### Setup Environment

Source: https://docs.liquid.ai/examples/customize-models/home-assistant

Clone the repository, install dependencies, and authenticate with HuggingFace and Modal.

```bash
git clone https://github.com/Liquid4All/leap-finetune
cd leap-finetune && uv sync && cd -
huggingface-cli login
modal setup
```

--------------------------------

### Clone Repository and Navigate

Source: https://docs.liquid.ai/examples/android/leap-koog-agent

Clone the example repository and navigate to the LeapKoogAgent directory to begin setup.

```bash
git clone https://github.com/Liquid4All/LeapSDK-Examples.git
cd LeapSDK-Examples/Android/LeapKoogAgent
```

--------------------------------

### Setup Python environment

Source: https://docs.liquid.ai/examples/laptop-examples/audio-car-cockpit

Clones the repository and installs necessary dependencies for the audio car cockpit demo.

```bash
git clone https://github.com/Liquid4All/cookbook.git
cd cookbook/examples/audio-car-cockpit
make setup
```

--------------------------------

### Install and Authenticate Environment

Source: https://docs.liquid.ai/examples/customize-models/satellite-vlm

Initial setup for the project environment, including dependency synchronization, Modal authentication, and Hugging Face login.

```bash
uv sync
uv run python -m modal setup
uv run huggingface-cli login
```

--------------------------------

### Complete Tool Use Example Setup

Source: https://docs.liquid.ai/lfm/key-concepts/tool-use

Initializes the model and tokenizer, defines tools using JSON, and sets up the conversation messages for tool use.

```python
import json
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "LiquidAI/LFM2.5-1.2B-Instruct",
    dtype="bfloat16",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("LiquidAI/LFM2.5-1.2B-Instruct")

tools = [{
    "name": "get_candidate_status",
    "description": "Retrieves the current status of a candidate in the recruitment process",
    "parameters": {
        "type": "object",
        "properties": {
            "candidate_id": {
                "type": "string",
                "description": "Unique identifier for the candidate",
            }
        },
        "required": ["candidate_id"],
    },
}]

messages = [
    {"role": "system", "content": f"List of tools: {json.dumps(tools)}"},
    {"role": "user", "content": "What is the current status of candidate ID 12345?"},
]
```

--------------------------------

### Install, Launch, and Query Sglang

Source: https://docs.liquid.ai/lfm/models/lfm2-350m

Setup for Sglang server and querying via the OpenAI-compatible API.

```bash
uv pip install "sglang>=0.5.10"
```

```bash
(toolCallParser ? `sglang serve \
    --model-path ${modelId} \
    --host 0.0.0.0 \
    --port 30000 \
    --tool-call-parser ${toolCallParser}` : `sglang serve \
    --model-path ${modelId} \
    --host 0.0.0.0 \
    --port 30000`)
```

```python
from openai import OpenAI

client = OpenAI(base_url="http://localhost:30000/v1", api_key="None")

response = client.chat.completions.create(
    model="${modelId}",
    messages=[{"role": "user", "content": "What is machine learning?"}],
    temperature=0.3,
)

print(response.choices[0].message.content)
```

--------------------------------

### Clone Repository and Setup Environment

Source: https://docs.liquid.ai/examples/customize-models/car-maker-identification

Clones the Liquid AI cookbook repository, navigates into the example directory, and synchronizes Python dependencies using uv.

```bash
git clone https://github.com/Liquid4All/cookbook.git
cd cookbook/examples/car-maker-identification
uv sync
```

--------------------------------

### Complete Custom Model Bundling Workflow

Source: https://docs.liquid.ai/deployment/tools/model-bundling/quick-start

This example demonstrates the complete workflow for bundling custom models, including installation, authentication, bundle creation, monitoring, and downloading.

```sh
pip install leap-bundle
leap-bundle login <api-key>
leap-bundle whoami

leap-bundle create <model-directory>

leap-bundle list

leap-bundle download <request-id>
```

--------------------------------

### Install Dependencies

Source: https://docs.liquid.ai/examples/laptop-examples/lfm2-english-to-korean

Install the project's dependencies using the 'uv sync' command. Ensure you have uv installed.

```bash
uv sync
```

--------------------------------

### Clone and Navigate to Example Repository

Source: https://docs.liquid.ai/examples/android/vision-language-model-example

Initial commands to retrieve the VLM example project from GitHub.

```bash
git clone https://github.com/Liquid4All/LeapSDK-Examples.git
cd LeapSDK-Examples/Android/VLMExample
```

--------------------------------

### Install and Run vLLM for LFM2-24B-A2B

Source: https://docs.liquid.ai/lfm/models/lfm2-24b-a2b

Install vLLM using pip. Then, load the model and run inference with specified sampling parameters. This example demonstrates a basic chat interaction.

```bash
pip install vllm==0.14
```

```python
from vllm import LLM, SamplingParams

llm = LLM(model="${modelId}")

sampling_params = SamplingParams(${samplingParams}max_tokens=512)

output = llm.chat("What is machine learning?", sampling_params)
print(output[0].outputs[0].text)
```

--------------------------------

### SGLang Quick Start

Source: https://docs.liquid.ai/lfm/models/lfm2-350m

This example shows how to use SGLang to interact with the LFM2-350M model. SGLang provides a Pythonic interface for defining and executing language model workloads.

```python
from lang import:

app = Lang<"LiquidAI/LFM2-350M">

response = app.chat(
    [
        {"role": "user", "content": "Project: /websites/liquid_ai\nContent:\n    [{"role": "user", "content": "What is machine learning?"}],"}
    ]
)
print(response)
```

--------------------------------

### Run Hand & Voice Racer locally

Source: https://docs.liquid.ai/examples/web/hand-voice-racer

Commands to install dependencies and start the development server for the project.

```bash
npm install
npm run dev
```

--------------------------------

### Launch the demo

Source: https://docs.liquid.ai/examples/laptop-examples/audio-car-cockpit

Starts the audio server and the web interface concurrently.

```bash
make -j2 audioserver serve
```

--------------------------------

### llama.cpp Quick Start

Source: https://docs.liquid.ai/lfm/models/lfm2-350m

Integrate the LFM2-350M model with llama.cpp for efficient inference. This example shows how to set sampling parameters for text generation.

```bash
./main -m LiquidAI/LFM2-350M-GGUF --temp 0.3 --min-p 0.15 --repeat-penalty 1.05 -p "Project: /websites/liquid_ai\nContent:\n    [{"role": "user", "content": "What is machine learning?"}],"
```

--------------------------------

### Install Dependencies

Source: https://docs.liquid.ai/examples/web/audio-webgpu-demo

Install the required project dependencies.

```bash
npm install
```

--------------------------------

### Install Dependencies and Push Model

Source: https://docs.liquid.ai/deployment/gpu-inference/baseten

Navigate to the cloned directory and install dependencies, then push the LFM2-8B model to Baseten.

```bash
cd bastenpip install trusstruss push lfm2-8b --publish
```

--------------------------------

### Install liquid-audio

Source: https://docs.liquid.ai/lfm/models/lfm2-audio-1.5b

Install the liquid-audio library and optional dependencies for demos and flash attention.

```bash
pip install liquid-audio
```

```bash
pip install "liquid-audio[demo]"  # optional, for demo dependencies
```

```bash
pip install flash-attn --no-build-isolation  # optional, for flash attention 2
```

--------------------------------

### LEAP Bundle CLI usage examples

Source: https://docs.liquid.ai/deployment/tools/model-bundling/bundle-creation

Various examples for creating bundles, including standard, JSON-formatted, and quantized output.

```sh
# Create bundle from directory
leap-bundle create ./my-model-directory

# Example output on success
ℹ Calculating directory hash...
ℹ Submitting bundle request...
✓ Bundle request created with ID: 18734
ℹ Starting upload...
Uploading directory... ✓
✓ Upload completed successfully! Request ID: 18734

# Example output when request already exists
⚠ A bundle request with the same input hash already exists: req_xyz789abc123

# Create bundle with JSON output
leap-bundle create ./my-model-directory --json

# Example JSON output on success
{"request_id": "18734", "status": "success"}

# Example JSON output when request already exists
{"error": "A bundle request with the same input hash already exists: req_xyz789abc123", "status": "exists"}

# Create GGUF bundle with specific quantization
leap-bundle create ./my-model-directory --quantization Q8_0
```

--------------------------------

### Install vLLM

Source: https://docs.liquid.ai/deployment/gpu-inference/vllm

Commands to install vLLM or pull the prebuilt Docker image.

```bash
uv pip install vllm==0.14
```

```bash
docker pull vllm/vllm-openai:latest
```

--------------------------------

### Install Weights & Biases

Source: https://docs.liquid.ai/examples/customize-models/car-maker-identification

Installs the Weights & Biases Python package using uv. Ensure you have uv installed and are within your virtual environment.

```bash
uv add wandb
```

--------------------------------

### Install and Run Transformers

Source: https://docs.liquid.ai/lfm/models/lfm2-350m

Installation and model loading using the Hugging Face Transformers library.

```bash
pip install "transformers>=5.2.0" torch accelerate
```

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "${modelId}"
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    dtype="bfloat16",
)
tokenizer = AutoTokenizer.from_pretrained(model_id)

inputs = tokenizer.apply_chat_template(
```

--------------------------------

### Install Transformers and PyTorch

Source: https://docs.liquid.ai/lfm/help/contributing

Install the necessary libraries for working with transformer models.

```bash
pip install transformers torch
```

--------------------------------

### Clone and build the project

Source: https://docs.liquid.ai/examples/laptop-examples/flight-search-assistant

Clone the repository and install the necessary project dependencies.

```bash
git clone https://github.com/Liquid4All/cookbook.git
cd cookbook/examples/flight-search-assistant
uv sync
```

--------------------------------

### Complete Model Download Management Example

Source: https://docs.liquid.ai/deployment/on-device/android/ai-agent-usage-guide

A comprehensive ViewModel example demonstrating how to check model status, download models if needed, manage storage, remove models, and cancel downloads. It integrates storage checks and provides user feedback via StateFlow.

```kotlin
class ModelManagementViewModel(application: Application) : AndroidViewModel(application) {
    private val downloader = LeapModelDownloader(application)

    private val _downloadStatus = MutableStateFlow<ModelDownloadStatus>(ModelDownloadStatus.NotOnLocal)
    val downloadStatus: StateFlow<ModelDownloadStatus> = _downloadStatus.asStateFlow()

    private val _errorMessage = MutableStateFlow<String?>(null)
    val errorMessage: StateFlow<String?> = _errorMessage.asStateFlow()

    // Check if model is already downloaded
    suspend fun checkModelStatus(modelSlug: String, quantizationSlug: String) {
        val status = downloader.queryStatus(modelSlug, quantizationSlug)
        _downloadStatus.value = status
    }

    // Download model if not cached
    fun downloadIfNeeded(modelSlug: String, quantizationSlug: String) {
        viewModelScope.launch {
            try {
                val status = downloader.queryStatus(modelSlug, quantizationSlug)

                if (status is ModelDownloadStatus.Downloaded) {
                    Log.d(TAG, "Model already downloaded")
                    return@launch
                }

                // Check storage before downloading
                if (getAvailableStorageGB() < 2) {
                    _errorMessage.value = "Insufficient storage space"
                    return@launch
                }

                // Start download
                downloader.downloadModel(
                    modelSlug = modelSlug,
                    quantizationSlug = quantizationSlug,
                    progress = { progressData ->
                        _downloadStatus.value = ModelDownloadStatus.DownloadInProgress(
                            progress = progressData.progress
                        )
                    }
                )

                _downloadStatus.value = ModelDownloadStatus.Downloaded

            } catch (e: Exception) {
                _errorMessage.value = "Download failed: ${e.message}"
                Log.e(TAG, "Download error", e)
            }
        }
    }

    // Remove model to free up space
    fun removeModel(modelSlug: String, quantizationSlug: String) {
        viewModelScope.launch {
            try {
                downloader.removeModel(modelSlug, quantizationSlug)
                _downloadStatus.value = ModelDownloadStatus.NotOnLocal
                Log.d(TAG, "Model removed")
            } catch (e: Exception) {
                _errorMessage.value = "Failed to remove model: ${e.message}"
            }
        }
    }

    // Cancel download
    fun cancelDownload(modelSlug: String, quantizationSlug: String) {
        downloader.cancelDownload(modelSlug, quantizationSlug)
        _downloadStatus.value = ModelDownloadStatus.NotOnLocal
    }

    private fun getAvailableStorageGB(): Long {
        val path = Environment.getDataDirectory()
        val stat = StatFs(path.path)
        val availableBytes = stat.availableBlocksLong * stat.blockSizeLong
        return availableBytes / (1024 * 1024 * 1024)
    }

    companion object {
        private const val TAG = "ModelManagement"
    }
}
```

--------------------------------

### Install make utility

Source: https://docs.liquid.ai/examples/customize-models/car-maker-identification

Installs the make utility on Debian-based systems using apt-get. This is a common build automation tool.

```bash
sudo apt-get install make
```

--------------------------------

### Launch SGLang Server

Source: https://docs.liquid.ai/lfm/models/lfm25-1.2b-base

Start the SGLang server to host the model.

```bash
sglang serve \
    --model-path LiquidAI/LFM2.5-1.2B-Base \
    --host 0.0.0.0 \
    --port 30000
```

--------------------------------

### Install PyLate

Source: https://docs.liquid.ai/lfm/models/lfm2-colbert-350m

Install the PyLate library using pip. This is the first step before using the LFM2-ColBERT-350M model for inference.

```bash
pip install -U pylate
```

--------------------------------

### Deploy vLLM Server

Source: https://docs.liquid.ai/deployment/gpu-inference/vllm

Methods for starting an OpenAI-compatible vLLM server.

```bash
vllm serve LiquidAI/LFM2.5-1.2B-Instruct \
    --host 0.0.0.0 \
    --port 8000 \
    --dtype auto
```

```bash
docker run --runtime nvidia --gpus all \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=$HF_TOKEN" \
    -p 8000:8000 \
    --ipc=host \
    vllm/vllm-openai:latest \
    --model LiquidAI/LFM2.5-1.2B-Instruct
```

--------------------------------

### Load Local Model Examples

Source: https://docs.liquid.ai/deployment/on-device/ios/model-loading

Provides examples for loading local models using the `Leap.load` function. One example demonstrates loading an ExecuTorch backend via a `.bundle` file, and the other loads a llama.cpp backend via a `.gguf` file.

```swift
// ExecuTorch backend via .bundle
let bundleURL = Bundle.main.url(forResource: "qwen3-0_6b", withExtension: "bundle")!
let runner = try await Leap.load(url: bundleURL)

// llama.cpp backend via .gguf
let ggufURL = Bundle.main.url(forResource: "qwen3-0_6b", withExtension: "gguf")!
let ggufRunner = try await Leap.load(url: ggufURL)
```

--------------------------------

### Install Modal

Source: https://docs.liquid.ai/examples/customize-models/home-assistant

Add the Modal package to your project and authenticate your environment.

```bash
uv add modal
```

```bash
uv run modal setup
```

--------------------------------

### Resume bundle request examples

Source: https://docs.liquid.ai/deployment/tools/model-bundling/bundle-creation

Examples of resuming uploads, including using specific IDs and JSON output format.

```sh
# Resume the latest failed upload
leap-bundle resume

# Example output on success
ℹ Retrieving latest bundle request...
✓ Resuming upload for request 18734...
ℹ Input path: /path/to/model/directory
ℹ Using parallel upload with multipart support...
✓ Upload completed successfully! Request ID: 18734

# Resume a specific request by ID
leap-bundle resume 18734

# Example output
ℹ Retrieving bundle request 18734...
✓ Resuming upload for request 18734...
ℹ Input path: /path/to/model/directory
ℹ Using parallel upload with multipart support...
✓ Upload completed successfully! Request ID: 18734

# Resume with JSON output
leap-bundle resume 18734 --json

# Example JSON output on success
{"request_id": "18734", "status": "success"}
```

--------------------------------

### Install Transformers dependencies

Source: https://docs.liquid.ai/deployment/gpu-inference/transformers

Install the necessary packages including transformers, torch, and accelerate using uv.

```bash
uv pip install "transformers>=5.2.0" torch accelerate
```

--------------------------------

### Install TRL and dependencies

Source: https://docs.liquid.ai/customization/finetuning-frameworks/trl

Install the core TRL library along with necessary dependencies for LoRA support and distributed training.

```bash
pip install trl>=0.9.0 transformers>=4.55.0 torch>=2.6 peft accelerate
```

--------------------------------

### Few-shot Prompting Example

Source: https://docs.liquid.ai/lfm/key-concepts/text-generation-and-prompting

Provide example interactions to guide the model's behavior. The model learns from the conversation history and applies patterns to new inputs.

```python
messages = [
    {"role": "system", "content": "You are a helpful assistant that formats dates."}, 
    {"role": "user", "content": "2024-01-15"}, 
    {"role": "assistant", "content": "January 15, 2024"}, 
    {"role": "user", "content": "2024-12-25"}, 
    {"role": "assistant", "content": "December 25, 2024"}, 
    {"role": "user", "content": "2024-03-08"}  # Model follows the pattern
]
```

--------------------------------

### Install leap-bundle CLI

Source: https://docs.liquid.ai/deployment/tools/model-bundling/quick-start

Install the Liquid Edge AI Platform (LEAP) model bundling tool using pip. Ensure you have Python 3.8 or higher.

```bash
pip install leap-bundle
```

--------------------------------

### Install and Run llama.cpp for LFM2-24B-A2B

Source: https://docs.liquid.ai/lfm/models/lfm2-24b-a2b

Install llama.cpp using Homebrew. Then, run the model using llama-cli, specifying the GGUF repository and context size. The -hf flag downloads the model from Hugging Face.

```bash
brew install llama.cpp
```

```bash
llama-cli -hf ${ggufRepo} -c 4096 --color -i \
    ${samplingFlags}
```

--------------------------------

### Install Weights & Biases

Source: https://docs.liquid.ai/examples/customize-models/home-assistant

Add the wandb package to your project and authenticate your account.

```bash
uv add wandb
```

```bash
uv run wandb login
```

--------------------------------

### Clone and Navigate to Repository

Source: https://docs.liquid.ai/examples/android/recipe-generator-constrained-output

Commands to clone the example repository and enter the project directory.

```bash
git clone https://github.com/Liquid4All/LeapSDK-Examples.git
cd LeapSDK-Examples/Android/RecipeGenerator
```

--------------------------------

### Clone and Setup Repository

Source: https://docs.liquid.ai/examples/web/audio-webgpu-demo

Commands to clone the cookbook repository and navigate to the audio-webgpu-demo directory.

```bash
git clone https://github.com/Liquid4All/cookbook.git
cd cookbook/examples/audio-webgpu-demo
```

--------------------------------

### Prefill for Structured Output

Source: https://docs.liquid.ai/lfm/key-concepts/text-generation-and-prompting

Start the model with a specific format or structure, such as a JSON opening brace, to guide it toward structured outputs.

```python
messages = [
    {"role": "system", "content": "Extract information and return as JSON."}, 
    {"role": "user", "content": "Extract the name and age from: John is 30 years old."}, 
    {"role": "assistant", "content": "{\n  \"name\": "}  # Prefill with JSON structure
```

--------------------------------

### Start Development Server

Source: https://docs.liquid.ai/examples/web/audio-webgpu-demo

Launch the local development server to run the application.

```bash
npm run dev
```

--------------------------------

### Install LiquidONNX

Source: https://docs.liquid.ai/deployment/on-device/onnx

Clone the repository and set up the environment using uv. Use the gpu extra for hardware-accelerated inference.

```bash
git clone https://github.com/Liquid4All/onnx-export.git
cd onnx-export
uv sync

# For GPU inference
uv sync --extra gpu
```

--------------------------------

### Prepare models

Source: https://docs.liquid.ai/examples/laptop-examples/audio-car-cockpit

Downloads the required audio and tool-calling models for the demo.

```bash
make LFM2.5-Audio-1.5B-GGUF LFM2-1.2B-Tool-GGUF
```

--------------------------------

### Multi-turn Conversation Prompting

Source: https://docs.liquid.ai/lfm/key-concepts/text-generation-and-prompting

Continue a conversation or provide example interactions to guide the model. The model learns from the conversation history and applies patterns to new inputs.

```python
messages = [
    {"role": "system", "content": "You are a helpful assistant."}, 
    {"role": "user", "content": "What are the benefits of exercise?"}, 
    {"role": "assistant", "content": "Exercise has many benefits including:\n1. Improved cardiovascular health\n2. "},  # Partial response to continue
    {"role": "user", "content": "Tell me more about cardiovascular health."}
```

--------------------------------

### Clone and prepare the repository

Source: https://docs.liquid.ai/examples/laptop-examples/invoice-extractor-tool-with-liquid-nanos

Download the cookbook repository and navigate to the invoice parser example directory.

```sh
git clone https://github.com/Liquid4All/cookbook.git
cd cookbook/examples/invoice-parser
```

--------------------------------

### Run LFM2-VL-1.6B with Transformers

Source: https://docs.liquid.ai/lfm/models/lfm2-vl-1.6b

Load and run the LFM2-VL-1.6B model using the Hugging Face Transformers library. This example demonstrates loading the model and processor, processing an image from a URL, and generating a response to a text query about the image. Ensure you have the necessary libraries installed.

```python
from transformers import AutoProcessor, AutoModelForImageTextToText
from transformers.image_utils import load_image

model_id = "LiquidAI/LFM2-VL-1.6B"
model = AutoModelForImageTextToText.from_pretrained(
    model_id,
    device_map="auto",
    dtype="bfloat16",
)

processor = AutoProcessor.from_pretrained(model_id)

url = "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
image = load_image(url)

conversation = [
    {
        "role": "user",
        "content": [
            {"type": "image", "image": image},
            {"type": "text", "text": "What is in this image?"},
        ],
    },
]

inputs = processor.apply_chat_template(
    conversation,
    add_generation_prompt=True,
    return_tensors="pt",
    return_dict=True,
    tokenize=True,
).to(model.device)

outputs = model.generate(**inputs, do_sample=True, temperature=0.1, min_p=0.15, repetition_penalty=1.05, max_new_tokens=256)
response = processor.batch_decode(outputs, skip_special_tokens=True)[0]
print(response)
```

--------------------------------

### Start the application server

Source: https://docs.liquid.ai/examples/customize-models/home-assistant

Launch the Uvicorn server to run the application locally.

```bash
uv run uvicorn app.server:app --port 5173 --reload
```

--------------------------------

### Clone the repository

Source: https://docs.liquid.ai/examples/customize-models/home-assistant

Download the cookbook repository to access the home assistant example files.

```bash
git clone https://github.com/Liquid4All/cookbook.git
cd cookbook/examples/home-assistant
```

--------------------------------

### Install uv on macOS/Linux

Source: https://docs.liquid.ai/examples/laptop-examples/meeting-summarization

Installs the uv package manager using a curl command. Ensure you have curl installed.

```bash
curl -LsSf https://astral.sh/uv/install.sh | sh
```

--------------------------------

### Install Ollama on Linux

Source: https://docs.liquid.ai/deployment/on-device/ollama

Use this command to install Ollama on Linux systems by downloading and executing the official installation script.

```bash
curl -fsSL https://ollama.com/install.sh | sh
```

--------------------------------

### Install llama.cpp via Homebrew

Source: https://docs.liquid.ai/deployment/on-device/llama-cpp

Use this command to install llama.cpp on macOS or Linux systems that have Homebrew package manager installed.

```bash
brew install llama.cpp
```

--------------------------------

### Install uv on Windows

Source: https://docs.liquid.ai/examples/laptop-examples/meeting-summarization

Installs the uv package manager using a PowerShell command. This command bypasses execution policy restrictions for installation.

```powershell
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
```

--------------------------------

### Preview Documentation Locally

Source: https://docs.liquid.ai/lfm/help/contributing

Use the mintlify CLI to preview your documentation changes locally before submitting them.

```bash
mintlify dev
```

--------------------------------

### List Installed Models

Source: https://docs.liquid.ai/deployment/on-device/ollama

View all models currently installed with Ollama.

```bash
ollama list
```

--------------------------------

### Basic Usage

Source: https://docs.liquid.ai/deployment/on-device/android/utilities

Demonstrates the fundamental steps to initialize the LeapModelDownloader and download a model.

```APIDOC
## Basic Usage

This section shows how to initialize the `LeapModelDownloader` and start downloading a model.

### Initialization

```kotlin
import ai.liquid.leap.model_downloader.LeapModelDownloader
import ai.liquid.leap.model_downloader.LeapModelDownloaderNotificationConfig

// Initialize (in onCreate or similar)
val modelDownloader = LeapModelDownloader(
    context,
    notificationConfig = LeapModelDownloaderNotificationConfig.build {
        notificationTitleDownloading = "Downloading AI model..."
        notificationTitleDownloaded = "Model ready!"
        notificationContentDownloading = "Please wait..."
    }
)
```

### Download and Load Model

```kotlin
import android.util.Log
import androidx.lifecycle.lifecycleScope
import kotlinx.coroutines.launch

lifecycleScope.launch {
    try {
        val modelRunner = modelDownloader.loadModel(
            modelSlug = "LFM2-1.2B",
            quantizationSlug = "Q5_K_M",
            progress = { progressData ->
                Log.d(TAG, "Progress: ${progressData.progress * 100}%")
            }
        )
        // Model is ready to use
    } catch (e: Exception) {
        Log.e(TAG, "Failed to load model: ${e.message}")
    }
}
```
```

--------------------------------

### Install uv

Source: https://docs.liquid.ai/examples/customize-models/home-assistant

Install the uv package manager for managing Python dependencies.

```bash
curl -LsSf https://astral.sh/uv/install.sh | sh
```

```powershell
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
```

--------------------------------

### Start llama-server

Source: https://docs.liquid.ai/deployment/on-device/llama-cpp

Launch the OpenAI-compatible server using either a remote Hugging Face ID or a local file path.

```bash
llama-server -hf LiquidAI/LFM2.5-1.2B-Instruct-GGUF -c 4096 --port 8080
```

```bash
llama-server -m lfm2.5-1.2b-instruct-q4_k_m.gguf -c 4096 --port 8080
```

--------------------------------

### Verify uv installation

Source: https://docs.liquid.ai/examples/laptop-examples/flight-search-assistant

Check that the uv package manager is installed on your system.

```bash
uv --version
```

--------------------------------

### Start Vision Model Server

Source: https://docs.liquid.ai/docs/inference/llama-cpp

Launch a local server to host the vision model for API access.

```bash
llama-server \
  -m LFM2-VL-1.6B-Q8_0.gguf \
  --mmproj mmproj-LFM2-VL-1.6B-Q8_0.gguf \
  -c 4096 \
  --port 8080 \
  -ngl 99
```

--------------------------------

### Install Transformers.js

Source: https://docs.liquid.ai/deployment/on-device/onnx

Install the required package for running ONNX models in the browser.

```bash
npm install @huggingface/transformers
```

--------------------------------

### Verify npm Installation

Source: https://docs.liquid.ai/examples/web/audio-webgpu-demo

Check the installed version of npm to ensure the environment is ready.

```bash
npm --version
```

--------------------------------

### Deploy and Register Model

Source: https://docs.liquid.ai/examples/customize-models/home-assistant

Start the local server and register custom models in the server configuration.

```bash
uv run uvicorn app.server:app --port 5173 --reload
```

```bash
open http://localhost:5173
```

--------------------------------

### Run liquid-audio Gradio Demo

Source: https://docs.liquid.ai/lfm/models/lfm2-audio-1.5b

Start the Gradio web server for the liquid-audio demo. The server will be accessible at http://localhost:7860/.

```bash
liquid-audio-demo
# Starts webserver on http://localhost:7860/
```

--------------------------------

### Example Analysis Output

Source: https://docs.liquid.ai/examples/android/vision-language-model-example

This is an example of the AI-generated analysis output for an image of a sunset over a beach.

```text
"The image shows a beautiful sunset scene at a beach. The sky displays
vibrant orange and pink hues as the sun sets on the horizon. The ocean
water reflects the warm colors of the sky. In the foreground, there are
silhouettes of people walking along the shoreline. The overall mood is
peaceful and serene."
```

--------------------------------

### Clone the SloganApp Repository

Source: https://docs.liquid.ai/examples/android/slogan-generator

Use Git to clone the example repository and navigate to the SloganApp directory.

```bash
git clone https://github.com/Liquid4All/LeapSDK-Examples.git
cd LeapSDK-Examples/Android/SloganApp
```

--------------------------------

### Validate a model directory example

Source: https://docs.liquid.ai/deployment/tools/model-bundling/bundle-creation

Example usage and output for validating a model directory.

```sh
# Validate a model directory
leap-bundle validate ./my-model-directory

# Example output on success
✓ Directory validation passed
✓ Directory is ready for bundle creation

# Example output on failure
✗ Validation failed: No config.json found in directory.
```

--------------------------------

### Install sglang

Source: https://docs.liquid.ai/lfm/models/lfm2-2.6b-transcript

Install sglang using pip. Ensure you have a version greater than or equal to 0.5.10.

```bash
uv pip install "sglang>=0.5.10"
```

--------------------------------

### Verify llama.cpp installation

Source: https://docs.liquid.ai/examples/customize-models/home-assistant

Confirm that the llama-server binary is correctly installed and accessible on your system PATH.

```bash
llama-server --version
```

--------------------------------

### Build for Production

Source: https://docs.liquid.ai/examples/web/audio-webgpu-demo

Create an optimized production build in the dist directory.

```bash
npm run build
```

--------------------------------

### Install MLX language model package

Source: https://docs.liquid.ai/deployment/on-device/mlx

Install the necessary library to interface with MLX models.

```bash
pip install mlx-lm
```

--------------------------------

### MDX Page Structure Example

Source: https://docs.liquid.ai/lfm/help/contributing

An example of the frontmatter and basic structure for an MDX documentation page.

```mdx
---
title: "Page Title"
description: "Brief description for SEO and previews (1-2 sentences)"
---

Introduction paragraph explaining what this page covers.

## Main Section

Content organized under clear headings...

### Subsection

More detailed content...

## Next Steps

<CardGroup cols={2}>
  <Card title="Related Topic" href="/path/to/page">
    Brief description
  </Card>
</CardGroup>
```

--------------------------------

### Start llama-server with Hugging Face model

Source: https://docs.liquid.ai/docs/inference/llama-cpp

Launches the llama.cpp server, automatically downloading the specified model from Hugging Face. Ensure the model ID is correct.

```bash
llama-server -hf LiquidAI/LFM2.5-1.2B-Instruct-GGUF -c 4096 --port 8080

```

--------------------------------

### Start llama-server with local GGUF model

Source: https://docs.liquid.ai/docs/inference/llama-cpp

Launches the llama.cpp server using a pre-downloaded GGUF model file. Specify the correct path to your model.

```bash
llama-server -m lfm2.5-1.2b-instruct-q4_k_m.gguf -c 4096 --port 8080

```

--------------------------------

### Example Function Definition

Source: https://docs.liquid.ai/deployment/on-device/ios/function-calling

A comprehensive example demonstrating how to define a LeapFunction with various parameter types.

```APIDOC
## Example Function Definition

This example shows how to define a function to query weather information.

### Code Example
```swift
LeapFunction(
  name: "get_weather",
  description: "Query the weather of cities",
  parameters: [
    LeapFunctionParameter(
      name: "cities",
      type: LeapFunctionParameterType.array(
        ArrayType(itemType: .string(StringType()))
      ),
      description: "Names of the cities to query weather for"
    ),
    LeapFunctionParameter(
      name: "unit",
      type: LeapFunctionParameterType.string(
        StringType(enumValues: ["celsius", "fahrenheit"])),
      description: "Temperature unit (celsius or fahrenheit)"
    ),
  ]
)
```
```

--------------------------------

### Example of defining a function

Source: https://docs.liquid.ai/deployment/on-device/ios/function-calling

A comprehensive example demonstrating how to construct a LeapFunction with array and string parameters.

```swift
LeapFunction(
  name: "get_weather",
  description: "Query the weather of cities",
  parameters: [
    LeapFunctionParameter(
      name: "cities",
      type: LeapFunctionParameterType.array(
        ArrayType(itemType: .string(StringType()))
      ),
      description: "Names of the cities to query weather for"
    ),
    LeapFunctionParameter(
      name: "unit",
      type: LeapFunctionParameterType.string(
        StringType(enumValues: ["celsius", "fahrenheit"])),
      description: "Temperature unit (celsius or fahrenheit)"
    ),
  ]
)
```