Technical guides for building AI agents
Recently published articles, sorted newest first.
How do you know a prompt change is better, not worse? A systematic guide to canary deployment, quality gates, auto-rollback architecture, and continuous behavioral drift monitoring for agents in production.
Why do LLM agents hallucinate? This article traces root causes and systematically reviews practical mitigation patterns: retrieval augmentation, confidence scoring, multi-agent cross-validation, forced citation backtracking, and observability with UpTrain, Giskard, RagaAI Catalyst, Comet Opik, and NVIDIA Garak.
A systematic comparison of the three categories of agent memory -- working, long-term, and shared -- covering storage media, lifecycle, retrieval methods, typical frameworks, and design patterns, fully addressing agent personalization and multi-agent collaboration engineering.
Build a production-grade observability stack for multi-step agents using OpenTelemetry: OpenLLMetry semantic conventions, hierarchical span correlation, token cost attribution, retrieval quality metrics, and layered alerting.
Topic-based article groups for reading through one area at a time.
Architecture, orchestration, memory, evaluation, and production patterns for agent systems.
Agent Canary Deployment and Production Monitoring: From Prompt A/B Testing to Automatic Rollback
Agent Hallucination Defense: Practical Mitigation Patterns Beyond Guardrails
Retrieval, vector databases, GraphRAG, chunking, and knowledge-graph practice.
Chat UI Platform Selection Guide: Open WebUI vs LobeChat vs LibreChat
Advanced RAG: Chunking Strategies and Retrieval Optimization Trade-offs
IDE agents, CLI coding assistants, review workflows, and developer automation.
AI Coding Assistant Comparison: Aider vs Continue vs Cursor
Complete Local Deployment Guide for AutoGPT
Full timeline with the newest articles first.
How do you know a prompt change is better, not worse? A systematic guide to canary deployment, quality gates, auto-rollback architecture, and continuous behavioral drift monitoring for agents in production.
Why do LLM agents hallucinate? This article traces root causes and systematically reviews practical mitigation patterns: retrieval augmentation, confidence scoring, multi-agent cross-validation, forced citation backtracking, and observability with UpTrain, Giskard, RagaAI Catalyst, Comet Opik, and NVIDIA Garak.
A systematic comparison of the three categories of agent memory -- working, long-term, and shared -- covering storage media, lifecycle, retrieval methods, typical frameworks, and design patterns, fully addressing agent personalization and multi-agent collaboration engineering.
Build a production-grade observability stack for multi-step agents using OpenTelemetry: OpenLLMetry semantic conventions, hierarchical span correlation, token cost attribution, retrieval quality metrics, and layered alerting.
Based on OWASP LLM Top 10 engineering practice, this article systematically explains the seven layers of defense-in-depth for agent prompt injection: input sanitization, instruction isolation, least-privilege, output auditing, guardrails frameworks, continuous red-teaming, and kill switches -- with actionable code and toolchains.
Exploring how small language models are fine-tuned and deployed for agent workloads at the edge, balancing latency, cost, and accuracy for production AI agents.
A systematic guide to seven tool-call fault tolerance patterns: timeout hierarchy, exponential backoff with jitter, circuit breakers, fallback provider chains, recoverable error classification, structured validation, and idempotency keys -- keeping agents stable in unstable real-world environments.
Long-conversation agents fail at context management, not model capability. A systematic comparison of sliding window, retrieval injection, and layered compression strategies with practical decay diagnosis and recovery patterns.
The biggest hidden cost in agent production is not token pricing but redundant calls and model mismatches. A quantifiable cost-control framework covering caching strategies, fallback chains, and routing rules.
A systematic comparison of three mainstream multi-agent collaboration patterns: Supervisor, Swarm, and Graph. Actionable selection criteria, applicable scenarios, typical frameworks, and mixed-use strategies.
A three-way comparison of open-source Chat UI platforms — Open WebUI (142k Stars), LobeChat, and LibreChat — across deployment, RAG, multi-model, user management, and extensibility.
Dify (145k Stars, $30M Pre-A) is the benchmark for open-source LLM app platforms. From Docker deployment to RAG pipelines, agent orchestration and MCP integration — this article takes you end-to-end.
Traditional vector RAG only finds "similar chunks." GraphRAG extracts entities and relationships into a knowledge graph, letting agents understand "who, where, when, and how things connect." Compare Microsoft GraphRAG vs LightRAG approaches.
Voice AI agents are the next frontier. LiveKit (11k Stars, powering ChatGPT's Advanced Voice) offers a complete framework. This article breaks down the pipeline and walks through building production-ready voice agents.
A practical walkthrough of Cline's three deployment modes — CLI, VS Code extension, and Kanban — with MCP integration and human-in-the-loop coding workflow patterns.
A production walkthrough of LiteLLM based on official docs and its 51k-star repo: Virtual Keys, Guardrails, routing, and Postgres-backed cost governance.
A code-grounded walkthrough of OpenManus architecture, tool layer, browser automation, MCP extension, local deployment, and comparison with LangGraph and CrewAI.
Using Ray's distributed runtime to scale agent prototypes from a single machine to horizontally scalable cluster deployments.
A side-by-side comparison of five sandbox technologies, weighing latency, security, and ops cost.
Five-layer defense plus red-team loop, built on five open-source projects you can copy.
Decomposes five open-source deep-research projects across three sub-stages with reusable query-rewriting code.
Four LLM gateways compared, with production patterns for fallback, smart routing, cost observability, and scheduling.
Most teams evaluate agents by checking a few examples. Real evaluation needs layered metrics, non-rotting datasets, and judges that push back. This article provides runnable code patterns and a practical decision framework.
Most agent workflows fail at the orchestration layer, not the model. A practical comparison of DAG, state machine, and visual builder approaches with production-ready code for error handling, human approval gates, and conditional branching.
A deep architectural comparison of seven open-source coding agents across three paradigms — CLI-first, IDE-integrated, and fully autonomous — examining context management, tool access, and autonomy levels to help you pick the right tool for each development scenario.
Breaking down three abstraction layers for browser automation—from raw Playwright to structured extraction—with production patterns, runnable code, and common pitfalls.
Most RAG pipelines fail at retrieval, not generation. This article covers five chunking strategies, hybrid search, reranking pipelines, and a production-ready decision framework.
A deep dive into the four-layer agent memory architecture, with practical code for vector retrieval and memory compression to help you build scalable long-term memory systems.
A systematic guide to the three pillars of agent observability — distributed tracing, metrics monitoring, and automated evaluation — for building production-grade agent monitoring.
A systematic walkthrough of three major attack surfaces in AI agents, with practical code examples for prompt injection defense, tool permission scoping, and output filtering.
Comparing container, WebAssembly, and process-level isolation approaches, with practical code for safely executing agent-generated code.
Build a production-grade MCP server from scratch, covering tool definition, authentication design, and testing strategies to turn any API into an agent-ready tool.
From protocol modeling and server design to permission isolation, this guide shows how to build a stable tool integration layer for AI agents with MCP.
Based on real production experience, this guide explains how to build a closed loop of tracing, evaluation, and cost analytics for AI agents with Langfuse.
Focused on structured outputs, tool calling, and error recovery, this article presents practical PydanticAI patterns for production systems.
A practical breakdown of browser-use strengths and limits in web task automation, with strategies for stable execution and failure recovery.
Production-focused best practices for index design, filtering, reranking, and evaluation when building RAG retrieval layers with Qdrant.
An in-depth guide on how MetaGPT achieves full software development automation through role-playing, including practical guidance for PM, Architect, Engineer collaboration.
A comprehensive comparison of popular open-source vector databases Milvus, Chroma, and Weaviate, helping you choose the best vector database for RAG applications.
Learn how to evaluate RAG systems using Ragas and DeepEval, including measuring key metrics like faithfulness, answer relevance, and context precision.
Learn how to build stateful AI agents with long-term memory using Letta (formerly MemGPT), solving the LLM context window limitation.
A detailed comparison of three popular AI coding assistants - Aider, Continue, and Cursor - to help you choose the best development tool based on features, experience, and pricing.
An in-depth comparison of mainstream AI agent frameworks including LangChain, LangGraph, CrewAI, and AutoGen to help you choose the best development stack.
A hands-on guide to building a complete AI agent from scratch, covering environment setup, core components, and tool integration.
A deep dive into principles, architecture patterns, and best practices for building efficient multi-agent collaboration systems.
A step-by-step tutorial for installing and running AutoGPT locally, including environment setup, Docker deployment, and common troubleshooting.
An in-depth explanation of Retrieval-Augmented Generation and how to build private knowledge bases for AI agents to improve accuracy and reliability.