1. Unified development language
Modern JavaScript runs across the application stack. A shared representation reduces cognitive overhead and supports isomorphic code in browser and Node.js environments.
TeqFW philosophy
Make application structure explicit and runtime-composable.
The TeqFW philosophy for modular-monolith web applications built with people and LLM agents.
Architectural context
Web architecture for an era of LLM agents.
Tequila Framework (TeqFW) is an architectural philosophy for building web applications in an era where LLM agents participate in writing and maintaining code. It grew from practical experience with modular monoliths, a single database, and one JavaScript codebase across browser and server environments.
The browser is the most universal runtime environment available today, and JavaScript is the language it executes natively. TeqFW therefore focuses on web applications rather than claiming a universal architecture for every software domain.
Structural transparency
Make relationships visible before they need to change.
Traditional architectures often rely on implicit relationships between modules, configuration, and build tools. Experienced developers can infer these relationships, but static imports, reflection-based discovery, and complex pipelines make them harder to reconstruct from source alone.
TeqFW favors explicit contracts, predictable namespaces, and runtime linking. Dependencies remain visible and analyzable, helping both developers and automated agents understand, generate, and evolve a system without hidden wiring.
Core principles
Structure that stays legible as software evolves.
2. Late binding
Container-based, namespace-addressed resolution keeps dependencies explicit, replaceable, and testable without rigid compile-time coupling.
3. Evolutionary resilience
Clearly defined interaction contracts and flexible processing help code absorb changing requirements with fewer modifications to existing components.
4. Functional data and logic
Data objects hold state while stateless handlers process structured data. The resulting processing network is easier to test, analyze, and extend.
Core principles, continued
Machine-readable architecture without a second program.
5. Namespaces for structure
Modules, packages, database tables, APIs, and configuration occupy predictable namespaces, reducing conflict and making systems easier to navigate.
6. JavaScript with JSDoc
JSDoc gives IDEs and analysis tools useful type information while source and runtime remain the same program, without a TypeScript compilation layer.
7. Code and documentation for agents
Predictable files, declared dependencies, module conventions, and consistent documentation let LLM agents assist with generation, refactoring, and maintenance while preserving architecture.
8. Explicit machine interfaces
Every TeqFW plugin distributes a version-matched agent skill at skills/<name>/SKILL.md. It describes public contracts, integration patterns, and architectural assumptions; consuming projects can mount it in .agents/skills/.
DI skill · Log skill · Cfg skill · CLI skill · DB skill · Web skill
Conclusion
Architecture for collaboration between people and intelligent agents.
TeqFW explores a way to build modular-monolith web applications where LLM agents assist in producing and maintaining code. Explicit contracts, namespace-driven organization, runtime linking, and machine-oriented documentation keep the architecture inspectable and adaptable.
It does not replace all traditional architectures. It shows how software can be structured differently when automated agents become permanent collaborators in development.
The maintained source remains available in teqfw/di; the skill catalogs complement human-facing README files and source code.