The lab notebook
Field notes on AI engineering
824 articles · page 25 of 28
LangGraph Subgraphs: Composing Complex Agents from Smaller Graphs
Learn how LangGraph subgraphs let you compose complex, multi-agent systems from smaller, testable graphs with clean state boundaries.
LangGraph Testing: Unit Testing Individual Nodes
Learn how to unit test LangGraph nodes in isolation with pytest, mocks, and state fixtures before your graph ever runs end to end.
LangGraph Time Travel Debugging: Replaying Past Agent Decisions
Learn how LangGraph's time travel debugging lets you replay, inspect, and fork past agent states using get_state_history() and update_state().
LangGraph Visualization: Understanding Your Graph at a Glance
Learn LangGraph visualization with draw_mermaid(), PNG rendering, ASCII diagrams, and xray mode so you can see and debug your agent graphs at a glance.
LangGraph vs Plain Python State Machines: Why Use a Framework
A practical comparison of LangGraph and hand-rolled Python state machines for agent workflows, with code and honest tradeoffs.
LangSmith for A/B Testing Prompt Variants in Production
Learn how to run LangSmith A/B testing on prompt variants in production with traffic splitting, feedback capture, and experiment analysis.
LangSmith for Agent Debugging: Tracing Multi-Step Decisions
Learn LangSmith agent debugging step by step: trace multi-step decisions, inspect tool calls, find silent failures, and turn broken traces into eval datasets.
LangSmith Annotation Queues: Human Review at Scale
Learn how LangSmith annotation queues turn ad-hoc trace review into a repeatable human labeling workflow that feeds datasets, evals, and LLM judges.
LangSmith Cost Tracking: Understanding Your LLM Spend
Learn how LangSmith cost tracking works, from token-based pricing maps to per-user attribution, so you always know exactly where your LLM budget goes.
LangSmith for Customer Support Quality Monitoring
Learn how to use LangSmith to trace, score, and monitor AI customer support conversations so quality problems surface before your customers complain.
LangSmith Datasets: Building and Versioning Test Data
Learn how to build, organize, and version LangSmith datasets with real SDK code so your LLM evaluations stay reproducible as your app evolves.
LangSmith Feedback Collection: Capturing User Ratings on Responses
Learn LangSmith feedback collection end to end: capture thumbs ratings, scores, and comments on LLM responses with create_feedback and feedback tokens.
LangSmith for Fine-Tuning Data Collection
Learn how to turn LangSmith traces, feedback, and annotation queues into clean, exportable langsmith fine-tuning data for training smaller, cheaper models.
LangSmith Latency Analysis: Finding Your Slowest Chain Steps
Learn how to use LangSmith latency analysis to trace slow chain steps, read waterfalls, filter laggy runs, and fix the bottlenecks that hurt users.
LangSmith for Multi-Agent Trace Analysis
Learn langsmith multi-agent tracing: instrument supervisor and worker agents, read nested trace trees, and debug handoffs, loops, and cost hotspots.
LangSmith Onboarding: Getting Your Team Set Up in a Day
A practical, hour-by-hour LangSmith onboarding guide: workspaces, API keys, tracing, roles, datasets, and alerts — your whole team productive in one day.
LangSmith Playground: Iterating on Prompts Interactively
Learn how to use the LangSmith Playground to edit, test, compare, and version prompts interactively — without redeploying code between every tweak.
LangSmith Prompt Hub: Versioning and Sharing Prompts
Learn how the LangSmith Prompt Hub versions, tags, and shares prompts across your team, with real pull and push code you can use in production today.
LangSmith for RAG Debugging: Finding Retrieval Failures Fast
Learn langsmith rag debugging step by step: trace pipelines, spot retrieval failures, build eval datasets, and fix hallucinations at the source.
LangSmith Regression Testing: Comparing Prompt Versions Automatically
Learn how to set up LangSmith regression testing with datasets, evaluate() runs, and CI gates so prompt changes never silently break your LLM app.
LangSmith Self-Hosting: When to Run It On-Premise
A practical guide to LangSmith self-hosting: architecture, deployment, day-two operations, and an honest framework for deciding on-premise vs cloud.
LangSmith Team Workflows: Roles and Access Control
Learn how LangSmith team workflows use organizations, workspaces, roles, and API keys to give every engineer the right access without slowing anyone down.
LangSmith Trace Filtering: Finding the Runs That Matter
Master LangSmith trace filtering with the filter query language, SDK filters, and saved views to isolate failing, slow, or costly runs in seconds.
LangSmith Webhooks and Alerts: Automating Incident Response
Learn how to configure LangSmith webhooks and alerts to detect LLM failures early and automate incident response with real code examples.
LangSmith Integration Without LangChain: Using the SDK Directly
Trace, evaluate, and debug any LLM app with LangSmith without LangChain — using @traceable, wrap_openai, and RunTree from the plain Python SDK.
Migrating Off LangChain: When and How to Simplify
A practical guide to knowing when LangChain has become overhead instead of leverage, and how to migrate to simpler code without a rewrite disaster.
n8n Access Control: Managing Team Permissions
A practical guide to n8n's roles, project-based permissions, and credential sharing for teams running automation at scale.
n8n Community Nodes: Extending n8n with Custom Integrations
A practical guide to installing, evaluating, and building n8n community nodes to integrate any API into your automation workflows.
n8n Community vs Enterprise: Which Plan Do You Need
A practical, honest breakdown of n8n Community vs Enterprise so you pick the right plan for your team instead of overpaying or hitting a wall.
n8n for Content Generation Pipelines
Learn how to build automated, AI-powered content generation pipelines in n8n, from research and drafting to publishing and distribution.