[Bug]: RamaLama provdier detected but not installed #790

Bug description

TUI detects RamaLama provider but it is not installed on my system:

% which ramalama
ramalama not found

However I have llama-swap running on localhost:8080 and when I turn it off TUI stops detecting RamaLama.

Looks like generic OpenAI API running on localhost is mistakenly detected as RamaLama provider.

Expected behavior

Generic OpenAI API running on localhost:8080 (address not exclusive to RamaLama) shouldn't be detected as RamaLama provider.

Actual behavior

RamaLama is not installed (no such binary in PATH) but other service running on localhost is detected as RamaLama.

Steps to reproduce

• make sure no RamaLama installed
• run other OpenAI API on localhost:8080
• check detected providers in llmfit TUI

Diagnostic report (llmfit doctor)

## Detected by llmfit

SystemSpecs {
    total_ram_gb: 32.0,
    available_ram_gb: 25.521865844726563,
    total_cpu_cores: 10,
    cpu_name: "Apple M1 Max",
    has_gpu: true,
    gpu_vram_gb: Some(
        32.0,
    ),
    total_gpu_vram_gb: Some(
        32.0,
    ),
    gpu_available_gb: Some(
        25.390625,
    ),
    gpu_name: Some(
        "Apple M1 Max",
    ),
    gpu_count: 1,
    unified_memory: true,
    backend: Metal,
    gpus: [
        GpuInfo {
            name: "Apple M1 Max",
            vram_gb: Some(
                32.0,
            ),
            backend: Metal,
            count: 1,
            unified_memory: true,
        },
    ],
    cluster_mode: false,
    cluster_node_count: 0,
}


## Provider app installs

LM Studio installed: false
Docker Desktop installed: false
ollama on PATH: true

llmfit version

llmfit 1.1.6

Operating system

macOS (Apple Silicon)

Affected component

Provider detection (Ollama, llama.cpp, MLX, Docker)

Mode used

TUI (default)

GPU details (if relevant)

Additional context

No response