Description: The describe_image function in the media understanding engine is not fully implemented - it returns placeholder text instead of making actual API calls to vision-capable LLM providers.

Location: crates/librefang-runtime/src/media_understanding.rs:29-54

Expected Behavior: The function should call vision-capable LLM APIs (Anthropic, OpenAI, or Gemini) to generate actual image descriptions.

Current Implementation:

// For now, return a structured result indicating the provider.
// Actual API call would go here using reqwest.
Ok(MediaUnderstanding {
    media_type: MediaType::Image,
    description: format!(
        "[Image description would be generated by {} provider]",
        provider
    ),
    provider: provider.to_string(),
    model: default_vision_model(provider).to_string(),
})

Root Cause: The function detects the provider but does not make actual API calls - the implementation is incomplete.

Note: The transcribe_audio function in the same file IS fully implemented and makes actual API calls to Groq/OpenAI.

Suggested Fix: Implement the actual vision API call logic for image description, similar to how transcribe_audio is implemented.