Lap Tran
f32f98575b
CI / build (push) Has been cancelled
runnable-test-suite (parent) — complete. Three sub-tasks all complete: - make-tests-runnable: add requirements.txt pinning pytest==7.4.4, sweep all docs/prompts from bare 'python' to 'python3' (stock macOS/Windows ships python3), add idempotent .venv install block to scripts/install.sh, and add 'from __future__ import annotations' to 3 dashboard modules using PEP 604 union syntax at definition time so they import on Python 3.9+. The PEP 604 bug was caught by the streak verifier itself during implementation. - vram-detect-cross-platform: scripts/vram_detect.py now branches on platform.system() for Linux/Darwin/Windows. macOS path uses system_profiler SPDisplaysDataType (Apple Silicon unified memory via sysctl hw.memsize; Intel Macs via 'VRAM (Total):'). Windows uses wmic path win32_VideoController get AdapterRAM with PowerShell fallback. Linux /proc/meminfo and nvidia-smi/lspci paths unchanged (regression test locks them). Added 14 local-LLM context-window entries (llama-3.1, qwen2.5, mistral, deepseek-r1/v3, glm-4/4.5, gemma-2, phi-3/4) with source-cited model cards. Added _probe_ollama_model() that runs 'ollama list' as a last-resort fallback. run_command() now wraps PowerShell cmdlets on Windows (['powershell', '-NoProfile', '-NoLogo', '-Command', ...]). - vram-detect-cross-platform-tests: 11 new monkeypatched tests in tests/test_vram_detect.py covering Linux/Darwin/Windows branches for detect_ram and detect_gpu_vram, prefix-match for unknown model names, ollama probe, Windows PowerShell wrapper, and a LOCKED regression test for _detect_ram_linux(). All external subprocess/sysctl/wmic calls are mocked; no live hardware probes. Suite total: 235 passed, 0 errors. Verified on this box: gpu_vram_gb 0 -> 32 on Apple M5 (32GB unified memory), target context correctly jumped 12k -> 42k. Subtask-2 implementation was authored by local LLM (gemma-4-26B-A4B-it via headroom proxy @ localhost:8787). The 10-consecutive-clean-pass streak verifier ran as the independent checker model (article #2/#9/#13 in 'WTF Is a Loop? Part 2'). One anti-spin rail fired: local LLM produced inline branches where subtask-3 tests expected private _detect_ram_linux() helper; extracted helper to match the test contract without weakening tests. Parent + all 3 subtasks complete. Prior opencode-subagent implementation of subtask-2 preserved in git stash for reference.
91 lines
4.7 KiB
Markdown
91 lines
4.7 KiB
Markdown
# IMPLEMENTATION — vram-detect-cross-platform-tests
|
|
|
|
Parent: `runnable-test-suite` (see PARENT_SPEC.md).
|
|
SPEC: `tasks/runnable-test-suite/subtasks/vram-detect-cross-platform-tests/SPEC.md`.
|
|
|
|
## File touched
|
|
|
|
- `tests/test_vram_detect.py` — extended with 11 new test functions + helpers.
|
|
No other file was modified.
|
|
|
|
## Test functions added (11)
|
|
|
|
| # | Function | What it verifies |
|
|
|---|----------|------------------|
|
|
| 1 | `test_detect_ram_linux` | `/proc/meminfo` parse → `(16_384_000, 8_192_000)` via `detect_ram()` with `platform.system()` patched to `Linux`. |
|
|
| 2 | `test_detect_ram_macos` | `sysctl -n hw.memsize` → `"34359738368"` → total_kb `33_554_432`, available == total. |
|
|
| 3 | `test_detect_ram_windows` | `wmic ComputerSystem` → `TotalPhysicalMemory=34359738368` → total_kb `33_554_432`. |
|
|
| 4 | `test_detect_gpu_vram_nvidia_linux` | `nvidia-smi` → `"24576\n"` → `(25_165_824, 25_165_824, 1)`. |
|
|
| 5 | `test_detect_gpu_vram_apple_silicon` | `system_profiler` Apple M2 snippet + `sysctl hw.memsize=17179869184` → total > 0, num >= 1. |
|
|
| 6 | `test_detect_gpu_vram_windows_wmic` | `wmic win32_VideoController` → `AdapterRAM=8589934592` → total_vram_kb `8_388_608`. |
|
|
| 7 | `test_lookup_model_context_unknown_returns_zero` | `_lookup_model_context("completely-unknown-model")` → `0`. |
|
|
| 8 | `test_lookup_model_context_prefix_match` | `deepseek-r1:7b` → `64_000`; `llama-3.1-8b-instruct` → `128_000`. |
|
|
| 9 | `test_detect_model_context_ollama_probe` | No config model; `ollama list` returns `llama-3.1-8b` → context `128_000`. |
|
|
| 10 | `test_run_command_windows_powershell_wrapper` | On Windows, `Get-CimInstance ...` is routed through `powershell -NoProfile -NoLogo -Command "..."`. |
|
|
| 11 | `test_detect_ram_linux_regression` (LOCKED) | Exact match `(32_768_000, 16_384_000)` for a 32GB `/proc/meminfo` fixture via `_detect_ram_linux()`. |
|
|
|
|
Counting parametrize cases: **11 new tests** (no parametrization used; each function
|
|
is a single case).
|
|
|
|
## Existing tests modified
|
|
|
|
None. All 9 pre-existing test functions (`test_lookup_model_context`,
|
|
`test_parse_token_value`, `test_extract_value`, `test_parse_config_model`,
|
|
`test_parse_config_model_skips_code_blocks`, `test_parse_vram_config_manual`,
|
|
`test_recommend_context_api_model`, `test_recommend_context_manual_mode`,
|
|
`test_extract_model_from_file_respects_10kb_limit`) were preserved byte-for-byte.
|
|
No diffs to existing code.
|
|
|
|
## Mocking strategy
|
|
|
|
Every external call is mocked via `monkeypatch.setattr` — no live subprocess,
|
|
`system_profiler`, `nvidia-smi`, or `wmic` invocation:
|
|
|
|
- `vram.platform.system` → lambda returning the target OS string.
|
|
- `vram.subprocess.run` → `_make_fake_run(responses)` mapping `cmd[0]` (or the
|
|
joined PowerShell string) to a canned `_FakeResult(stdout, returncode=0)`.
|
|
- `vram.shutil.which` → lambda returning a truthy name (or `None` for non-ollama
|
|
in the ollama probe test).
|
|
- `Path.exists` / `Path.read_text` → `_patch_meminfo` serves the fixture only
|
|
for `/proc/meminfo` and falls through to the original for any other path
|
|
(keeps `tmp_path` and pytest internals working during the test).
|
|
- `vram.Path.home` → `tmp_path` in the ollama probe test so the real
|
|
`~/.automaton/config.md` is never consulted.
|
|
|
|
Module-level multiline string fixtures: `MEMINFO_LINUX_16GB`,
|
|
`MEMINFO_LINUX_32GB`, `APPLE_M2_PROFILER`, `WMIC_VIDEOCONTROLLER`,
|
|
`WMIC_COMPUTERSYSTEM`, `OLLAMA_LIST`.
|
|
|
|
## Acceptance criteria
|
|
|
|
| Criterion | Result |
|
|
|-----------|--------|
|
|
| `pytest tests/test_vram_detect.py -v` exits 0, all new tests pass | PASS — 20/20 (9 existing + 11 new) |
|
|
| `pytest tests/ -v` exits 0 | PASS — 235 passed, 0 errors |
|
|
| Streak: 10 consecutive clean `pytest tests/ -v` runs | PASS — `STREAK_COMPLETE attempt=1 clean=10/10` |
|
|
| No live subprocess against real hardware | PASS — every `subprocess.run` / `shutil.which` / `Path` I/O patched |
|
|
| Every new test < 500ms | PASS — entire file 0.01s; slowest 60 durations < 0.005s |
|
|
| No `pytest.mark.skip` | PASS — none used |
|
|
| No new deps | PASS — stdlib + pytest only |
|
|
| Did NOT touch `scripts/vram_detect.py` | CONFIRMED — only `tests/test_vram_detect.py` edited |
|
|
|
|
## Final pass counts
|
|
|
|
- Baseline (`tests/test_vram_detect.py`): 9 passed.
|
|
- After implementation (`tests/test_vram_detect.py`): 20 passed (+11).
|
|
- Full suite baseline: 224 passed.
|
|
- Full suite after: **235 passed** (+11), 0 errors, 0 skipped.
|
|
|
|
## Streak result
|
|
|
|
```
|
|
STREAK_COMPLETE attempt=1 clean=10/10
|
|
```
|
|
|
|
## STOP-and-report triggers hit
|
|
|
|
None. All 11 SPEC-required tests passed against subtask-2's `vram_detect.py`
|
|
without any signature mismatch. No edits to `scripts/vram_detect.py` were
|
|
required or made. The `ollama list` probe (SPEC test 9) is present in
|
|
`vram_detect.py:525-539` and works as specified.
|