# Model Architecture > Every principle in this category is listed as a record. Page: Ontology ยท Principles Canonical: https://banes-lab.com/ontology#arch-category-model-architecture Every principle in this category is listed as a record. Each record carries its kind, its severity, the scopes it applies at and the layer it lives in, then the edge relations that join it to other records, the records that point back at it, the contracts that answer to it and the tensions it takes part in. The descriptors say how it is violated, detected, measured, repaired and enforced. Where the record carries one, an exemplar shows the shape before and after the principle is applied. Relations diagram The relations inside this category. ```mermaid flowchart LR n_artificial_intelligence_architecture["Artificial Intelligence Architecture"] n_machine_learning_architecture["Machine Learning Architecture"] n_model_governance["Model Governance"] n_model_evaluation["Model Evaluation"] n_model_inference["Model Inference"] n_retrieval_augmented_generation["Retrieval-Augmented Generation (RAG)"] n_vector_search["Vector Search"] n_knowledge_graphs["Knowledge Graphs"] n_explainability["Explainability"] n_model_safety["Model Safety"] n_prompt_engineering["Prompt Engineering"] n_model_drift_monitoring["Model Drift Monitoring"] n_agentic_architecture["Agentic Architecture"] n_artificial_intelligence_architecture --> n_model_governance n_artificial_intelligence_architecture --> n_model_evaluation n_artificial_intelligence_architecture --> n_model_safety n_artificial_intelligence_architecture -.-> n_explainability n_machine_learning_architecture --> n_model_governance n_model_governance --> n_model_safety n_model_evaluation --> n_model_safety n_model_inference --> n_artificial_intelligence_architecture n_retrieval_augmented_generation --> n_explainability n_vector_search --> n_retrieval_augmented_generation n_knowledge_graphs --> n_explainability n_model_safety --> n_model_governance n_prompt_engineering --> n_model_inference n_prompt_engineering --> n_model_evaluation n_model_drift_monitoring --> n_model_evaluation n_model_drift_monitoring --> n_model_governance n_agentic_architecture --> n_model_inference n_agentic_architecture --> n_explainability n_agentic_architecture --> n_model_safety ``` ### Artificial Intelligence Architecture - Kind: [model](https://banes-lab.com/records/kind/model.md) - Severity: contextual/mandatory for model-backed systems - Scope: model-backed system, application, platform - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Model Governance](https://banes-lab.com/records/arch/model-governance.md), [Data/Model Boundaries](https://banes-lab.com/records/lex/data-model-boundaries.md) Reinforces [Model Evaluation](https://banes-lab.com/records/arch/model-evaluation.md), [Model Safety](https://banes-lab.com/records/arch/model-safety.md) Enables [Model-Integrated Systems](https://banes-lab.com/records/lex/model-integrated-systems.md) In tension with [Determinism](https://banes-lab.com/records/arch/determinism.md), [Explainability](https://banes-lab.com/records/arch/explainability.md) Conflicts with [Opaque Ungoverned Model Use](https://banes-lab.com/records/lex/opaque-ungoverned-model-use.md) Referenced by [Model Inference](https://banes-lab.com/records/arch/model-inference.md) Tensions [Artificial Intelligence Architecture Determinism](https://banes-lab.com/records/tension/artificial-intelligence-architecture-determinism.md), [Artificial Intelligence Architecture Explainability](https://banes-lab.com/records/tension/artificial-intelligence-architecture-explainability.md) Violated by model behavior integrated without evaluation/governance Detected by model calls without tests, [logging](https://banes-lab.com/records/arch/logging.md), [fallback](https://banes-lab.com/records/lex/fallback.md), [policy](https://banes-lab.com/records/lex/policy.md) Measured by model quality/safety/evaluation coverage Refactored by Add Evaluation Harness, Add Model Boundary, Add Guardrails Enforced by model governance gates Before ```typescript async function answerFoo(prompt: string) { return model.generate(prompt); } ``` After ```typescript async function answerFoo(request: FooRequest) { const input = FooRequestSchema.parse(request); const context = await fooRetriever.retrieve(input.query); const output = await fooModel.generate(buildFooPrompt(input, context)); return FooResponseSchema.parse(output); } ``` ### Machine Learning Architecture - Kind: [model](https://banes-lab.com/records/kind/model.md) - Severity: contextual - Scope: ML pipeline, model serving, data - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Data Pipeline](https://banes-lab.com/records/lex/data-pipeline.md), [Training/Inference Separation](https://banes-lab.com/records/lex/training-inference-separation.md) Reinforces [Reproducibility](https://banes-lab.com/records/arch/reproducibility.md), [Model Governance](https://banes-lab.com/records/arch/model-governance.md) Enables [Reliable Model Lifecycle](https://banes-lab.com/records/lex/reliable-model-lifecycle.md) In tension with [Experimentation Speed](https://banes-lab.com/records/lex/experimentation-speed.md) Conflicts with [Ad-Hoc Notebook-to-Production](https://banes-lab.com/records/lex/ad-hoc-notebook-to-production.md) Tensions [Machine Learning Architecture Experimentation Speed](https://banes-lab.com/records/tension/experimentation-speed-machine-learning-architecture.md) Violated by unversioned data/model/config Detected by missing lineage, untracked training inputs Measured by [reproducibility](https://banes-lab.com/records/arch/reproducibility.md), drift, evaluation metrics Refactored by Add ML Pipeline, Version Data/Model/Config Enforced by MLOps gates Before ```typescript const model = trainFoo(loadAllData()); serve(model); ``` After ```typescript const dataset = datasetRegistry.load("foo", "v3"); const features = fooFeaturePipeline.transform(dataset); const model = trainFoo(features, versionedTrainingConfig); modelRegistry.register(model, evaluateFooModel(model, validationSet)); ``` ### Model Governance - Kind: [activity](https://banes-lab.com/records/kind/activity.md) - Severity: mandatory for production models - Scope: model lifecycle, model-backed system - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Model Registry](https://banes-lab.com/records/lex/model-registry.md), [Evaluation](https://banes-lab.com/records/lex/evaluation.md), [Approval Policy](https://banes-lab.com/records/lex/approval-policy.md) Reinforces [Compliance](https://banes-lab.com/records/arch/compliance.md), [Model Safety](https://banes-lab.com/records/arch/model-safety.md) Enables [Controlled Model Deployment](https://banes-lab.com/records/lex/controlled-model-deployment.md) In tension with [Experiment Velocity](https://banes-lab.com/records/lex/experiment-velocity.md) Conflicts with [Unapproved Model Deployment](https://banes-lab.com/records/lex/unapproved-model-deployment.md), [Model Version Ambiguity](https://banes-lab.com/records/arch/model-version-ambiguity.md) Referenced by [Artificial Intelligence Architecture](https://banes-lab.com/records/arch/artificial-intelligence-architecture.md), [Machine Learning Architecture](https://banes-lab.com/records/arch/machine-learning-architecture.md), [Model Safety](https://banes-lab.com/records/arch/model-safety.md), [Model Drift Monitoring](https://banes-lab.com/records/arch/model-drift-monitoring.md) Tensions [Model Governance Experiment Velocity](https://banes-lab.com/records/tension/experiment-velocity-model-governance.md) Violated by deploying unapproved/untracked models Detected by model without lineage/approval/eval Measured by governance coverage Refactored by Add Registry, Add Approval Workflow, Add Eval Gates Enforced by CI/CD model gates Before ```typescript deployModel(newestModelFile()); ``` After ```typescript const candidate = modelRegistry.get("foo-model", "1.4.0"); requireApproval(candidate, ["model-owner", "risk-owner"]); requirePolicyCompliance(candidate, fooModelPolicies); deployModel(candidate); ``` ### Model Evaluation - Kind: [activity](https://banes-lab.com/records/kind/activity.md) - Severity: mandatory - Scope: model, model-backed feature, pipeline - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Dataset](https://banes-lab.com/records/lex/dataset.md), [Metrics](https://banes-lab.com/records/lex/metrics.md), [Acceptance Criteria](https://banes-lab.com/records/lex/acceptance-criteria.md) Reinforces [Model Safety](https://banes-lab.com/records/arch/model-safety.md), [Correctness](https://banes-lab.com/records/arch/correctness.md) Enables [Model Selection/Regression Detection](https://banes-lab.com/records/lex/model-selection-regression-detection.md) In tension with [Metric Completeness](https://banes-lab.com/records/lex/metric-completeness.md) Conflicts with [Untested Model Deployment](https://banes-lab.com/records/lex/untested-model-deployment.md) Referenced by [Artificial Intelligence Architecture](https://banes-lab.com/records/arch/artificial-intelligence-architecture.md), [Prompt Engineering](https://banes-lab.com/records/arch/prompt-engineering.md), [Model Drift Monitoring](https://banes-lab.com/records/arch/model-drift-monitoring.md) Tensions [Model Evaluation Metric Completeness](https://banes-lab.com/records/tension/metric-completeness-model-evaluation.md) Violated by model change without evaluation Detected by missing eval report/gate Measured by task metrics, safety metrics, regression rate Refactored by Add Eval Suite, Add Regression Dataset Enforced by model CI gates Before ```typescript if (model.accuracy > 0.8) deploy(model); ``` After ```typescript const evaluation = evaluateModel(model, { datasets: [fooValidationSet, fooStressSet], metrics: [precision, recall, calibration, latencyP95], slices: ["foo-kind", "foo-region"], }); requireThresholds(evaluation, fooModelThresholds); ``` ### Model Inference - Kind: [capability](https://banes-lab.com/records/kind/capability.md) - Severity: contextual - Scope: service, model serving - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Model Artifact](https://banes-lab.com/records/lex/model-artifact.md), [Input/Output Contract](https://banes-lab.com/records/lex/input-output-contract.md) Reinforces [Artificial Intelligence Architecture](https://banes-lab.com/records/arch/artificial-intelligence-architecture.md) Enables [Runtime Prediction/Generation](https://banes-lab.com/records/lex/runtime-prediction-generation.md) In tension with [Latency/Cost](https://banes-lab.com/records/lex/latency-cost.md) Conflicts with [Training-Time-Only Model Logic](https://banes-lab.com/records/lex/training-time-only-model-logic.md) Referenced by [Prompt Engineering](https://banes-lab.com/records/arch/prompt-engineering.md), [Agentic Architecture](https://banes-lab.com/records/arch/agentic-architecture.md) Tensions [Model Inference Latency/Cost](https://banes-lab.com/records/tension/latency-cost-model-inference.md) Violated by inference without validation/observability/fallback Detected by raw model calls in business logic Measured by [latency](https://banes-lab.com/records/arch/latency.md), error rate, output quality Refactored by Add Inference Service, Add Adapter/Contract Enforced by serving standards Before ```typescript const output = model.predict(input as any); ``` After ```typescript const input = FooInferenceSchema.parse(rawInput); const output = await inferenceRuntime.predict(fooModelVersion, input, { timeoutMs: 500, traceId, }); return FooPredictionSchema.parse(output); ``` ### Retrieval-Augmented Generation (RAG) - Kind: [pattern](https://banes-lab.com/records/kind/pattern.md) - Severity: contextual - Scope: language-model system, knowledge retrieval - Aliases: RAG - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Retriever](https://banes-lab.com/records/lex/retriever.md), [Document Store](https://banes-lab.com/records/lex/document-store.md), [Grounding Strategy](https://banes-lab.com/records/lex/grounding-strategy.md) Reinforces [Explainability](https://banes-lab.com/records/arch/explainability.md), [Knowledge Freshness](https://banes-lab.com/records/lex/knowledge-freshness.md) Enables [Contextual Generation](https://banes-lab.com/records/lex/contextual-generation.md) In tension with [Retrieval Quality/Latency](https://banes-lab.com/records/lex/retrieval-quality-latency.md) Conflicts with [Ungrounded Generation](https://banes-lab.com/records/lex/ungrounded-generation.md) Referenced by [Vector Search](https://banes-lab.com/records/arch/vector-search.md) Tensions [Retrieval-Augmented Generation (RAG) Retrieval Quality/Latency](https://banes-lab.com/records/tension/retrieval-augmented-generation-rag-retrieval-quality-latency.md) Violated by answers generated without relevant retrieved context where required Detected by missing citations/context in grounded tasks Measured by retrieval precision/recall, groundedness Refactored by Add Retriever, Add Reranker, Add Citation Grounding Enforced by RAG evals Before ```typescript const answer = await model.generate(`Answer: ${question}`); ``` After ```typescript const query = normalizeFooQuery(question); const documents = await fooRetriever.search(query, { topK: 8 }); const groundedPrompt = buildGroundedFooPrompt(question, documents); const answer = await model.generate(groundedPrompt); return attachCitations(answer, documents); ``` ### Vector Search - Kind: [mechanism](https://banes-lab.com/records/kind/mechanism.md) - Severity: contextual - Scope: retrieval, search, RAG - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Embeddings](https://banes-lab.com/records/lex/embeddings.md), [Vector Index](https://banes-lab.com/records/lex/vector-index.md) Reinforces [Retrieval-Augmented Generation (RAG)](https://banes-lab.com/records/arch/retrieval-augmented-generation.md), [Semantic Search](https://banes-lab.com/records/lex/semantic-search.md) Enables [Similarity Retrieval](https://banes-lab.com/records/lex/similarity-retrieval.md) In tension with [Explainability/Recall](https://banes-lab.com/records/lex/explainability-recall.md) Conflicts with [Exact Keyword Search Only](https://banes-lab.com/records/lex/exact-keyword-search-only.md) Tensions [Vector Search Explainability/Recall](https://banes-lab.com/records/tension/explainability-recall-vector-search.md) Violated by semantic retrieval requirement implemented with only brittle keyword matching Detected by poor semantic recall Measured by retrieval metrics, [latency](https://banes-lab.com/records/arch/latency.md) Refactored by Add Embeddings, Add Vector Index, Tune Retrieval Enforced by retrieval evaluation Before ```typescript const results = foos.filter(foo => foo.text.includes(query)); ``` After ```typescript const queryVector = await embedder.embed(query); const results = await fooVectorIndex.search(queryVector, { topK: 10, filter: { tenantId }, }); ``` ### Knowledge Graphs - Kind: [pattern](https://banes-lab.com/records/kind/pattern.md) - Severity: contextual - Scope: knowledge modeling, retrieval, reasoning - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Entities](https://banes-lab.com/records/lex/entities.md), [Relations](https://banes-lab.com/records/lex/relations.md), [Schema/Ontology](https://banes-lab.com/records/lex/schema-ontology.md) Reinforces [Semantic Consistency](https://banes-lab.com/records/arch/semantic-consistency.md), [Explainability](https://banes-lab.com/records/arch/explainability.md) Enables [Relationship-Aware Retrieval/Reasoning](https://banes-lab.com/records/lex/relationship-aware-retrieval-reasoning.md) In tension with [Curation Cost](https://banes-lab.com/records/lex/curation-cost.md) Conflicts with [Flat Document-Only Knowledge](https://banes-lab.com/records/lex/flat-document-only-knowledge.md) Tensions [Knowledge Graphs Curation Cost](https://banes-lab.com/records/tension/curation-cost-knowledge-graphs.md) Violated by relation-heavy domain modeled only as unstructured text Detected by repeated need for entity relationship traversal Measured by graph coverage, query accuracy Refactored by Extract Entities/Relations, Build Graph Enforced by schema/ontology validation Before ```typescript const fooLinks = new Map(); fooLinks.set(foo.id, [bar.id, baz.id]); ``` After ```typescript const graph = new KnowledgeGraph(); graph.addNode(foo.id, "Foo", foo); graph.addNode(bar.id, "Bar", bar); graph.addEdge(foo.id, "DEPENDS_ON", bar.id); graph.addEdge(bar.id, "PRODUCES", baz.id); ``` ### Explainability - Kind: [quality-attribute](https://banes-lab.com/records/kind/quality-attribute.md) - Severity: contextual/mandatory in regulated domains - Scope: model, model-backed system, decision flow - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Traceability](https://banes-lab.com/records/arch/traceability.md), [Rationale/Evidence](https://banes-lab.com/records/lex/rationale-evidence.md) Reinforces [Governance](https://banes-lab.com/records/arch/governance.md), [Trust](https://banes-lab.com/records/lex/trust.md) Enables [Audit and Debugging](https://banes-lab.com/records/lex/audit-and-debugging.md) In tension with [Model Complexity](https://banes-lab.com/records/lex/model-complexity.md) Conflicts with [Opaque Black-Box Decisions](https://banes-lab.com/records/lex/opaque-black-box-decisions.md) Referenced by [Artificial Intelligence Architecture](https://banes-lab.com/records/arch/artificial-intelligence-architecture.md), [Retrieval-Augmented Generation (RAG)](https://banes-lab.com/records/arch/retrieval-augmented-generation.md), [Knowledge Graphs](https://banes-lab.com/records/arch/knowledge-graphs.md), [Agentic Architecture](https://banes-lab.com/records/arch/agentic-architecture.md) Tensions [Explainability Model Complexity](https://banes-lab.com/records/tension/explainability-model-complexity.md) Violated by consequential model decisions without explanation/evidence Detected by missing rationale/feature attribution/citations Measured by explanation coverage/quality Refactored by Add Explanation Layer, Add Evidence Trace Enforced by model governance gates Before ```typescript return model.predict(foo.features); ``` After ```typescript const prediction = await model.predict(foo.features); const explanation = await explainer.explain({ modelVersion: model.version, input: foo.features, prediction, }); return { prediction, explanation }; ``` ### Model Safety - Kind: [quality-attribute](https://banes-lab.com/records/kind/quality-attribute.md) - Severity: mandatory for model-backed systems - Scope: model-backed system, model, application - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Evaluation](https://banes-lab.com/records/lex/evaluation.md), [Guardrails](https://banes-lab.com/records/lex/guardrails.md), [Monitoring](https://banes-lab.com/records/arch/monitoring.md) Reinforces [Model Governance](https://banes-lab.com/records/arch/model-governance.md), [Security](https://banes-lab.com/records/lex/security.md) Enables [Safe Model Deployment](https://banes-lab.com/records/lex/safe-model-deployment.md) In tension with [Capability/Utility](https://banes-lab.com/records/lex/capability-utility.md) Conflicts with [Unguarded Model Autonomy](https://banes-lab.com/records/lex/unguarded-model-autonomy.md) Referenced by [Artificial Intelligence Architecture](https://banes-lab.com/records/arch/artificial-intelligence-architecture.md), [Model Governance](https://banes-lab.com/records/arch/model-governance.md), [Model Evaluation](https://banes-lab.com/records/arch/model-evaluation.md), [Agentic Architecture](https://banes-lab.com/records/arch/agentic-architecture.md) Tensions [Model Safety Capability/Utility](https://banes-lab.com/records/tension/capability-utility-model-safety.md) Violated by unsafe outputs/actions without guardrails Detected by safety eval failures, missing policy filters Measured by safety incident rate, eval pass rate Refactored by Add Guardrails, Add Human Review, Add Safety Evals Enforced by safety gates, runtime monitors Before ```typescript return model.generate(userPrompt); ``` After ```typescript const input = await safety.validateInput(userPrompt); const draft = await model.generate(input); const checked = await safety.validateOutput(draft, { policy: "foo-assistant-v2" }); if (!checked.allowed) return safeRefusal(checked.reasons); return checked.output; ``` ### Prompt Engineering - Kind: [technique](https://banes-lab.com/records/kind/technique.md) - Severity: contextual/mandatory for model-backed systems - Scope: model-backed system, language-model system, application - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Model Inference](https://banes-lab.com/records/arch/model-inference.md) Reinforces [Model Evaluation](https://banes-lab.com/records/arch/model-evaluation.md), [Reproducibility](https://banes-lab.com/records/arch/reproducibility.md) Enables [Structured, Versioned Prompts](https://banes-lab.com/records/lex/structured-versioned-prompts.md) In tension with [Robustness](https://banes-lab.com/records/lex/robustness.md) Conflicts with [Prompt Sprawl](https://banes-lab.com/records/arch/prompt-sprawl.md) Tensions [Prompt Engineering Robustness](https://banes-lab.com/records/tension/prompt-engineering-robustness.md) Violated by prompts inlined and duplicated across call sites Detected by scattered prompt string literals Measured by duplicated prompt count Refactored by Centralize and Version Prompts Enforced by model design review Before ```typescript const answer = await model.generate("summarize: " + text); ``` After ```typescript const prompt = fooPromptTemplate.render({ task: "summarize", input: text, format: "bullet-points", maxWords: 100 }); const answer = await model.generate(prompt, { temperature: 0, stop: ["\n\n"] }); ``` ### Model Drift Monitoring - Kind: [activity](https://banes-lab.com/records/kind/activity.md) - Severity: mandatory for production models - Scope: model-backed system, model lifecycle, operations - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Model Evaluation](https://banes-lab.com/records/arch/model-evaluation.md) Reinforces [Observability](https://banes-lab.com/records/arch/observability.md), [Model Governance](https://banes-lab.com/records/arch/model-governance.md) Enables [Degradation Detection](https://banes-lab.com/records/lex/degradation-detection.md), [Retraining Triggers](https://banes-lab.com/records/lex/retraining-triggers.md) In tension with [Monitoring Cost](https://banes-lab.com/records/lex/monitoring-cost.md) Conflicts with [Deploy-and-Forget Models](https://banes-lab.com/records/lex/deploy-and-forget-models.md) Tensions [Model Drift Monitoring Monitoring Cost](https://banes-lab.com/records/tension/model-drift-monitoring-monitoring-cost.md) Violated by model quality assumed stable after deployment Detected by no ongoing evaluation of live model outputs Measured by drift in accuracy/quality metrics over time Refactored by Instrument Drift Monitoring Enforced by model governance review Before ```typescript serveModel(fooModel); ``` After ```typescript monitor.track(fooModel, { metrics: [inputDistribution, predictionConfidence, groundTruthLag], alertOn: { populationStabilityIndex: 0.2 }, }); ``` ### Agentic Architecture - Kind: [pattern](https://banes-lab.com/records/kind/pattern.md) - Severity: contextual/mandatory in regulated domains - Scope: model-backed system, reasoning, decision flow - Layer: [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) Details Requires [Model Inference](https://banes-lab.com/records/arch/model-inference.md), [Tool Interface](https://banes-lab.com/records/lex/tool-interface.md) Reinforces [Explainability](https://banes-lab.com/records/arch/explainability.md), [Model Safety](https://banes-lab.com/records/arch/model-safety.md) Enables [Bounded Tool-Using Agents](https://banes-lab.com/records/lex/bounded-tool-using-agents.md), [Governed Autonomy](https://banes-lab.com/records/lex/governed-autonomy.md) In tension with [Determinism](https://banes-lab.com/records/arch/determinism.md) Conflicts with [Ungrounded Content](https://banes-lab.com/records/arch/ungrounded-content.md) Tensions [Agentic Architecture Determinism](https://banes-lab.com/records/tension/agentic-architecture-determinism.md) Violated by an unbounded model loop acting with no guardrails Detected by agent actions without tool scoping or step limits Measured by unguarded agent action rate Refactored by Bound the Agent with Tools, Limits, and Review Enforced by model safety review Before ```typescript const answer = await model.generate(question); ``` After ```typescript const agent = createFooAgent({ tools: [searchFoos, calculator, fooStore], maxSteps: 8 }); const answer = await agent.run(question); ``` ## Links to - [model](https://banes-lab.com/records/kind/model.md) - [Correctness Core](https://banes-lab.com/records/layer/correctness-core.md) - [Model Governance](https://banes-lab.com/records/arch/model-governance.md) - [Data/Model Boundaries](https://banes-lab.com/records/lex/data-model-boundaries.md) - [Model Evaluation](https://banes-lab.com/records/arch/model-evaluation.md) - [Model Safety](https://banes-lab.com/records/arch/model-safety.md) - [Model-Integrated Systems](https://banes-lab.com/records/lex/model-integrated-systems.md) - [Determinism](https://banes-lab.com/records/arch/determinism.md) - [Explainability](https://banes-lab.com/records/arch/explainability.md) - [Opaque Ungoverned Model Use](https://banes-lab.com/records/lex/opaque-ungoverned-model-use.md) - [Model Inference](https://banes-lab.com/records/arch/model-inference.md) - [Artificial Intelligence Architecture / Determinism](https://banes-lab.com/records/tension/artificial-intelligence-architecture-determinism.md) - [Artificial Intelligence Architecture / Explainability](https://banes-lab.com/records/tension/artificial-intelligence-architecture-explainability.md) - [Logging](https://banes-lab.com/records/arch/logging.md) - [Fallback](https://banes-lab.com/records/lex/fallback.md) - [Policy](https://banes-lab.com/records/lex/policy.md) - [Data Pipeline](https://banes-lab.com/records/lex/data-pipeline.md) - [Training/Inference Separation](https://banes-lab.com/records/lex/training-inference-separation.md) - [Reproducibility](https://banes-lab.com/records/arch/reproducibility.md) - [Reliable Model Lifecycle](https://banes-lab.com/records/lex/reliable-model-lifecycle.md) - [Experimentation Speed](https://banes-lab.com/records/lex/experimentation-speed.md) - [Ad-Hoc Notebook-to-Production](https://banes-lab.com/records/lex/ad-hoc-notebook-to-production.md) - [Machine Learning Architecture / Experimentation Speed](https://banes-lab.com/records/tension/experimentation-speed-machine-learning-architecture.md) - [activity](https://banes-lab.com/records/kind/activity.md) - [Model Registry](https://banes-lab.com/records/lex/model-registry.md) - [Evaluation](https://banes-lab.com/records/lex/evaluation.md) - [Approval Policy](https://banes-lab.com/records/lex/approval-policy.md) - [Compliance](https://banes-lab.com/records/arch/compliance.md) - [Controlled Model Deployment](https://banes-lab.com/records/lex/controlled-model-deployment.md) - [Experiment Velocity](https://banes-lab.com/records/lex/experiment-velocity.md) - [Unapproved Model Deployment](https://banes-lab.com/records/lex/unapproved-model-deployment.md) - [Model Version Ambiguity](https://banes-lab.com/records/arch/model-version-ambiguity.md) - [Artificial Intelligence Architecture](https://banes-lab.com/records/arch/artificial-intelligence-architecture.md) - [Machine Learning Architecture](https://banes-lab.com/records/arch/machine-learning-architecture.md) - [Model Drift Monitoring](https://banes-lab.com/records/arch/model-drift-monitoring.md) - [Model Governance / Experiment Velocity](https://banes-lab.com/records/tension/experiment-velocity-model-governance.md) - [Dataset](https://banes-lab.com/records/lex/dataset.md) - [Metrics](https://banes-lab.com/records/lex/metrics.md) - [Acceptance Criteria](https://banes-lab.com/records/lex/acceptance-criteria.md) - [Correctness](https://banes-lab.com/records/arch/correctness.md) - [Model Selection/Regression Detection](https://banes-lab.com/records/lex/model-selection-regression-detection.md) - [Metric Completeness](https://banes-lab.com/records/lex/metric-completeness.md) - [Untested Model Deployment](https://banes-lab.com/records/lex/untested-model-deployment.md) - [Prompt Engineering](https://banes-lab.com/records/arch/prompt-engineering.md) - [Model Evaluation / Metric Completeness](https://banes-lab.com/records/tension/metric-completeness-model-evaluation.md) - [capability](https://banes-lab.com/records/kind/capability.md) - [Model Artifact](https://banes-lab.com/records/lex/model-artifact.md) - [Input/Output Contract](https://banes-lab.com/records/lex/input-output-contract.md) - [Runtime Prediction/Generation](https://banes-lab.com/records/lex/runtime-prediction-generation.md) - [Latency/Cost](https://banes-lab.com/records/lex/latency-cost.md) - [Training-Time-Only Model Logic](https://banes-lab.com/records/lex/training-time-only-model-logic.md) - [Agentic Architecture](https://banes-lab.com/records/arch/agentic-architecture.md) - [Model Inference / Latency/Cost](https://banes-lab.com/records/tension/latency-cost-model-inference.md) - [Latency](https://banes-lab.com/records/arch/latency.md) - [pattern](https://banes-lab.com/records/kind/pattern.md) - [Retriever](https://banes-lab.com/records/lex/retriever.md) - [Document Store](https://banes-lab.com/records/lex/document-store.md) - [Grounding Strategy](https://banes-lab.com/records/lex/grounding-strategy.md) - [Knowledge Freshness](https://banes-lab.com/records/lex/knowledge-freshness.md) - [Contextual Generation](https://banes-lab.com/records/lex/contextual-generation.md) - [Retrieval Quality/Latency](https://banes-lab.com/records/lex/retrieval-quality-latency.md) - [Ungrounded Generation](https://banes-lab.com/records/lex/ungrounded-generation.md) - [Vector Search](https://banes-lab.com/records/arch/vector-search.md) - [Retrieval-Augmented Generation (RAG) / Retrieval Quality/Latency](https://banes-lab.com/records/tension/retrieval-augmented-generation-rag-retrieval-quality-latency.md) - [mechanism](https://banes-lab.com/records/kind/mechanism.md) - [Embeddings](https://banes-lab.com/records/lex/embeddings.md) - [Vector Index](https://banes-lab.com/records/lex/vector-index.md) - [Retrieval-Augmented Generation (RAG)](https://banes-lab.com/records/arch/retrieval-augmented-generation.md) - [Semantic Search](https://banes-lab.com/records/lex/semantic-search.md) - [Similarity Retrieval](https://banes-lab.com/records/lex/similarity-retrieval.md) - [Explainability/Recall](https://banes-lab.com/records/lex/explainability-recall.md) - [Exact Keyword Search Only](https://banes-lab.com/records/lex/exact-keyword-search-only.md) - [Vector Search / Explainability/Recall](https://banes-lab.com/records/tension/explainability-recall-vector-search.md) - [Entities](https://banes-lab.com/records/lex/entities.md) - [Relations](https://banes-lab.com/records/lex/relations.md) - [Schema/Ontology](https://banes-lab.com/records/lex/schema-ontology.md) - [Semantic Consistency](https://banes-lab.com/records/arch/semantic-consistency.md) - [Relationship-Aware Retrieval/Reasoning](https://banes-lab.com/records/lex/relationship-aware-retrieval-reasoning.md) - [Curation Cost](https://banes-lab.com/records/lex/curation-cost.md) - [Flat Document-Only Knowledge](https://banes-lab.com/records/lex/flat-document-only-knowledge.md) - [Knowledge Graphs / Curation Cost](https://banes-lab.com/records/tension/curation-cost-knowledge-graphs.md) - [quality-attribute](https://banes-lab.com/records/kind/quality-attribute.md) - [Traceability](https://banes-lab.com/records/arch/traceability.md) - [Rationale/Evidence](https://banes-lab.com/records/lex/rationale-evidence.md) - [Governance](https://banes-lab.com/records/arch/governance.md) - [Trust](https://banes-lab.com/records/lex/trust.md) - [Audit and Debugging](https://banes-lab.com/records/lex/audit-and-debugging.md) - [Model Complexity](https://banes-lab.com/records/lex/model-complexity.md) - [Opaque Black-Box Decisions](https://banes-lab.com/records/lex/opaque-black-box-decisions.md) - [Knowledge Graphs](https://banes-lab.com/records/arch/knowledge-graphs.md) - [Explainability / Model Complexity](https://banes-lab.com/records/tension/explainability-model-complexity.md) - [Guardrails](https://banes-lab.com/records/lex/guardrails.md) - [Monitoring](https://banes-lab.com/records/arch/monitoring.md) - [Security](https://banes-lab.com/records/lex/security.md) - [Safe Model Deployment](https://banes-lab.com/records/lex/safe-model-deployment.md) - [Capability/Utility](https://banes-lab.com/records/lex/capability-utility.md) - [Unguarded Model Autonomy](https://banes-lab.com/records/lex/unguarded-model-autonomy.md) - [Model Safety / Capability/Utility](https://banes-lab.com/records/tension/capability-utility-model-safety.md) - [technique](https://banes-lab.com/records/kind/technique.md) - [Structured, Versioned Prompts](https://banes-lab.com/records/lex/structured-versioned-prompts.md) - [Robustness](https://banes-lab.com/records/lex/robustness.md) - [Prompt Sprawl](https://banes-lab.com/records/arch/prompt-sprawl.md) - [Prompt Engineering / Robustness](https://banes-lab.com/records/tension/prompt-engineering-robustness.md) - [Observability](https://banes-lab.com/records/arch/observability.md) - [Degradation Detection](https://banes-lab.com/records/lex/degradation-detection.md) - [Retraining Triggers](https://banes-lab.com/records/lex/retraining-triggers.md) - [Monitoring Cost](https://banes-lab.com/records/lex/monitoring-cost.md) - [Deploy-and-Forget Models](https://banes-lab.com/records/lex/deploy-and-forget-models.md) - [Model Drift Monitoring / Monitoring Cost](https://banes-lab.com/records/tension/model-drift-monitoring-monitoring-cost.md) - [Tool Interface](https://banes-lab.com/records/lex/tool-interface.md) - [Bounded Tool-Using Agents](https://banes-lab.com/records/lex/bounded-tool-using-agents.md) - [Governed Autonomy](https://banes-lab.com/records/lex/governed-autonomy.md) - [Ungrounded Content](https://banes-lab.com/records/arch/ungrounded-content.md) - [Agentic Architecture / Determinism](https://banes-lab.com/records/tension/agentic-architecture-determinism.md) ## Linked from - [The layer topology](https://banes-lab.com/ontology/schema/the-layer-topology.md) - [The membership](https://banes-lab.com/ontology/schema/the-membership.md)