Tracing Agent Runs: LangSmith, Langfuse, and OpenTelemetry
Tracing agent runs: spans for model calls, tools, retrievals, and state changes in LangSmith, Langfuse, or OpenTelemetry, without…
Author
539 articles
-
-
How to Evaluate AI Agents: Task Success, Trajectory, and Cost
Evaluate AI agents beyond answer quality: task success, tool and argument correctness, trajectory quality, cost per successful task,…
-
Context Window Management: Summarization, Compression, Pruning
Context window management for AI agents: budgeting, summarization, compression, and pruning, so agent runs keep the right state…
-
RAG for Agents: Retrieval That Adapts to the Task
RAG for agents vs chatbot retrieval: task-aware queries, grounding, access control, freshness, and retrieval as a tool the…
-
Multi-Agent vs Single-Agent: When Multi-Agent Wins (and When It Doesn’t)
A cost-benefit test for multi-agent vs single-agent: when specialization and isolation pay off, and when one agent loop…
-
Multi-Agent Architectures: Supervisor, Swarm, and Hierarchical
Supervisor, swarm, and hierarchical multi-agent architecture compared on coordination, shared state, latency, cost, and how failures spread between…
-
MCP for AI Agents: How the Model Context Protocol Standardizes Tools
MCP for AI agents: how the Model Context Protocol standardizes tools, resources, and prompts, and what it leaves…
-
CrewAI vs AutoGen vs LangGraph: Choosing a Framework
CrewAI vs AutoGen vs LangGraph compared on architecture, state, multi-agent support, and operations, plus where Microsoft Agent Framework…
-
LangGraph for Agents: Stateful, Cyclic Workflows
How LangGraph models agents as stateful, cyclic graphs: state schemas, nodes, checkpointing, human review interrupts, and when to…
-
Human-in-the-Loop: Approvals, Escalation, and Guardrails
Designing human-in-the-loop agents: approval boundaries, escalation, review, override, audit trails, and how guardrails differ from permission checks.
-
Planning and Reasoning: ReAct, Plan-and-Execute, and Tree-of-Thought
AI agent planning and reasoning patterns: how ReAct, plan-and-execute, and tree-of-thought work, and when each one earns its…
-
Agent Memory: Short-Term, Long-Term, and Episodic
Agent memory for AI agents: short-term, long-term, and episodic stores, with write policies, retrieval, freshness, and how memory…
-
Tool Use and Function Calling: Giving Agents the Ability to Act
How tool use and function calling work in AI agents: tool contracts, JSON schemas, validation, authorization, and error…
-
When to Use an AI Agent (and When a Workflow Is Enough)
When to use an AI agent instead of a workflow: a decision framework comparing predictability, cost, and failure…