One search. Connected ideas.

Provider failover chooses an allowed alternative after eligible failure. Multi-tenancy scopes records and operations; security enforces trust boundaries. Observability connects execution paths to their outcomes and resource costs.

An AI gateway sits between your application and model providers. It can normalize access, choose approved models, enforce quotas, collect usage, and handle eligible fallback. It is similar to an API gateway specialized for model traffic.

Beginner — What do you trace in an AI request?

Evaluate no evidence, stale policies, malformed documents, ambiguous dates, denied access, injected content, restarts, duplicate submissions, and uncertain writes. Track source/model/prompt/graph versions, p95 latency, tokens, cost per successful task,…

An agent loop alternates decision, action, and observation. ReAct is a pattern that combines deciding what information/action is needed with executing tools and observing results. “Reasoning” here means useful task planning and decision-making, not a…

Vision models interpret image content, such as objects, layout, handwriting, and visual relationships. Image understanding produces descriptions or structured observations; it does not automatically verify authenticity or determine business liability. Quality…

An AI agent is software that uses a model to choose actions toward a goal, observes results, and decides what to do next within application limits. A chatbot describes a conversational interface; it may or may not contain an agent. A workflow defines the…

An agentic workflow allows a model to choose some steps based on observations. The surrounding application limits the actions, resources, and impact of those choices. It is useful for variable tasks but offers less predictable paths and call counts.

Worked decision: “Automatically approve claims from uploaded photos.” First separate observation from adjudication. Vision can identify proposed observations, OCR can extract document fields, and RAG can retrieve applicable terms. Approval requires…

Message roles identify the purpose and source of information. System or developer instructions describe application behavior, user messages contain requests, assistant messages contain model responses, and tool results supply external observations. Exact role…

A graph can separate model decisions, tool execution, and stop checks into different nodes. The model node proposes calls; the tool node uses approved business services. Conditional routing returns to the model only while permitted progress remains.

AI and LLM fundamentals

An agent loop alternates decision, action, and observation. ReAct is a pattern that combines deciding what information/action is needed with executing tools and observing results. “Reasoning” here means useful task planning and decision-making, not a…

Vision models interpret image content, such as objects, layout, handwriting, and visual relationships. Image understanding produces descriptions or structured observations; it does not automatically verify authenticity or determine business liability. Quality…

An AI agent is software that uses a model to choose actions toward a goal, observes results, and decides what to do next within application limits. A chatbot describes a conversational interface; it may or may not contain an agent. A workflow defines the…

An agentic workflow allows a model to choose some steps based on observations. The surrounding application limits the actions, resources, and impact of those choices. It is useful for variable tasks but offers less predictable paths and call counts.

Worked decision: “Automatically approve claims from uploaded photos.” First separate observation from adjudication. Vision can identify proposed observations, OCR can extract document fields, and RAG can retrieve applicable terms. Approval requires…

Message roles identify the purpose and source of information. System or developer instructions describe application behavior, user messages contain requests, assistant messages contain model responses, and tool results supply external observations. Exact role…

A graph can separate model decisions, tool execution, and stop checks into different nodes. The model node proposes calls; the tool node uses approved business services. Conditional routing returns to the model only while permitted progress remains.

A trace connects the operations serving one request or task. Each timed operation is a span, such as an LLM call, retrieval, or tool execution. An agent run may include several spans and persisted state transitions across multiple requests.

Provider failover chooses an allowed alternative after eligible failure. Multi-tenancy scopes records and operations; security enforces trust boundaries. Observability connects execution paths to their outcomes and resource costs.

AI and LLM fundamentals

A planner proposes steps; an executor performs permitted steps and reports results. Splitting the roles separates deciding from acting. The executor still enforces constraints rather than blindly following the plan.

These platforms help collect traces, datasets, scores, and experiment results. LangSmith integrates closely with the LangChain ecosystem; Langfuse provides AI observability and evaluation workflows; Phoenix supports tracing and evaluation analysis. They are…

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module

AI Engineering learning module