What Happens When Both Groq and OpenAI API Keys Are Configured in claude-video

When both Groq and OpenAI API keys are configured in claude-video, the system automatically selects Groq as the transcription backend because it appears first in the priority list within the load_api_key() function.

The claude-video repository provides a video transcription skill that supports multiple Whisper backends for converting speech to text. When you configure both Groq and OpenAI API keys in your environment variables or .env file, the application must decide which provider to use for processing audio. Understanding this automatic selection logic helps you predict costs, latency, and availability based on which API key takes precedence.

How the Backend Selection Logic Works

According to the source code in skills/watch/scripts/whisper.py, the transcription backend is determined by a priority-based lookup in the load_api_key() function.

The Candidate List Priority

At lines 65-71, the function constructs an ordered tuple of candidate providers:

candidates = [
    ("GROQ_API_KEY", "groq"),
    ("OPENAI_API_KEY", "openai")
]

The function iterates through this list and returns the first key it finds in the environment or .env file. Because GROQ_API_KEY appears first in the tuple, Groq is preferred whenever both keys are present, making it the default transcription provider.

Integration with the Transcription Flow

The transcribe_video() function invokes load_api_key() at lines 24-28 to discover the available backend. If no explicit backend argument is passed to the script, this automatic detection determines whether Groq or OpenAI handles the audio processing, and the selected backend name is reported back to the caller.

Verifying the Active Backend

You can confirm which provider is active using the setup status command. When a Groq key is present, the JSON output from skills/watch/scripts/setup.py reports "whisper_backend": "groq", as verified in the test suite at tests/test_setup.py lines 73-80.

Run the following to check your configuration:

python3 skills/watch/scripts/setup.py --json

Example output when both keys are configured:

{
  "whisper_backend": "groq",
  "status": "ready"
}

Overriding the Default Selection

If you need to use OpenAI despite having both keys configured, the script provides a command-line override. As implemented at lines 76-78 of skills/watch/scripts/whisper.py, passing the --backend flag bypasses the automatic detection.

Force OpenAI when both keys exist:

watch <video-url-or-path> --backend openai

This forces the transcription to use OpenAI's Whisper endpoint, ignoring the Groq key that would otherwise be selected by default.

Summary

• When both Groq and OpenAI API keys are configured in claude-video, Groq is automatically selected as the transcription backend due to its position as the first item in the candidate list.
• The selection logic resides in skills/watch/scripts/whisper.py within the load_api_key() function, which checks for GROQ_API_KEY before OPENAI_API_KEY.
• The active backend is reported in the setup status JSON as "whisper_backend": "groq" when the Groq key is present.
• Users can force OpenAI by passing the --backend openai command-line argument to override the automatic priority.

Frequently Asked Questions

Why does claude-video prefer Groq over OpenAI?

The codebase explicitly orders Groq first in the provider candidate list within the load_api_key() function at lines 65-71 of skills/watch/scripts/whisper.py. This design prioritizes Groq's typically faster inference and lower costs for Whisper transcription tasks, though users retain the flexibility to override this default with the --backend flag.

Can I use both backends simultaneously in the same command?

No, the transcription script processes each video using a single backend. The load_api_key() function returns exactly one API key and backend identifier. You must run separate commands with different --backend flags if you need to compare outputs from both providers.

How do I check which API key is actually being used?

Run the setup status command: python3 skills/watch/scripts/setup.py --json. The whisper_backend field indicates the active provider. Additionally, when running the watch command, the backend selection is reported in the output, allowing you to verify whether Groq or OpenAI is processing your request.

What happens if neither API key is configured?

If neither GROQ_API_KEY nor OPENAI_API_KEY is found in the environment or .env file, the load_api_key() function returns None for both the key and backend. The transcription will fail with an error indicating that no valid API credentials were detected, prompting you to configure at least one provider.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}