Skip to content

TEST-COVERAGE.md

Latest commit

 

History

History
157 lines (114 loc) · 6.55 KB

TEST-COVERAGE.md

File metadata and controls

157 lines (114 loc) · 6.55 KB

Test Coverage Report

Auto-generated analysis. Last updated: 2026-03-28

Summary

Suite Test files Tests Time (warm)
Rust (20 crates) ~50 ~840 ~15 s
Frontend (Vitest) 28 748 ~8 s
LLM E2E 1 1 ~12 s
Total ~79 ~1 589 ~35 s

Rust Crates — Coverage by Crate

✅ Well-Tested

Crate Tests Inline tests Notes
skill-tools 189 10/27 src files Tool parsing, extraction, schemas
skill-data 109 5/16 src files Chat store, hooks, embeddings
skill-llm 48 + E2E 4/23 src files Catalog types, chat store, think tracker
skill-devices 45 9/11 src files Session, scores, adapters
skill-eeg 5 + proptests 8/10 src files DSP pipeline, band analysis, filters
skill-constants 18 1/1 Channel maps, sample rates
skill-exg 28 1/1 ExG signal processing
skill-settings 11 3/4 Settings API, keychain
skill-health 9 1/2 Health checks

⚠️ Partially Tested

Crate Tests Gap
skill-history 2 lib.rs (997 lines) has session listing/loading logic untested
skill-screenshots 5 Only integration tests; capture.rs (1357 lines), platform.rs (550 lines) untested
skill-commands 3 SVG/DOT graph renderers (svg.rs 927 lines, svg_3d.rs 377 lines) untested
skill-tts 2 neutts.rs (705 lines), lib.rs (511 lines), kitten.rs (205 lines) untested
skill-headless ~5 engine.rs (1193 lines) — the core headless inference engine — untested

❌ Missing or Minimal Tests

Crate Src files Lines (est.) What needs testing
skill-calendar 6 ~500 Platform calendar integrations (macOS/Linux/Windows)
skill-lsl 4 ~310 LSL + Iroh adapters (existing tests have unused variables)
skill-iroh 8 ~205 untested device_receiver.rs — Iroh device data receiver
skill-tray 1 small System tray (platform-specific, hard to unit test)
skill-vision 1 Obj-C macOS Vision OCR bridge (can't unit test easily)

Rust — Specific Gaps to Address

High Priority (core logic, easy to test)

  1. skill-history/src/lib.rs (997 lines) Session discovery, listing, date grouping, cache invalidation. Only 2 tests exist (path construction). Missing: listing with filters, date range queries, cache hit/miss, corrupt session handling.

  2. skill-data/src/screenshot_store.rs (682 lines) Has 4 basic insert/count tests. Missing: search queries, embedding updates, pagination, HNSW ID assignment, migration paths.

  3. skill-data/src/session_csv.rs (637 lines) / session_parquet.rs (581 lines) Session data export. No tests. Should test: round-trip write/read, column types, empty sessions, large sessions, corrupt file handling.

  4. skill-data/src/dnd.rs (643 lines) Drag-and-drop data handling. No tests.

  5. skill-tools/src/parse/coerce.rs (300 lines) / extract.rs (474 lines) Tool argument coercion and extraction from LLM output. Some tests exist in adjacent files but these specific modules lack coverage.

  6. skill-tools/src/exec/ — Tool execution layer

    • tools_fs.rs (365 lines) — file read/write/edit tools
    • tools_system.rs (381 lines) — bash, date tools
    • tools_web.rs (416 lines) — web search/fetch tools
    • safety.rs (166 lines) — path validation, sandboxing
    • truncate.rs (98 lines) — output truncation

Medium Priority

  1. skill-llm/src/engine/tool_orchestration.rs (891 lines) Multi-round tool-calling loop. Tested indirectly via E2E but no unit tests for edge cases: max rounds, error recovery, parallel tool calls.

  2. skill-llm/src/engine/sampling.rs (229 lines) Temperature, top-p, repetition penalty. No unit tests.

  3. skill-llm/src/catalog/download.rs (518 lines) Model download, resume, shard assembly. No unit tests (only E2E).

  4. skill-commands/src/graph/svg.rs (927 lines) + svg_3d.rs (377 lines) EEG graph SVG rendering. Complex math, no tests.

  5. skill-settings/src/screenshot_config.rs (184 lines) Screenshot configuration. No tests.

Low Priority (platform-specific / hard to unit test)

  1. skill-screenshots/src/capture.rs (1357 lines) / platform.rs (550 lines) Screen capture — requires display server, platform APIs.

  2. skill-headless/src/engine.rs (1193 lines) Headless inference engine — needs mock LLM server.

  3. skill-calendar — All platform backends need integration tests.

  4. skill-tts — TTS engines need audio hardware or mocking.

Frontend — Coverage

✅ Well-Tested (28 test files, 748 tests)

Tested modules: calibration, chat-types, chat-utils, compare-logic, compare-types, constants, dashboard-logic, devices-logic, electrodes, format, goals-logic, graph3d-logic, history-helpers, hooks-logic, i18n, llm-helpers, llm-tab-logic, markdown-normalize, onboarding-logic, screenshots-logic, search-interactive-logic, search-logic, search-types, sleep-analysis, theme, umap-helpers, umap-viewer-logic, utils.

⚠️ Untested Frontend Modules (>50 lines)

File Lines What it does
compare-canvas.ts 493 EEG comparison canvas rendering
history-canvas.ts 350 Session history canvas rendering
stores/theme.svelte.ts 338 Theme store (partial — theme.test.ts exists)
stores/chart-colors.svelte.ts 205 Chart color palette logic
stores/titlebar.svelte.ts 106 Titlebar state management
types.ts 185 Shared type definitions
use-canvas.ts 130 Canvas rendering hook

Recommendations

Quick Wins (highest value per effort)

  1. skill-history lib.rs — Add session listing/filtering tests (pure logic, no I/O)
  2. skill-data session_csv/parquet — Round-trip tests with tempdir
  3. skill-tools exec/ safety.rs — Path validation tests (security-critical)
  4. skill-tools parse/ coerce+extract — More edge cases for LLM output parsing
  5. skill-data screenshot_store — Search, pagination, embedding update tests

Test Infrastructure Improvements

  • cargo-llvm-cov — Add code coverage reporting to CI
  • Property testing — Expand proptest usage (currently only in skill-eeg and skill-tools)
  • Snapshot tests — For SVG graph rendering (skill-commands)
  • Mock LLM — Lightweight mock for tool_orchestration unit tests