Invidious: interface, media paths and operational responsibility
An Invidious learning project: explain search decisions before adding more automation
Design an offline query-route explainer with synthetic fixtures, clear evidence labels and acceptance tests, without claiming an unbuilt interactive tool.
What you will learn
- Build an explanation of a real branch boundary
- Keep synthetic inputs and evidence labels visible
- Define completion before proposing extra visuals
Before you start
- Basic HTTP and container concepts
- Ability to distinguish application state from media traffic
Explain the dependency and trust boundaries, prepare a verifiable trial and interpret source/model evidence without overclaiming.
Key takeaways
- Choose an interactive exercise that explains a real source boundary.
- Keep model output distinct from upstream execution and deployment evidence.
- An idea or static illustration is not a shipped interactive explorer.
Build an explanation of a real branch boundary
A useful learning project is a small offline explainer for search classification: show normalized input, the smart-feature inhibitor, mode/filter gates and the subsequent sanitizer boundary. This matches a concrete source-reading difficulty rather than adding a decorative 3D scene unrelated to the code. A plain accessible table or SVG trace can convey the relationship without loading a graphics engine.
The sixteen independent model cases provide initial teaching fixtures, including a root URL that is URL-like but not a video and filtered inputs that remain ordinary search. They are not an implementation oracle. A production-quality explainer would need tests in the actual source language and reviewed behavior for mode/filter parsing and final destination normalization.
Keep synthetic inputs and evidence labels visible
Use invented, non-personal identifiers and perform no network navigation from the fixture panel. Show separate labels for documented behavior, source-inspected behavior, independent-model output and actual upstream execution. This prevents a green classification badge from implying that a video exists, a redirect is safe or a deployed service has passed a security test.
An advanced version could compare a source revision with a later revision and highlight changed branches, but only after both are pinned and their semantics inspected. Do not automatically ingest a user’s viewing history to populate examples. Exported account data is outside the needs of this exercise, and a static lesson should not quietly become a telemetry collector.
Define completion before proposing extra visuals
Acceptance should include keyboard-readable controls, text alternatives, full language labels, deterministic fixtures and a clear separation between classification and final routing. If animation is added, it should explain ordered decisions and respect reduced-motion preferences. Three-dimensional rendering offers no obvious advantage for these small conditional branches and would require additional accessibility and performance justification.
This series ships original static diagrams and a runnable independent JavaScript teaching-model harness in the EasyAI repository. It does not ship an interactive query explorer, execute Crystal or test the full sanitizer. The proposed project is complete only when its actual interface, source-language behavior and accessibility checks exist and are verified; the idea itself is not a delivered feature.
Implementation steps
- 1
Specify the exact query decision to explain.
- 2
Start with synthetic offline fixtures and visible evidence labels.
- 3
Add source-language and final-destination tests before claiming equivalence.
- 4
Verify keyboard, text alternatives and translations if an interface is built.
Copy-ready example
{
"proposal": "offline query-route explainer",
"modelCasesAvailable": 16,
"interactiveExplorerShipped": false,
"sourceLanguageSuiteExecuted": false,
"networkNavigationEnabled": false,
"userHistoryRequired": false
}Frequently asked questions
Is there already a Three.js explorer in this series?
No. The delivered illustrations are static SVGs. An interactive explorer is explicitly a proposed follow-up, not a shipped feature.
Can the JavaScript model replace upstream tests?
No. It is a bounded teaching aid; source-language tests and final-route checks would be needed for stronger equivalence claims.
Sources
- invidious/src/invidious/search/query.crSource checked 2026-09-08
- invidious/spec/invidious/search/query_spec.crSource checked 2026-09-08
- invidious/src/invidious/routes/search.crSource checked 2026-09-08
- invidious/src/invidious/yt_backend/url_sanitizer.crSource checked 2026-09-08