| name | check-install |
| description | Verify that GAFIME is correctly installed and all components are working. Use when the user just installed GAFIME, wants to verify their setup, asks "is GAFIME working", "test my installation", "health check", "verify install", "check if GPU is working with GAFIME", or encounters import errors after installation. |
Installation Health Check
Run a comprehensive verification of the GAFIME installation.
Instructions
-
Run the health check script:
python .claude/skills/check-install/scripts/health_check.py
-
The script performs these checks in order:
- Python version compatibility (3.10+)
- Core imports (
gafime, polars)
- Rust helper alias (
from gafime import subfunctions)
- Optional imports (
scikit-learn)
- Backend availability (vendor GPU payloads when installed, C++ core)
- Vendor payload package status (
gafime-cuda, gafime-rocm)
- CLI functionality (
gafime --init can run)
- Functional test: runs a tiny synthetic analysis end-to-end
- Discrete function test: enables
enable_discrete_functions=True
- Throughput benchmark: measures combinations per second
-
Present results as a clear checklist with pass/fail status.
-
If any check fails, provide the specific fix:
- Python too old: "Upgrade to Python 3.10+"
- Missing polars: "pip install polars"
- Missing subfunctions: "Rebuild or reinstall the native Rust helper extension"
- Missing sklearn: "pip install gafime[sklearn]"
- Missing CUDA payload on an NVIDIA system:
pip install "gafime[cuda]"
- Missing ROCm payload on an AMD Linux x86_64 system:
pip install "gafime[rocm]"
- AMD system on Windows: use C++ Core in v0.4.7; ROCm payload wheels are Linux-only in this release
- Backend not loading: Refer to the
troubleshoot-backend skill
- Functional test fails: Likely a packaging issue, try reinstalling
- Discrete function test fails: Check that v0.4.7 Python files and native helpers match
- CUDA selector missing: Rebuild native extensions so
gafime_discrete_selection_adaptive_cuda is present
- Missing template batches: Rebuild the Rust helper so
BatchScheduler.create_batches is present
Example
User says: "I just did pip install gafime, is it working?"
Actions: Run health_check.py
Result:
GAFIME Health Check
[PASS] Python 3.11.5 (>= 3.10 required)
[PASS] gafime v0.4.7 imported
[PASS] Rust subfunctions alias
[PASS] polars 0.20.31
[PASS] scikit-learn 1.4.0
[PASS] CUDA payload recommendation (installed)
[SKIP] Metal backend (not macOS arm64)
[PASS] C++ core backend
[PASS] Functional test: signal detected in synthetic data
[PASS] Discrete function test
[PASS] Throughput: 18,432 combinations/sec
All checks passed! GAFIME is ready to use.
