← Back to README · Architecture Guide · Benchmarks Matrix · Roadmap
Every component in intel-arc-models-runner is governed by strict Triad Testing (Positive, Negative, and Edge-Case scenarios) derived from Claude Global Rules (33-test-case-roadmap.md, 35-positive-testing-standards.md, 36-negative-testing-standards.md, 37-edge-case-testing-standards.md, 38-test-mocking-strategy.md).
As of the latest release, the test suite contains 569 passing automated tests across 36 test modules with 0 failures and 0 errors.
| # | Test Module | Tests | Coverage Area | Invariant / Proof Referenced |
|---|---|---|---|---|
| 1 | test_devices.py |
8 | Level Zero & OpenVINO hardware discovery & device prober | Level Zero API enumeration |
| 2 | test_dll_loader.py |
6 | Windows DriverStore DLL lazy loader & path discovery | Level Zero DLL loading |
| 3 | test_engine.py |
28 | OpenVINO GenAI execution engine, streaming, KV-cache | Proof M61 (Transformer roofline) |
| 4 | test_speculative.py |
20 | Dual-silicon speculative decoding (NPU draft + Arc verify) | Proof M6 & Leviathan equation |
| 5 | test_aot_cache.py |
8 | Ahead-of-Time NPU compilation & warm-load speedup | Proof M93 & M112 |
| 6 | test_benchmark.py |
12 | TTFT/TPOT measurement, warmup discard, Student's t CI | Proof M65 & M81 |
| 7 | test_hub.py |
12 | Model Hub registry, downloader, manifest verification | EMA bandwidth smoothing |
| 8 | test_gguf_parser.py |
10 | Binary GGUF v1/v2/v3 parsing, tensor extraction | GGUF container specification |
| 9 | test_ollama_importer.py |
8 | Ollama model discovery & OpenVINO profile mapping | Ollama manifest parsing |
| 10 | test_server.py |
24 | FastAPI Ollama + OpenAI parity endpoints | HTTP API contracts |
| 11 | test_embeddings.py |
13 | NPU embeddings, L2 norm, mean pooling, cosine similarity | Proof M1, M2, M14 |
| 12 | test_vision.py |
9 | VLMPipeline vision engine, visual token calculation | Proof M5 (Visual token roofline) |
| 13 | test_transpiler.py |
18 | GGUF→OpenVINO transpiler, INT4 dequant, SQNR, XMX pack | Proof M7, M8, M9, M21 |
| 14 | test_transpiler_stream.py |
14 | SSE streaming engine, O(1) RAM, SIMD quantization | Proof M15–M20, P1–P6 |
| 15 | test_continuous_batcher.py |
9 | PagedBlockTable, RadixCache, Little's Law admission | Proof M11, M12, M13, M30 |
| 16 | test_batch_pipeline.py |
16 | Continuous batch pipeline, Little's Law, dynamic reclamation | Proof M28–M35 |
| 17 | test_batch_api.py |
10 | Batch status API, preemption routes, 429 saturation | Little's Law saturation |
| 18 | test_ui_endpoints.py |
24 | Web UI SPA, SSE stream endpoints, embeddings test, convert | Web UI API contracts |
| 19 | test_lora.py |
16 | Multi-LoRA manager, safetensors zero-copy, LRU cache | Proof M24, M26 |
| 20 | test_lora_api.py |
10 | LoRA REST API endpoints, hot-swapping dispatch, isolation | Proof M23, M27 |
| 21 | test_real_hardware.py |
21 | Real hardware integration (Arc Xe-LPG, AI Boost NPU, USM) | Proof M36–M43 |
| 22 | test_audio_engine.py |
32 | OpenVINO Whisper ASR & Kokoro TTS, VAD silence gating | Proof M59–M72 |
| 23 | test_image_engine.py |
24 | OpenVINO Text2Image (SD 1.5, LCM, SD-Turbo, SDXL), DPAS | Proof M73–M93 |
| 24 | test_video_engine.py |
24 | OpenVINO Text2Video (LTX-Video, CogVideoX, AnimateDiff) | Proof M94–M114 |
| 25 | test_multimodal_orchestrator.py |
5 | End-to-end Studio DAG pipeline, heterogeneous scheduling | Multi-engine fork-join |
| 26 | test_studio_director.py |
8 | Director prompt decomposition, regex AST auto-repair | Dual-stage AST repair |
| 27 | test_studio_models.py |
14 | Pydantic v2 schemas, telemetry serialization | Pydantic v2 invariants |
| 28 | test_studio_workers.py |
7 | AudioWorker (Kokoro TTS), VisualWorker (Compositor/DiT) | USM zero-copy bridge |
| 29 | test_studio_muxer.py |
5 | Closed-loop pacing convergence, hold-frame padding | Audio-video sync convergence |
| 30 | test_studio_routes.py |
5 | /v1/studio/generate REST endpoint, SSE pipeline events |
SSE streaming protocol |
| 31 | test_studio_ui_triad.py |
33 | 3-tier Triad matrix for Tab 7 Studio Web UI | Triad (Pos, Neg, Edge) |
| 32 | test_slide_compositor_triad.py |
38 | 16:9 HD Slide Compositor, Double-Slit, Bohr atom, Voronoi | Dynamic physics & typography |
| 33 | test_nextgen_orchestration_triad.py |
42 | Milestones 14.1–14.7 (QSV, Alpha dissolve, Reactive DAG) | Proofs M7–M12 |
| 34 | test_pypi_packaging_triad.py |
11 | Standalone wheel packaging, PEP 561, dual CLI entrypoints | Proof M115 (Wheel Deflate Invariant) |
| 35 | test_docker_container_triad.py |
11 | Ubuntu 24.04 multi-stage Dockerfile, Rule 29, DRI /dev/dri |
Proofs M116–M120 |
| 36 | test_flash_stream_triad.py |
14 | Flash-Stream O(1) RAM streaming, NPU/GPU disaggregation, SwiGLU | Proofs M121–M134 |
| Total | 36 Modules | 569 | 100% Green · 0 Failures · 0 Errors | Complete Invariant Verification |
python -m pytest -v --tb=short
# → 533 passed in ~24.5spython -m pytest tests/integration/test_real_hardware.py -v
# → 21 passed in 4.82s on physical Intel Core Ultra siliconpython -m pytest tests/test_studio_ui_triad.py tests/test_slide_compositor_triad.py -vpython -m pytest tests/test_nextgen_orchestration_triad.py -varc-runner test-hardware --iterations 10