Testing Plan

RUSVEL targets ~1200+ tests across Rust and frontend, with staged workspace coverage toward 55-65%. This page summarizes the strategy; the full checklist lives in docs/plans/comprehensive-testing-plan.md.

Current state

Area	Count	Notes
Rust tests (workspace sum)	~645	Sum of `running N tests` from `cargo test --workspace`; see `docs/status/current-state.md`
Rust integration tests	~92	Strong API smoke tests, engine round-trips
Benchmarks	2	Criterion: DB open + registry load
Frontend visual tests	27 routes	Playwright screenshot comparison
Frontend unit tests	0	No Vitest setup yet
Property-based tests	0	No proptest yet
Fuzz tests	0	No cargo-fuzz targets yet
Snapshot tests	0	No insta snapshots yet

CI coverage floor: 42% (cargo-llvm-cov, --fail-under-lines).

Testing pyramid

                    /\
                   /  \          E2E (Playwright)
                  / 27  \        Visual regression + interaction flows
                 /--------\
                /          \     Integration Tests
               /   ~150     \    API handlers, engine pipelines, cross-crate
              /--------------\
             /                \   Unit Tests
            /     ~800+        \  Domain types, port impls, utils, components
           /--------------------\
          /                      \  Property + Fuzz
         /        ~50+            \  Parsers, serialization, scoring
        /__________________________\

Most tests at the unit level. Integration tests for cross-boundary flows. E2E for critical user journeys only.

Sprint overview

Sprint	Theme	New Tests
0	Baseline capture	0 (measurement)
1	Rust unit tests (rusvel-core to 90%, all crates > 50%)	~200
2	Frontend Vitest setup + api.ts + stores	~100
3	API negative path tests (400/401/404/405 for all routes)	~120
4	Rust engine contract tests (13 engines)	~80
5	Snapshot tests (insta) + property-based tests (proptest)	~70
6	Frontend component tests (@testing-library/svelte)	~60
7	Benchmarks (Criterion) + CI hardening	~15
8	Fuzz testing (cargo-fuzz, nightly)	~8

Coverage targets by layer

These align with docs/testing/coverage-strategy.md:

Layer	Current	Target
rusvel-core	~60%	85-95%
Engines (13)	~50%	70-90%
Adapters	~35%	60-80%
rusvel-api	~45%	50-70%
rusvel-app	~25%	30-50%
Frontend (lib/)	0%	50%
Workspace total	~42%	~55-65%

Test infrastructure

Rust

Test harness: crates/rusvel-api/tests/common/mod.rs provides TestHarness with temp SQLite, stub LLM, mock platform adapters
Mock pattern: Port-based mocking via FakeStorage, RecordingEvents, StubLlm structs
Framework: tokio async runtime, tempfile for isolation, criterion for benchmarks
Coverage: cargo-llvm-cov with HTML reports via ./scripts/coverage.sh

Frontend

Visual regression: Playwright + Claude Vision analysis (27 routes)
Unit tests (planned): Vitest + @testing-library/svelte
Test data: Global setup seeds sessions and goals via API

Key principles

Port-based mocking – engines receive Arc<dyn Trait> mocks, never real adapters in unit tests
No mocking the database in integration tests – use temp SQLite for realistic behavior
Negative paths are required – every API endpoint needs 400, 401, 404, 405 tests
Property tests for parsers – proptest on code-engine parser, harvest scoring, cron expressions
Snapshots for serialization – insta catches accidental field renames in API responses

Running tests

# Rust
cargo test                          # Full workspace
cargo test -p rusvel-core           # Single crate
cargo bench -p rusvel-app --bench boot  # Benchmarks
./scripts/coverage.sh               # HTML coverage report

# Frontend
cd frontend
pnpm test                           # Vitest unit tests (planned)
pnpm test:visual                    # Playwright visual regression
pnpm test:e2e                       # All Playwright tests
pnpm test:coverage                  # Vitest coverage (planned)

Keyboard shortcuts

RUSVEL