feat: add Astraflow provider support to voice input example

[ucloudnb666]

Summary

Adds Astraflow (by UCloud / 优刻得) as a named provider shortcut in examples/voice_input/funasr_input.py.

Astraflow is an OpenAI-compatible AI model aggregation platform supporting 200+ models. Because the voice input tool already uses the openai Python client with a configurable base_url, wiring in Astraflow requires only:

1. A PROVIDER_CONFIGS dict that maps provider names → (base_url, api_key_env).
2. A --provider CLI argument that overrides --server / --apikey when set.
3. Env-var validation with a helpful signup URL when the key is missing.

Usage:

Global endpoint (outside China):

export ASTRAFLOW_API_KEY=your_key_here
python funasr_input.py --provider astraflow --model sensevoice

China endpoint:

export ASTRAFLOW_CN_API_KEY=your_key_here
python funasr_input.py --provider astraflow-cn --model sensevoice

Type of change

• Bug fix
• Documentation
• Example or demo
• Runtime or deployment
• Benchmark or evaluation
• Model/training change

Validation

Changes are limited to examples/voice_input/funasr_input.py: adds PROVIDER_CONFIGS, --provider / --apikey CLI args, provider resolution logic, and updated usage examples in the docstring. No new files, no new dependencies, no changes to core FunASR model code.

• python -m compileall funasr examples tests
• Docs or links checked
• Runtime/deployment command tested

N/A (no screenshots or recordings applicable)

User impact

Users who want to use Astraflow (by UCloud / 优刻得) as their LLM backend — both global users (ASTRAFLOW_API_KEY) and China-region users (ASTRAFLOW_CN_API_KEY) — can now select it via a single --provider astraflow or --provider astraflow-cn flag without manually specifying --server and --apikey. This lowers the barrier for new users onboarding with an OpenAI-compatible aggregation platform supporting 200+ models.

Notes for reviewers

• This is a purely additive change to one example file; no core library code is touched.
• The new --provider argument short-circuits --server / --apikey when set, so existing CLI usage is fully backward-compatible.
• Provider endpoint details:

Field	Value
Global endpoint	https://api-us-ca.umodelverse.ai/v1
China endpoint	https://api.modelverse.cn/v1
Global signup	https://astraflow.ucloud-global.com
China signup	https://astraflow.ucloud.cn
Protocol	OpenAI-compatible (drop-in base_url replacement)

• Follow-up: other named providers could be added to PROVIDER_CONFIGS using the same pattern.

[gemini-code-assist]

Code Review

This pull request introduces a provider shortcut system for the FunASR voice input example, allowing users to switch between local and Astraflow backends via a new --provider flag. It also adds support for API keys through command-line arguments or environment variables. The reviewer suggested simplifying the API key assignment logic by removing an unnecessary getattr call and redundant default value handling.

[gemini-code-assist]

The use of getattr(args, "apikey", None) is unnecessary here because apikey is a defined argument in the parser and is guaranteed to exist on the args object. You can simplify this to a direct attribute access. Additionally, the default value logic here makes the else block at lines 88-89 redundant.

    api_key = args.apikey or "not-needed"

[LauraGPT]

LGTM — adds Astraflow as an alternative voice input provider. Clean addition to the example.

[LauraGPT]

Thanks for the contribution! However, we'd prefer not to hardcode specific third-party provider endpoints in the codebase.

The existing --server flag already supports any OpenAI-compatible API, including Astraflow:

python funasr_input.py --server https://api-us-ca.umodelverse.ai/v1

Also note there's an indentation bug in the PR (the try: block is mis-indented after the provider resolution code).

We'll close this for now, but feel free to add Astraflow as an example in the documentation or README instead. Thanks for supporting FunASR!