Real-world applications
How Context Architecture Scales
From replacing the Prometheus Helm chart to modeling Kantian reasoning frameworks. Same composition mechanism. Every depth.
Prometheus Helm Chart Replacement
The community Prometheus Helm chart: the definitive real-world test of infrastructure composition at scale.
The Problem
- •2,734 lines across 17 template files
- •500+ control flow directives (if/else/loop)
- •572+ open GitHub issues
- •Environment drift is structural
The Solution
- ✓6 composable layers replace all 17 files
- ✓Zero control flow — pure composition
- ✓3ms composition vs 500ms-2s Helm template
- ✓Drift is structurally impossible
The 5-Layer Fractal Stack
K8s Substrate
37 points — Reusable Kubernetes primitives
Prometheus Language
50 points — Domain-specific shapes and semantics
Base Defaults
194 points — Sensible defaults for all environments
Production
219 points — Production-specific overrides and performance tuning
Team: Payments
229 points — Team-specific configuration and resource allocation
Key insight: The same composition mechanism that scales infrastructure (K8s → Prometheus → Production → Team) scales to any domain. Environment forking, team-specific layers, and independent overrides emerge from pure composition, not special-case logic.
Kantian Reasoning Framework
Kantext models whole reasoning frameworks, not just infrastructure. The tree IS the architectonic—philosophy and context scale the same way.
The Architecture
The framework tree branches through layers of reasoning:
The Insight
A reasoning framework is contextual. The model doesn't memorize the framework — it navigates the composed tree. Domain-specific adaptations (like medical ethics) reshape foundational formulas for clinical dignity, not by override but by adapted composition.
The 4-Layer Fractal Stack
Eval Substrate
21 points — Core evaluation primitives and rules
Kantian Scales
115 points — Categorical imperatives and synthetic judgment
The Architectonic
254 points — Systematic structure of reason itself
Medical Ethics
329 points — Humanity formula adapted for clinical dignity
Key insight: Framework adaptation isn't override — it's composition. The medical ethics layer doesn't replace the Humanity formula, it reshapes it through domain-specific boundaries. The model navigates the same compositional mechanism at every depth.
5 Use Case Categories
Real problems, architectural solutions.
Platform Engineering
Config composition, environment forking, multi-environment drift prevention.
AI/ML Context
Reasoning frameworks, model evaluation, grounded context for LLMs.
Compliance & Audit
Grounded provenance, blake3 hash chains, sealed identity audit trails.
Multi-Team Coordination
Independent overrides, team-specific layers, shared vocabulary.
DevOps & Infrastructure
Unified config source, template rendering, Terraform/Helm/K8s projections.
In Depth
Platform Engineering
Configuration as Composed Context
Multi-environment configuration drift is structural in YAML. Dev, staging, and production each maintain separate copies with invisible divergence. Kantext solves this by composing one vocabulary scoped to boundaries. Development and Production declare what they mean. Values adapt automatically through composition — not through templating, conditionals, or manual synchronization.
Example: Prometheus stack with independent staging/production branches, team-specific resource allocation per service.
AI/ML Context
Grounded Context for LLMs & Agents
LLMs receive raw data and must infer meaning. Kantext grounds context—every value carries its meaning, constraints, and purpose in a composed structure. The MCP server means any Claude-compatible agent can compose and query context natively, without prompt engineering. Reasoning frameworks like Kantian evaluation compose on substrates where the model navigates the tree itself.
Example: Kantian reasoning framework adapting to medical ethics; evaluation models querying grounded context through HQL.
Compliance & Audit
Provenance by Design
Audit trails are reconstructed after the fact—git blame, commit messages, and tribal knowledge. Kantext makes provenance structural. Every sealed composition is a cryptographic triple: WHERE (git repo, file, commit), WHAT (composed frame, blake3 hash), WHO (sealed identity). The audit trail isn't reconstructed. It IS the composition. Every value traces to its source file, line number, and sealer.
Example: Change tracking with immutable provenance chains; sealed frames with blake3 hashes; WHO signature binding identity to composition.
Multi-Team Coordination
Shared Vocabulary, Local Adaptation
Teams share infrastructure but maintain separate configurations. Platform team and application team files drift apart with no structural connection. Kantext composes a shared vocabulary as a base layer. Teams define local Boundaries for adaptation. Stacking compositions merges them back. The Holograph shows exactly where teams align and where they diverge—as structural mergisms with full provenance.
Example: Platform base layer + application team overrides stacked together; structural visibility into alignment and divergence.
DevOps & Infrastructure
Single Grounded Source
Terraform, Helm, Kubernetes manifests, and CI/CD pipelines each carry fragments of context in different formats with different assumptions. Kantext composes infrastructure context once. Render to any format via templates. The Holograph is the single grounded source; all tool-specific outputs are projections. Change context, regenerate projections. Configuration becomes unified.
Example: Single Prometheus composition renders to Helm values, Terraform state, and K8s manifests; truth lives in the Holograph.
Explore Kantext
Try the playground. Read the architecture. See what composable context means in practice.