# Engineering References and Applied Principles

Reference documents provide evidence and design guidance; they do not override approved project specs or agent rules.

## Designing Data-Intensive Applications — Martin Kleppmann

Source: <https://martin.kleppmann.com/2017/03/27/designing-data-intensive-applications.html>

Applied here through explicit reliability, scalability, and maintainability trade-offs; immutable import provenance; idempotent processing; derived-data rebuildability; and separating the logical data model from physical indexes/materializations chosen from measured access patterns.

## Site Reliability Engineering — Google

Source: <https://sre.google/sre-book/table-of-contents/>

Applied here through service-level objectives before operational tuning, actionable monitoring, release health checks, rollback readiness, capacity awareness, and restore rehearsal rather than backup existence alone.

## Release It!, Second Edition — Michael T. Nygard

Source: <https://pragprog.com/titles/mnee2/release-it-second-edition/>

Applied here through bounded resources, timeouts, backpressure, failure isolation, explicit readiness/liveness, safe startup/shutdown, and treating production behavior as part of design rather than a deployment afterthought.

## Building Effective Agents — Anthropic

Source: <https://www.anthropic.com/engineering/building-effective-agents>

Applied to the post-MVP AI capability: begin with deterministic analytics and the simplest bounded LLM workflow, add agentic autonomy only where evaluation demonstrates value, keep tools/data access explicit, and retain human-readable traces and budgets.

## AI Agents and Cloud Patterns — course slides

Local source: `C:/Users/Eight/Downloads/ai_agents_cloud_patterns_1405.pdf`

Applied to future AI execution through durable job/session state, idempotency, queue boundaries, traceability, execution budgets, authentication/authorization, and an outbox-style handoff where cross-system delivery must be reliable.

## AvalAI documentation

Source: <https://docs.avalai.ir:8443/en/>

Applied as a post-MVP integration constraint: AvalAI credentials remain backend-only; provider/model configuration is environment-driven and pinned; requests receive timeouts and bounded retries; prompts receive minimized aggregate data; and provider responses cannot become the analytical source of truth.

## ZarinPal visual references

Source: <https://www.zarinpal.com/> plus the user-supplied authenticated desktop/mobile screenshots.

Applied in the frontend repository's `DESIGN.md`: Persian RTL layout, right-side navigation/drawer, spacious light canvas, white cards, blue action/active states, dark ink surfaces, restrained yellow accent, rounded geometry, and responsive behavior. Proprietary assets/fonts are used only if legally supplied.

## Framework documentation

- SvelteKit introduction: <https://svelte.dev/docs/kit/introduction>
- Go documentation: <https://go.dev/doc/>

Framework decisions remain subordinate to capability requirements. SvelteKit is used as a TypeScript application framework without assuming SEO/SSR; Go is used for the backend with architecture, DI, database access, and API choices approved through the specification process.