thin, while strength training can increase muscle circ...
AI Investigation #81: Fat Loss Science - What Body Fat Pe...
Body fat percentage affects both appearance and health directly. Healthy body fat for men is 10%~20%, ideal around 15%; for women it's between 18%~28%. Asian standards are stricter than Western sta...
AI Investigation #80: Why Fat Loss Must Be Science-Based ...
The core mechanism of fat loss is energy deficit: when calorie intake is less than consumption, the body uses fat stores for energy. Total daily energy...
AI Investigation #78: LFP Lithium Battery - Shallow Charg...
Shallow charge frequent and deep discharge charge have significantly different impacts on LFP battery life. Research shows that shallow cycling can effectively...
AI Investigation #79: LFP Battery Degradation Patterns Un...
The widespread use of LFP batteries in Tesla and other models has accumulated large amounts of real-world lifetime data. Research shows that capacity retention rate is generally above 95% for the f...
AI Investigation #77: LFP Lithium Iron Phosphate vs Terna...
Lithium Iron Phosphate (LFP) batteries, with their stable olivine crystal structure, significantly outperform ternary lithium (NMC/NCA) batteries in cycle life and safety. LFP volume change during ...
AI Investigation #76: When Robots Enter Life - Embodied A...
The widespread application of embodied AI is profoundly changing social structure. Automation will replace many repetitive jobs while creating new occupations, leading to job polarization in the em...
AI Investigation #75: From LLM to LBM - Robot Hierarchica...
The integration of Large Language Models (LLM) with robot real-time control is driving intelligent upgrades in robotics. LLMs show great potential in...
AI Investigation #74: Robot Learning Breakthroughs - Meta...
This article explores fast learning capabilities of embodied AI agents, focusing on meta-learning and few-shot learning methods, as well as technical challenges and solutions for simulation-to-real...
AI Investigation #73: Embodied AI Future Trends - From Te...
In the next decade, embodied AI will undergo paradigm shifts: centered on 'pre-trained world models + online learning', software-hardware collaboration and interdisciplinary fusion accelerating, di...
AI Investigation #72: Embodied AI Development Challenges ...
Embodied AI development faces six core challenges: data scarcity, hardware limitations, training efficiency, cost bottlenecks, standardization and industrial ecosystem, safety and ethics. Only by a...
AI Investigation #71: Embodied AI Case Studies - From ROS...
Typical practices of embodied AI in architecture, capabilities and applications. In open source, ROS robot operating system has become the de facto industry...
AI Investigation #70: Embodied AI Industry Ecology and De...
Embodied AI is leading a new round of technological revolution. The market size is expected to grow from $2.53 billion in 2024 to $8.76 billion in 2033, with a CAGR of 15%.
AI Investigation #69: Embodied AI Key Capabilities - Algo...
Systematic overview of five key capability dimensions of embodied AI: intelligent algorithms, high-performance hardware, simulation and virtual environments, embedded and software systems, and data...
AI Investigation #68: Embodied AI Application Landscape -...
Systematically explore the application prospects and development trends of embodied AI in multiple fields, covering core scenarios including home, industry, healthcare, transportation and virtual i...
AI Investigation #67: Embodied AI Core Technology - Perce...
Embodied AI technology can be summarized as a 'perception-decision-control-learning-interaction' closed-loop system. The perception layer consists of various sensors, providing environment modeling...
AI Investigation #66: Robotic Arm Software Algorithm Syst...
Robotic arm software algorithm system covers core technologies including kinematics, trajectory planning, robot vision, artificial intelligence and machine learning, force control and compliance st...
AI Investigation #65: Robotic Arm Control Technology Evol...
Implementation of robotic arm systems requires software-hardware collaboration, with strict requirements from control layer to safety mechanisms. Control systems have evolved from PLC to MPC.
AI Investigation #64: Building a Robotic Arm from Scratch...
Building and controlling a high-performance robotic arm requires integrating multiple core modules including motor drive, reducer, mechanical structure, sensor system, controller and end effector.
AI Investigation #63: Full-scenario Applications of Robot...
Modern robotic arms undertake various key tasks in industrial automation: handling, assembly, welding, spraying, grinding, unstructured grasping, machining, inspection and testing.
AI Investigation #62: Five Major Robotic Arm Application ...
Robotic arms, as important tools for modern automation, have been widely applied in manufacturing, healthcare, agriculture, services and scientific research.
AI Investigation #61: Complete Industrial Robot Types Ana...
Complete industrial robot spectrum: articulated robotic arms suitable for complex trajectory tasks like automotive manufacturing; SCARA for high-speed planar...
AI Investigation #60: Robotic Arm Technology Development ...
Robotic arm technology development shows five major trends: high precision, collaboration, lightweight, intelligence, and connectivity. Industrial, service and DIY segments each have their own char...
AI Investigation #86: Balance of Autonomic Nervous System...
Systematically expounds the balance mechanism of sympathetic and parasympathetic nerves in the human autonomic nervous system, deeply analyzing their physiological effects, pros and cons in differe...
AI Investigation #59: Robotics Career Map - Development P...
The robotics industry is experiencing rapid development, with surging demand for interdisciplinary talent with mechanical, electronic, control and software capabilities. Career path starts from ass...
AI Investigation #58: Robotics - From Factories to Homes,...
As a general-purpose technology, robots are widely used in manufacturing, healthcare, agriculture, logistics and home sectors. The industry shows three major trends: intelligence, networking, and c...
AI Investigation #57: Five Major Categories of Robots - I...
Robots can be divided into five major categories based on function and form: industrial robots, service robots, humanoid robots, mobile robots, and specialty robots.
AI Investigation #56: Robotics Technology Iteration - Evo...
Robotics has undergone profound evolution from early hydraulic drive and analog control to modern electric drive, digital control and perception systems.
AI Investigation #55: Robotics - A Century of Evolution f...
Since the term 'Robot' was first introduced in 1921, robotics has undergone a century of evolution from science fiction to reality. Starting with Unimate in 1959 pioneering the industrial robot era...
AI Investigation #54: Big Data Industry Applications and ...
Big data has achieved deep integration in finance, e-commerce, internet, communications, manufacturing, healthcare, education and other industries, becoming the core engine for business innovation.
AI Investigation #53: Big Data Talent Landscape - Experie...
The talent structure in the big data industry shows characteristics of youth and rapid growth. The 25-30 age group is the main force, while 30-35 year-olds are gradually becoming the core strength.
AI Investigation #52: Big Data Technology Landscape - Lak...
Big data technology is undergoing a new wave of transformation. Lakehouse architecture combines the advantages of data lakes and data warehouses. Data Mesh...
AI Investigation #51: Big Data Technology Evolution - Obs...
Big data technology evolution: MapReduce replaced by Spark, Storm replaced by Flink, Pig/Hive gradually phased out. This article analyzes why these technologies were eliminated and the technical re...
AI Investigation #50: Big Data Evolution - Two Decades of...
Two decades of big data evolution: from 2006 MapReduce batch processing to 2013 Spark in-memory computing, to 2019 Flink real-time computing. Architecture evolved from monolithic Hadoop to YARN mul...
AI Research 49 - Big Data Survey Report: Development Hist...
Big data development began in 1997 when NASA proposed the concept, 2003-2006 Google published GFS, MapReduce, Bigtable three major papers leading distributed computing revolution. 2005 saw Hadoop b...
AI Research 01 - Is Mindfulness Meditation Effective? Hea...
Mindfulness meditation is a scientifically validated mind-body regulation method. Research shows it can reduce anxiety, depression, and stress levels, improve sleep quality, enhance immune function...
AI Research 48 - Traditional Chinese and Western Medicine...
Compare differences and advantages of Chinese and Western medicine in treating common diseases and chronic diseases, explore application prospects of...
AI Research 47 - Multi-Dimensional Survey Report: Systema...
Extra nutrient supplementation should be based on individual assessment, not blind following. If daily diet and multivitamin supplementation are still insufficient, or there are specific physiologi...
AI Research 46 - Multi-Dimensional Survey Report: Multivi...
Multiple large-scale clinical studies show that for ordinary healthy people, long-term intake of multivitamin supplements has limited overall effect in preventing cardiovascular disease, cancer, di...
AI Research 45 - Multi-Dimensional Survey Report: Key Dif...
Explore main differences in nutritional supplementation between men and women, focusing on different needs for key nutrients like iron, calcium, and folate.
AI Research 44 - Multi-Dimensional Survey Report: Should ...
Survey research on whether nutritional supplements like fish oil, calcium tablets, and vitamin C need extra intake. Analyze applicable scenarios and precautions for each nutritional supplement.
AI Research 43 - Multi-Dimensional Survey Report: Should ...
Multivitamin supplements can fill nutritional gaps caused by unbalanced diet, suitable for elderly, vegetarians, pregnant women and other specific groups.
AI Research 42 - Multimodal Large Model Quantization: Fro...
Survey outline for multimodal large model quantization schemes: from FP32 to INT4. Core goal is model capability retention, compression efficiency 50-75%, inference speedup 2-4x. Analyze comparison...
AI Research 41 - Multimodal Large Model Quantization: Qwe...
Qwen2.5-VL is the new generation multimodal large model launched by Alibaba, significantly leading in visual understanding, video analysis, and cross-modal reasoning. Provides multiple versions fro...
AI Research 40 - Multimodal Large Model Quantization: Pat...
Multimodal large models are developing rapidly, with representative models like BLIP-2, MiniGPT-4, Flamingo, LLaVA, and Qwen2.5-VL emerging. Analyze each model's architectural innovations, performa...
AI Research 39 - Multimodal Large Model Quantization: How...
In multimodal large model optimization, the order choice of fine-tuning and quantization directly affects the final model's performance and efficiency. There are three main strategies: fine-tune fi...
AI Research 38 - Multimodal Large Model Quantization: Ana...
To systematically evaluate the impact of model quantization on performance, need to combine multiple vision-language datasets and metrics. Commonly used datasets include Flickr30k and MS COCO, usin...
AI Research 37 - Multimodal Large Model Quantization: Imp...
Model quantization compresses FP32 weights into low-precision representations, significantly reducing inference resource consumption. Experiments show quantized models have 60% lower latency and 70...
AI Research 36 - Comprehensive Analysis of Multimodal Lar...
This comprehensive overview systematically introduces mainstream quantization techniques in multimodal models, including principles and practices of...
AI Research 35 - Coffee Price War: Taste and Consumption ...
The taste experience differences between homemade coffee and chain budget coffee stem from multiple factors including ingredient quality, production process, and psychological feelings. Analyze cof...
AI Research 34 - Coffee Price War: User Preferences and B...
China's chain coffee market is experiencing beverage-ization change—coffee products increasingly becoming milk-tea-like. Brands like Luckin and Starbucks innovate with milk foam, syrups, and toppin...
AI Research 33 - Coffee Price War: Comprehensive Analysis...
Brands like Luckin and Cotti use subsidies to gain market share, driving user growth, but also face risks of cost inversion and profit pressure. Analyze price war review, cost structure, and brand ...
AI Research 32 - Coffee Price War: 9.9 Yuan Era - Can Cof...
China's brewed coffee market is shifting from space experience to price+efficiency driven, with local brands like Luckin, Cotti, and Lucky Cup accelerating penetration, and market structure being d...
AI Research 31 - Programmers Don't Have Bad Memory, They ...
8 evidence-based strategies for programmers to combat forgetting: Ebbinghaus forgetting curve can be quantified, active recall is better than passive review,...
AI Research 30 - How Programmers Can Scientifically Comba...
Programmers often experience 'forgetting after learning'. Effective strategies include: note organization, spaced repetition (Anki), active recall, project practice, and output sharing. Core perspe...
AI Research 29 - Why Does the 'Perfect Partner' During Da...
Dating focuses on emotional attraction, marriage values responsibility, communication, and values more. Differences cause conflicts such as expectation gaps,...
AI Research 28 - Comprehensive Comparison of Dating Persp...
Dating perspective differences between China and US: first love age is approximately 17-18 in China, 15-17 in US. Number of relationships before marriage: 2-3 in China, 4-6 in US. Mate preferences:...
AI Research 27 - [Time Management] Heavy vs Light Users: ...
Heavy social media users have higher rates of depression, anxiety, and loneliness, plus more sleep problems. Time after reducing usage can be reallocated to exercise, reading, family, hobbies, and ...
AI Research 26 - [Time Management] Social Media Usage Tim...
Global netizens spend an average of 2 hours and 21 minutes on social media daily. Chinese users spend 1.55 hours on Douyin, with short video usage peaking at midnight. Young women aged 16-24 are th...
AI Research 25 - [Money Management] Savings vs Spending: ...
Should young people save more or spend more? Experts recommend 'save first, then spend', following the 50/30/20 rule: 50% needs, 30% wants, 20% savings. Build 3-6 months emergency fund.
AI Research 24 - [Money Management] Survey of Young Peopl...
Comparison of savings levels, consumption structure, and financial concepts among young people in China, the US, and Japan. Nearly 30% of Chinese are 'monthly spenders', and 61.1% pay more attentio...
AI Research 23 - [Time Management] Strategies and Metrics...
Time management can significantly improve efficiency and sense of achievement. Evaluating time management effectiveness requires objective indicators (task completion rate, goal achievement) and su...
AI Research 22 - [Time Management] Reasonable Planning an...
More than 80% of people do not have a formal time management system. Building a time management solution requires: clear goals → choose tool combinations → design time templates → establish task cl...
AI Research 21 - [Time Management] Necessity and Practica...
Effective time management can improve efficiency, reduce stress, and improve work-life balance. Scientific evidence includes psychology, behavioral science, cognitive science, and more.
AI Research 20 - [Mindfulness Meditation] Effects on Heal...
Mindfulness meditation is a scientifically validated mind-body practice. Research shows it can significantly reduce anxiety, depression, and stress hormone levels, and improve sleep quality and cog...
AI Research 19 - Conflict Analysis Between Programmers an...
Reducing conflicts between programmers and product managers requires: improving requirements process, enhancing communication skills, aligning goals, agile practices, managing expectations, senior ...
AI Research 18 - Conflict Analysis Between Programmers an...
Resolving conflicts between programmers and product managers requires: establishing clear communication mechanisms, improving document quality, fairly negotiating priorities, building mutual unders...
AI Research 17 - Conflict Analysis Between Programmers an...
Conflicts between programmers and product managers show a 'bell curve' distribution. Influencing factors include project scale, team experience, company type, work mode, and management systems.
AI Research 16 - Conflict Analysis Between Programmers an...
Conflicts between programmers and product managers most frequently occur in four key stages: requirements analysis, development implementation, schedule planning, and testing acceptance. Main confl...
AI Research 13 - LLM and Agent Research: The Rise and Dev...
2024 is called the 'Year of Agents'. LLM trends show parallel development of 'bigger and stronger' and 'smaller and more specialized'. OpenAI o1 series, Claude, and other multimodal models continue...
AI Research 15 - Methodology, Validation Process, and Val...
Building products from 0 to 1 requires mastering methodologies like Lean Startup and Design Thinking, and validating directions through requirement verification, user interviews, and competitive an...
AI Research 14 - How to Write Excellent Research Reports,...
How to write excellent research reports, slide presentations, and technical sharing sessions? Need to build a systematic knowledge system, follow the IMRaD structure, and focus on scientifically ev...
AI Research 12 - LLM and Agent Research: Overview of Majo...
Major LLM application directions in 2024-2025 include enterprise applications (code assistance, customer service, knowledge management) and consumer applications (general conversation, content crea...
AI Research 11 - Running Analysis Research: Nutrition Str...
Running nutrition strategies should focus on carbohydrates (50-65% of daily energy), protein (1.2-1.6g/kg), and fat (20-30%). Consume 15-25g of protein within 30 minutes after training.
AI Research 09 - Running Analysis and Research: Optimal R...
The optimal way to run is to keep heart rate between 60% and 80% of maximum heart rate. Post-run nutrition should include carbohydrates and protein promptly.
AI Research 07 - Running Benefits: Effects of Different R...
Running has significant benefits for cardiovascular health, weight management, and psychological state. Different running distances (3km, 5km, 10km, half marathon, full marathon) each have their ow...
AI Research 06 - The Physiological Health Effects and Mec...
As a health intervention, cold showers can provide benefits such as refreshing, immune enhancement, and recovery promotion when practiced scientifically. It is recommended to transition from warm w...
AI Research 05 - The Physiological Health Effects and Mec...
Cold showers have positive effects on mental health, including improved alertness, stress and depression relief, and enhanced psychological resilience. Hot showers help with relaxation and sleep, e...
AI Research 04 - The Physiological Health Effects and Mec...
Cold showers typically refer to showering with water temperature at or below 20°C, with positive effects on immune function, blood circulation, metabolic rate, inflammation, and muscle soreness.
AI Research 03 - Does Technical Time Investment Correlate...
Reasons for stagnant income despite accumulated technical capabilities include position saturation, increased talent supply, industry cycle impacts,...
AI Research 02 - Does Technical Time Investment Correlate...
Technical investment and income are generally positively correlated, but returns do not grow infinitely. Most programmers receive their largest salary...
LangChain-26 Custom Agent Complete Tutorial Building a Cu...
This article demonstrates how to create a chat agent using the Langchain library and GPT-4 model in Python by defining tool functions and integrating them with LLM to achieve queries for informatio...
LangChain-24 AgentExecutor Comprehensive Guide
This article introduces how to use the Langchain library in Python for document retrieval, load web content, configure OpenAIEmbeddings, and integrate GPT-3.5-turbo model for Q&A. It demonstrates h...
LangChain-25 ReAct Framework Detailed Explanation Integra...
This article introduces ReAct, a framework that uses logical reasoning and action sequences to achieve goal-oriented tasks through LLM decision-making and operations. The core components include Th...
LangChain-22 Text Embedding and FAISS Practical Explanation
This article introduces the key role of TextEmbedding in NLP, how to convert text into real number vectors to represent semantic relationships, and how to combine OpenAIEmbeddings and FAISS for eff...
LangChain-23 Vector AI Semantic Search System Vector Data...
This article introduces how to use Chroma vector database to process and retrieve high-dimensional vector embeddings from documents, vectorize them using...
LangChain-20 Document Loaders TextLoader, CSVLoader, PyPD...
This article introduces various document loaders provided by the LangChain library, such as TextLoader, CSVLoader, DirectoryLoader, etc., demonstrating how to load and process data in various formats.
LangChain Text Splitter: Character, Word, HTML and Code-b...
This article introduces various TextSplitters in the LangChain library, including character-based, word-based, HTML tag-based, and programming language-based splitters, as well as their application...
LangChain Cache Mechanism: InMemoryCache and SQLiteCache ...
LangChain provides a comprehensive caching mechanism to significantly reduce LLM call latency and costs. Its core includes InMemoryCache (in-memory cache) and SQLiteCache (persistent cache).
LangChain-19 TokenUsage Callback Function Explained
Explains how to integrate OpenAI GPT-3 model in Python through LangChain library, demonstrating how to use the `get_openai_callback` function to obtain callbacks and execute requests.
LangChain-16 Using Tools: Mastering LLM Tool Calling
LangChain is currently one of the most popular LLM application development frameworks, specifically designed for building intelligent assistants, automation...
LangChain-17 Function Calling AI Function Calling Explained
Function Calling is a core technology for Large Language Models (like GPT-4, Claude, Gemini) to interact with external systems. It enables AI to not only understand language but also execute tasks,...
LangChain-14 OpenAI Content Moderation (Moderation) Expla...
Content moderation is a core component of modern internet platform safety and compliance, used to identify, filter, and manage user-generated content (UGC) to prevent the spread of illegal, low-qua...
LangChain-15 Intelligent Knowledge Retrieval: AgentExecut...
Build an intelligent knowledge retrieval system using Wikipedia search plugin, AgentExecutor, and LangChain tools. Covers agent initialization, tool binding, and multi-step reasoning workflows.
LangChain-12 Routing By Semantic Similarity
This article introduces a method using large models (like OpenAI) and Prompt templates to handle unexpected inputs in program design by calculating the similarity between queries and preset templates.
LangChain-13 Memory ConversationBufferMemory: Conversatio...
This article introduces how to use tools in the LangChain library to manage conversation context of large models in Python. Through components like...
LangChain-11 Code Writing FunctionCalling: Autoregressive...
This article introduces how to use the GPT-3.5-Turbo model to write Python code to solve users' abstract calculation problems, such as 2+2 and complex mathematical expressions, demonstrating the mo...
LangChain 09 - Query SQL DB with RUN GPT
This article introduces how to use Python libraries like langchain and ChatOpenAI (GPT-3.5-turbo) combined with SQLite database to create a program to execute SQL queries and return results in natu...
LangChain 10 - Agents Langchainhub Guide
This article introduces how to use LangChainHub's Hub mechanism through Python code to easily access and share Prompts. Although the project hasn't been...
LangChain 07 - Multiple Chains
How to use Runnable and Prompts in LangChain to create chainable conversation flows for multi-stage question answering, with practical examples of sequential and parallel chain composition.
LangChain 08 - Query SQL DB with GPT
This article introduces how to use LangChain framework to import Chinook SQLite database through Python script and use GPT model to execute SQL queries, such as calculating employee count.
LangChain 05 - RAG Enhanced Conversational Retrieval
This article introduces how to use tools in LangChain library, such as OpenAIEmbeddings and ChatModels, combined with document retrieval technology, to create a program that generates answers based...
LangChain 06 - RAG with Source Document
Retrieval-Augmented Generation (RAG) with Source Document is an AI technology framework that combines retrieval with large language model generation. Its core...
LangChain 03 - astream_events Streaming Output with FAISS...
This article introduces how to use DocArrayInMemorySearch to vectorize text data, combined with OpenAIEmbeddings and GPT-3.5 model, to implement relevant information retrieval and answer generation...
LangChain 04 - RAG Retrieval-Augmented Generation
This article explains in detail how to use RAG technology in LangChain, combined with OpenAI's GPT-3.5 model, to improve text generation quality through retrieval and generation. Provides installat...
LangChain 01 - Getting Started: Quick Hello World Guide
This article introduces how to use the LangChain library with OpenAI API and GPT-3.5-turbo model to create a template for generating jokes about specific topics (like cats). The author demonstrates...
LangChain 02 - JsonOutputParser and Streaming JSON Data P...
This article explains how to install and use LangChain and OpenAI API in Python, retrieve specified country and its population data through async functions, and demonstrates the process of progress...