voice-to-notes/docs/TESTING.md

Voice to Notes — Testing Strategy

Overview

Each layer has its own test approach. Agents should run tests after every significant change before considering work complete.

---

1. Rust Backend Tests

Tool: cargo test (built-in) Location: src-tauri/src/ (inline #[cfg(test)] modules)

cd src-tauri && cargo test

What to test:

• SQLite schema creation and migrations
• Database CRUD operations (projects, media files, segments, words, speakers)
• IPC message serialization/deserialization (serde)
• llama-server port allocation logic
• File path handling across platforms

Lint/format check:

cd src-tauri && cargo fmt --check && cargo clippy -- -D warnings

---

2. Python Sidecar Tests

Tool: pytest Location: python/tests/

cd python && python -m pytest tests/ -v

What to test:

• IPC protocol: JSON-line encode/decode, message routing
• Transcription service (mock faster-whisper for unit tests)
• Diarization service (mock pyannote for unit tests)
• AI provider adapters (mock HTTP responses)
• Export formatters (SRT, WebVTT, text output correctness)
• Hardware detection logic

Lint check:

cd python && ruff check . && ruff format --check .

---

3. Frontend Tests

Tool: vitest (Vite-native, works with Svelte) Location: src/lib/**/*.test.ts

npm run test

What to test:

• Svelte store logic (transcript, playback, project state)
• Tauri bridge service (mock tauri::invoke)
• Audio sync calculations (timestamp → word mapping)
• Export option formatting

Lint/format check:

npm run check   # svelte-check (TypeScript)
npm run lint    # eslint

---

4. Integration Tests

IPC Round-trip Test

Verify Rust ↔ Python communication works end-to-end:

# From project root
cd src-tauri && cargo test --test ipc_integration

This spawns the real Python sidecar, sends a ping message, and verifies the response.

Tauri App Launch Test

Verify the app starts and the frontend loads:

cd src-tauri && cargo test --test app_launch

---

5. Quick Validation Script

Agents should run this after any significant change:

#!/bin/bash
# scripts/validate.sh — Run all checks
set -e

echo "=== Rust checks ==="
cd src-tauri
cargo fmt --check
cargo clippy -- -D warnings
cargo test
cd ..

echo "=== Python checks ==="
cd python
ruff check .
ruff format --check .
python -m pytest tests/ -v
cd ..

echo "=== Frontend checks ==="
npm run check
npm run lint
npm run test -- --run

echo "=== All checks passed ==="

---

6. Manual Testing (Final User Test)

These require a human to verify since they involve audio playback and visual UI:

• Import a real audio file and run transcription
• Verify waveform displays correctly
• Click words → audio seeks to correct position
• Rename speakers → changes propagate
• Export caption files → open in VLC/subtitle editor
• AI chat → get responses about transcript content

---

7. Test Fixtures

Store small test fixtures in tests/fixtures/:

• short_clip.wav — 5-second audio clip with 2 speakers (for integration tests)
• sample_transcript.json — Pre-built transcript data for UI/export tests
• sample_ipc_messages.jsonl — Example IPC message sequences

Agents should create mock/fixture data rather than requiring real audio files for unit tests.