### Install Dependencies and Download Model Weights (Bash)

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Install necessary Python packages, download the Qwen3-ASR model weights, and generate the tokenizer. This is a prerequisite for building the project from source.

```bash
pip install huggingface_hub transformers

huggingface-cli download Qwen/Qwen3-ASR-0.6B --local-dir Qwen3-ASR-0.6B

python -c "
from transformers import AutoTokenizer
tok = AutoTokenizer.from_pretrained('Qwen3-ASR-0.6B', trust_remote_code=True)
tok.backend_tokenizer.save('Qwen3-ASR-0.6B/tokenizer.json')
"
```

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

### Quick Install Qwen3 ASR Skill

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/skills/install.md

Use this script for a quick and recommended installation of the Qwen3 ASR skill. It clones the repository, copies necessary files, and downloads platform-specific releases and models.

```bash
SKILL_DIR="${HOME}/.openclaw/skills/audio_asr"
mkdir -p "$SKILL_DIR"

# Clone the repo
git clone --depth 1 https://github.com/second-state/qwen3_asr_rs.git /tmp/qwen3-asr-repo
cp -r /tmp/qwen3-asr-repo/skills/* "$SKILL_DIR"
rm -rf /tmp/qwen3-asr-repo

# Download platform-specific release and model
"${SKILL_DIR}/bootstrap.sh"
```

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

### Verify Qwen3 ASR Skill Installation

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/skills/install.md

After installation, use this command to verify that the Qwen3 ASR skill is working correctly by transcribing an audio file.

```bash
~/.openclaw/skills/audio_asr/scripts/asr \
  ~/.openclaw/skills/audio_asr/scripts/models/Qwen3-ASR-0.6B \
  /path/to/audio.wav
```

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

### Manual Installation: Download Model

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/skills/install.md

Manually download the Qwen3-ASR-0.6B model using `huggingface-cli` and save it to the specified local directory.

```bash
huggingface-cli download Qwen/Qwen3-ASR-0.6B \
  --local-dir ~/.openclaw/skills/audio_asr/scripts/models/Qwen3-ASR-0.6B
```

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

### Install Qwen3 ASR CLI

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Run this script to automatically detect your platform, download the correct release binary, model weights, and a sample audio file. Follow the prompts for model size and GPU usage.

```bash
curl -sSf https://raw.githubusercontent.com/second-state/qwen3_asr_rs/main/install.sh | bash
```

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

### Start Qwen3 ASR API Server

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Run the `asr-server` binary, specifying the path to the model directory. The server defaults to host 0.0.0.0 and port 8080.

```bash
asr-server --model-dir ./Qwen3-ASR-0.6B
```

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

### Perform First Transcription

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

After installation, navigate to the project directory and use the `asr` command with the model and a sample audio file to generate text.

```bash
cd qwen3_asr_rs
./asr ./Qwen3-ASR-0.6B sample.wav
```

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

### CLI Transcription Usage

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Examples of how to use the `asr` CLI tool for basic and language-specific audio transcription.

```APIDOC
## CLI Transcription Usage

### Basic transcription (auto-detect language)
```bash
asr ./Qwen3-ASR-0.6B input.wav
```

### Force language
```bash
asr ./Qwen3-ASR-0.6B input.wav chinese
asr ./Qwen3-ASR-0.6B input.wav english
```

### Enable debug logging
```bash
RUST_LOG=debug asr ./Qwen3-ASR-0.6B input.wav
```

### Output Format
```
Language: Chinese
Text: 你好世界
```
```

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

### Build for Linux with LibTorch (Bash)

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Install dependencies and build the project for Linux using the libtorch backend. Ensure libtorch has been downloaded and environment variables are set.

```bash
cargo build --release
```

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

### Manual Installation: Copy Scripts

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/skills/install.md

If automatic download fails, manually extract the downloaded zip file for your platform and copy its contents to the scripts directory. Ensure the 'asr' executable has execute permissions.

```bash
SCRIPTS=~/.openclaw/skills/audio_asr/scripts
mkdir -p "$SCRIPTS"
unzip asr-<platform>.zip
cp -r asr-<platform>/* "$SCRIPTS/"
chmod +x "$SCRIPTS/asr"
```

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

### Manual Installation: Generate Tokenizer

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/skills/install.md

Generate the `tokenizer.json` file required for the Qwen3 ASR skill. This involves using `transformers.AutoTokenizer` to load the model and save its backend tokenizer.

```python
from transformers import AutoTokenizer
import os
path = os.path.expanduser('~/.openclaw/skills/audio_asr/scripts/models/Qwen3-ASR-0.6B')
tok = AutoTokenizer.from_pretrained(path, trust_remote_code=True)
tok.backend_tokenizer.save(f'{path}/tokenizer.json')
print(f'Saved {path}/tokenizer.json')
```

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

### API Server - POST /v1/audio/transcriptions

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Documentation for the OpenAI-compatible transcription endpoint, detailing parameters, request examples, and response formats.

```APIDOC
## POST /v1/audio/transcriptions

OpenAI-compatible transcription endpoint. Accepts multipart form data.

### Parameters

#### Request Body

- **file** (binary) - Required - Audio file (any format supported by FFmpeg)
- **language** (string) - Optional - Language hint (e.g., `english`, `chinese`)
- **response_format** (string) - Optional - `json` (default), `text`, or `verbose_json`
- **model** (string) - Optional - Accepted for compatibility, ignored
- **temperature** (float) - Optional - Accepted for compatibility, ignored
- **prompt** (string) - Optional - Accepted for compatibility, ignored

### Request Example

```bash
# JSON response (default)
curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@recording.wav

# {"text":"Thank you for your contribution..."}

# Plain text response
curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@recording.wav \
  -F response_format=text

# Thank you for your contribution...

# Verbose JSON with language and duration
curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@recording.wav \
  -F response_format=verbose_json

# {"task":"transcribe","language":"English","duration":7.999,"text":"Thank you..."}

# Force language
curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@recording.wav \
  -F language=chinese
```

### Response

#### Success Response (200)

- **text** (string) - The transcribed text.
- **task** (string) - The type of task performed (e.g., `transcribe`).
- **language** (string) - The detected or specified language.
- **duration** (float) - The duration of the audio in seconds.

#### Response Example

```json
{
  "task": "transcribe",
  "language": "English",
  "duration": 7.999,
  "text": "Thank you for your contribution to the most recent issue of Computer."
}
```
```

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

### POST /v1/audio/transcriptions JSON Response

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Example of a JSON response from the transcription endpoint when `response_format` is set to `json` (default).

```bash
curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@recording.wav

# {"text":"Thank you for your contribution..."}
```

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

### POST /v1/audio/transcriptions Plain Text Response

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Example of a plain text response from the transcription endpoint when `response_format` is set to `text`.

```bash
curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@recording.wav \
  -F response_format=text

# Thank you for your contribution...
```

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

### POST /v1/audio/transcriptions Force Language

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Example of forcing a specific language for transcription using the `language` parameter in the POST request.

```bash
curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@recording.wav \
  -F language=chinese
```

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

### Enable Debug Logging for ASR

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/CLAUDE.md

Run the ASR tool with debug logging enabled to get detailed information about the transcription process. This is helpful for troubleshooting.

```bash
RUST_LOG=debug ./target/release/asr ./Qwen3-ASR-0.6B test_audio/sample1.wav
```

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

### POST /v1/audio/transcriptions Verbose JSON Response

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Example of a verbose JSON response from the transcription endpoint when `response_format` is set to `verbose_json`. Includes task, language, duration, and text.

```bash
curl -X POST http://localhost:8080/v1/audio/transcriptions \
  -F file=@recording.wav \
  -F response_format=verbose_json

# {"task":"transcribe","language":"English","duration":7.999,"text":"Thank you..."}
```

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

### Build Commands for Qwen3-ASR-RS

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/CLAUDE.md

Commands to build the Qwen3-ASR-RS project. Use the MLX backend on macOS for Apple Silicon, and the libtorch backend on Linux. Features like 'mlx' and 'build-ffmpeg' are mutually exclusive and affect the build process.

```bash
# macOS (MLX backend — recommended for Apple Silicon)
git submodule update --init --recursive
cargo build --release --no-default-features --features mlx,build-ffmpeg
```

```bash
# Linux (libtorch backend)
export LIBTORCH=$(pwd)/libtorch
export LIBTORCH_BYPASS_VERSION_CHECK=1
cargo build --release --features build-ffmpeg
```

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

### Troubleshooting: Check Platform

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/skills/install.md

Determine your operating system and architecture using `uname` to ensure compatibility with the available Qwen3 ASR skill releases.

```bash
echo "OS: $(uname -s), Arch: $(uname -m)"
```

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

### List Available Models (Bash)

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Use this command to list the models available through the API. Ensure the API server is running at http://localhost:8080.

```bash
curl http://localhost:8080/v1/models
# {"object":"list","data":[{"id":"qwen3-asr","object":"model","owned_by":"qwen"}]}
```

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

### Qwen3 ASR API Server Options

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Configuration options for the `asr-server`, including model directory, host, port, default language, and verbosity.

```bash
--model-dir <PATH>      Path to the Qwen3-ASR model directory (required)
--host <ADDR>           Host address to bind to (default: 0.0.0.0)
--port <PORT>           Port to listen on (default: 8080)
--language <LANG>       Default language for transcription (e.g., chinese, english)
-v, -vv                 Verbose output (debug, trace)
```

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

### Basic Transcription with Qwen3 ASR CLI

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Use the `asr` command to transcribe an audio file. The tool automatically detects the language.

```bash
asr ./Qwen3-ASR-0.6B input.wav
```

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

### Build for macOS with MLX (Bash)

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Build the project for macOS using the MLX backend. Ensure Git submodules are updated before building.

```bash
git submodule update --init --recursive
cargo build --release --no-default-features --features mlx
```

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

### Troubleshooting: Check Network Connectivity

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/skills/install.md

Use `curl` to check if you can reach the GitHub releases page, which can help diagnose download failures.

```bash
curl -I "https://github.com/second-state/qwen3_asr_rs/releases/latest"
```

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

### Download LibTorch for Linux (Bash)

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Download and extract the appropriate libtorch version for your Linux environment (CPU or CUDA, x86_64 or ARM64). This is required for building with the libtorch backend.

```bash
# Linux x86_64 (CPU)
curl -LO https://github.com/second-state/libtorch-releases/releases/download/v2.7.1/libtorch-cxx11-abi-x86_64-2.7.1.tar.gz
tar xzf libtorch-cxx11-abi-x86_64-2.7.1.tar.gz

# Linux x86_64 (CUDA 12.6)
curl -LO https://github.com/second-state/libtorch-releases/releases/download/v2.7.1/libtorch-cxx11-abi-x86_64-cuda12.6-2.7.1.tar.gz
tar xzf libtorch-cxx11-abi-x86_64-cuda12.6-2.7.1.tar.gz

# Linux ARM64 (CPU)
curl -LO https://github.com/second-state/libtorch-releases/releases/download/v2.7.1/libtorch-cxx11-abi-aarch64-2.7.1.tar.gz
tar xzf libtorch-cxx11-abi-aarch64-2.7.1.tar.gz

# Linux ARM64 (CUDA 12.6 / Jetson)
curl -LO https://github.com/second-state/libtorch-releases/releases/download/v2.7.1/libtorch-cxx11-abi-aarch64-cuda12.6-2.7.1.tar.gz
tar xzf libtorch-cxx11-abi-aarch64-cuda12.6-2.7.1.tar.gz
```

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

### Enable Debug Logging for Qwen3 ASR CLI

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

To enable debug logging, set the `RUST_LOG` environment variable to `debug` before running the `asr` command.

```bash
RUST_LOG=debug asr ./Qwen3-ASR-0.6B input.wav
```

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

### Basic Transcription with ASR

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/CLAUDE.md

Execute a basic transcription of an audio file using the ASR tool. Specify the model path and the audio file path.

```bash
./target/release/asr ./Qwen3-ASR-0.6B test_audio/sample1.wav
```

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

### Transcribe Audio File

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/skills/SKILL.md

Use the `asr` binary to transcribe an audio file. Specify the model path and the audio file path as arguments. Any FFmpeg-supported audio format is accepted.

```shell
{baseDir}/scripts/asr \
  {baseDir}/scripts/models/Qwen3-ASR-0.6B \
  <audio_file>
```

```shell
{baseDir}/scripts/asr \
  {baseDir}/scripts/models/Qwen3-ASR-0.6B \
  recording.wav
```

```shell
{baseDir}/scripts/asr \
  {baseDir}/scripts/models/Qwen3-ASR-0.6B \
  /path/to/audio.wav
```

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

### Force Language Transcription with Qwen3 ASR CLI

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Specify the language for transcription by providing the language code (e.g., `chinese`, `english`) as an argument after the audio file.

```bash
asr ./Qwen3-ASR-0.6B input.wav chinese
asr ./Qwen3-ASR-0.6B input.wav english
```

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

### List Available Models

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Retrieves a list of all available ASR models supported by the service.

```APIDOC
## GET /v1/models

### Description
Lists available models.

### Method
GET

### Endpoint
/v1/models

### Response
#### Success Response (200)
- **object** (string) - Type of the response object, typically "list".
- **data** (array) - An array of model objects.
  - **id** (string) - The unique identifier for the model.
  - **object** (string) - The type of the object, typically "model".
  - **owned_by** (string) - The owner of the model.

### Response Example
```json
{
  "object": "list",
  "data": [
    {
      "id": "qwen3-asr",
      "object": "model",
      "owned_by": "qwen"
    }
  ]
}
```
```

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

### Health Check Endpoint (Bash)

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Check the health status of the API server by sending a request to the /health endpoint. A response of {"status":"ok"} indicates the server is running.

```bash
curl http://localhost:8080/health
# {"status":"ok"}
```

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

### Force Language Transcription

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/CLAUDE.md

Perform transcription while explicitly setting the expected language. This is useful for multilingual models or when language detection might be uncertain.

```bash
./target/release/asr ./Qwen3-ASR-0.6B test_audio/sample3.wav chinese
```

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

### Set Environment Variables for LibTorch (Bash)

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Set the LIBTORCH environment variable to the path where libtorch was extracted and bypass version checks. These variables are necessary for the build process on Linux.

```bash
export LIBTORCH=$(pwd)/libtorch
export LIBTORCH_BYPASS_VERSION_CHECK=1
```

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

### Health Check

Source: https://github.com/second-state/qwen3_asr_rs/blob/main/README.md

Checks the health status of the ASR service.

```APIDOC
## GET /health

### Description
Health check endpoint.

### Method
GET

### Endpoint
/health

### Response
#### Success Response (200)
- **status** (string) - The health status of the service, typically "ok".

### Response Example
```json
{
  "status": "ok"
}
```
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.