# lynx SDK

On-device computer-vision SDK with bindings for Swift (iOS/macOS), Kotlin (Android), Java, Python, and C. Load a small signed `.lnx` model and run inference locally — detection, segmentation, 17-keypoint pose, depth, oriented boxes, OCR, tracking — no server round-trip, no GPU required. The docs below are agent-grade: exact APIs and complete, paste-and-run integration recipes. Start at index.md, then the recipe for your platform+task.

## Docs
- [api/c.md](/sdk-docs/api/c.md): VENDORED — generated from LYNX-SDK origin/main :docs/api/c.md 
- [api/conventions.md](/sdk-docs/api/conventions.md): Fallible calls throw in every binding . You never check a return code by hand. The thrown error carries:
- [api/java.md](/sdk-docs/api/java.md): VENDORED — generated from LYNX-SDK origin/main :docs/api/java.md 
- [api/kotlin.md](/sdk-docs/api/kotlin.md): VENDORED — generated from LYNX-SDK origin/main :docs/api/kotlin.md 
- [api/python.md](/sdk-docs/api/python.md): VENDORED — generated from LYNX-SDK origin/main :docs/api_reference.md 
- [api/swift.md](/sdk-docs/api/swift.md): VENDORED — generated from LYNX-SDK origin/main :docs/api/swift.md 
- [index.md](/sdk-docs/index.md): lynx is an on-device computer-vision SDK: you load a small signed model  and run inference locally — no server round-trip, no GPU required. One model can expose
- [install/android.md](/sdk-docs/install/android.md): Add the Gradle dependency  with google + mavenCentral repos:
- [install/c.md](/sdk-docs/install/c.md): Download the C SDK  from a build run and compile against it:
- [install/ios.md](/sdk-docs/install/ios.md): Add the Swift Package :
- [install/java.md](/sdk-docs/install/java.md): Add the dependency :
- [install/python.md](/sdk-docs/install/python.md): bash
- [models/catalog.md](/sdk-docs/models/catalog.md): Two models are available. Both are keyless , production version 2.0, and ship for linux-x64 , ios-arm64 , android-arm64-v8a , input 640×640.
- [models/no-model.md](/sdk-docs/models/no-model.md): You're on this path when the task needs a class or capability the shipped models don't cover, e.g.:
- [recipes/android-detection.md](/sdk-docs/recipes/android-detection.md): Goal: load a model, run detection on a Bitmap, get boxes + labels + scores. Model: lynx-basic ).
- [recipes/c-detection.md](/sdk-docs/recipes/c-detection.md): Goal: load a model, run detection on an image, print each detection's class name + score + box. Model: lynx-basic ).
- [recipes/ios-detection.md](/sdk-docs/recipes/ios-detection.md): Goal: load a model, run detection on an image, get boxes + labels + scores. Model: lynx-basic ).
- [recipes/java-detection.md](/sdk-docs/recipes/java-detection.md): Goal: load a model, run detection on an image, get boxes + labels + scores. Model: lynx-basic ).
- [recipes/python-classification.md](/sdk-docs/recipes/python-classification.md): Goal: run a whole-image classification head — "what is this image?" , not "where are the objects?".
- [recipes/python-depth.md](/sdk-docs/recipes/python-depth.md): Goal: run detection and depth in one pass, read the dense depth map plus a per-object depth, and turn depth into a picture you can display. Model: lynx-basic ).
- [recipes/python-detection.md](/sdk-docs/recipes/python-detection.md): Goal: load a model, run detection on an image, get boxes + labels + scores. Model: lynx-basic ).
- [recipes/python-ocr.md](/sdk-docs/recipes/python-ocr.md): LYNX does OCR as detection: each character is a normal detection whose class is
- [recipes/python-server.md](/sdk-docs/recipes/python-server.md): Goal: run lynx-basic behind an HTTP endpoint so a client that can't run the SDK on-device — a React Native / JS app, a Go/Rust/Java service, plain curl — can PO
- [recipes/python-streaming.md](/sdk-docs/recipes/python-streaming.md): Goal: open a webcam, run tracked detection on every frame in a loop, and print the detections  per frame. Model: lynx-basic ).
- [recipes/python-tracking.md](/sdk-docs/recipes/python-tracking.md): Goal: run detection across a video/frame sequence and follow each object with a stable id as it moves frame to frame. Model: lynx-basic ).
- [support.md](/sdk-docs/support.md): LYNX is one C inference core with a thin, idiomatic binding per language — the same models and the same open → predict → detections shape everywhere, from a Swi