Integrate the speech recognition model

Call the Token Plan speech recognition API to transcribe audio into text.

Audio file or URL: $ARGUMENTS

## Steps

1. Confirm that the input is a local audio path or a publicly accessible URL. The audio must be no longer than 5 minutes. For format and size limits, see the audio specifications in the QwenCloud speech recognition documentation.
2. Install dependencies: pip install requests
3. Use the Bash tool to run the following Python script. The script reads TOKEN_PLAN_API_KEY first, then ANTHROPIC_AUTH_TOKEN. Do not display the key or write it to files or logs.

```python
import base64
import json
import os
import sys
from pathlib import Path

import requests

MIME_TYPES = {
    "aac": "audio/aac",
    "amr": "audio/amr",
    "flac": "audio/flac",
    "m4a": "audio/mp4",
    "mp3": "audio/mpeg",
    "ogg": "audio/ogg",
    "opus": "audio/opus",
    "wav": "audio/wav",
    "webm": "audio/webm",
}

source = sys.argv[1]
key = os.environ.get("TOKEN_PLAN_API_KEY") or os.environ.get("ANTHROPIC_AUTH_TOKEN")
if not key:
    raise SystemExit("Token Plan API Key not found.")

if source.startswith(("http://", "https://")):
    audio_data = source
    audio_format = Path(source.split("?")[0]).suffix.lstrip(".").lower() or "mp3"
else:
    path = Path(source)
    audio_format = path.suffix.lstrip(".").lower()
    mime_type = MIME_TYPES.get(audio_format, f"audio/{audio_format}")
    audio_data = f"data:{mime_type};base64,{base64.b64encode(path.read_bytes()).decode()}"

payload = {
    "model": "qwen-audio-3.0-asr-flash",
    "input": {
        "messages": [
            {
                "role": "user",
                "content": [
                    {"type": "input_audio", "input_audio": {"data": audio_data}}
                ],
            }
        ]
    },
    "parameters": {"format": audio_format},
}

response = requests.post(
    "https://token-plan.maas.qwencloudapi.com/api/v1/services/aigc/multimodal-generation/generation",
    headers={"Authorization": f"Bearer {key}", "Content-Type": "application/json"},
    json=payload,
    timeout=300,
)
if not response.ok:
    raise SystemExit(f"{response.status_code}: {response.text}")
result = response.json()
print(result.get("text") or json.dumps(result, ensure_ascii=False, indent=2))
```

4. Output the transcription. If detailed results are needed, show the `sentence` and `words` fields in the response JSON.