Domain: AI Platform Patterns

AI Platform Patterns

This domain covers the platform surface around LLM systems: request flows, retrieval, tool use, evaluation, tracing, model routing, and the control boundaries that keep the system operable.

mindmap
  ((AI Platform Patterns))
    (Pipeline architecture)
    (Retrieval and tool use)
    (Tooling landscape)
    (Evaluation)
    (Observability)
    (Governance)

01-llm-pipeline-architecture

The architectural building blocks of production LLM systems, including request paths, retrieval paths, execution paths, fallbacks, retries, routing, and safe degradation.

02-rag-retrieval-and-tool-use

Retrieval, reranking, embeddings, citations, tool calling, MCP-style tool connectivity, and the decision boundary between asking a model to reason and asking a system to fetch or act.

03-modern-ai-tooling-landscape

A current view of inference gateways, vector search stacks, rerankers, orchestration frameworks, eval harnesses, tracing platforms, and guardrail tooling.

01-ai-evaluation-and-quality-assurance

Offline and online evaluation, golden datasets, retrieval metrics, extraction scoring, confidence thresholds, and review routing.

02-ai-observability-and-operations

Prompt traces, retrieval traces, token and cost telemetry, provenance, incident response, rollback criteria, and operational debugging of AI workflows.

01-ai-governance-and-security

Prompt injection, excessive agency, PII handling, secret leakage, vendor boundary decisions, approval gates, and financial-domain governance controls.