Complete API Reference
100% Full Class, Struct, and Method Documentation
Python SDK Core API Reference
| Module / Function | Signature | Description |
|---|---|---|
termux_vision.vlm.load |
load(model_id, device='auto', ngl=None) |
Context manager initializing multimodal VLM engine with 5-backend hardware offload. |
VLMEngine.describe |
describe(image_path, prompt, quality='optimal', max_tokens=150) |
Executes multimodal visual inference and returns VLMResponse with token metrics. |
termux_vision.cv.canny |
canny(image_gray, low_threshold=40, high_threshold=120, backend='auto') |
Hardware-accelerated Canny edge detector: 100% Vulkan GPU (0.23ms) or ARM64 NEON (3.02ms). |
termux_vision.installer.ensure_installed |
ensure_installed(force=False, from_source=False, dry_run=False) |
Idempotent prebuilt installer provisioning native C++/Vulkan acceleration binaries. |
termux_vision.detect.detect_faces |
detect_faces(image, score_threshold=0.70, backend='auto') |
UltraFace SSD Deep Learning Neural Face Detection (11.94ms, 100% false-positive elimination) with automatic face margin padding. |
termux_vision.cli.doctor.run_doctor |
run_doctor(probe_vulkan=True) |
Probes Vulkan ICD driver, RAM headroom, big.LITTLE CPU topology, and native binary health. |