Feature request

### Problem Description Currently, the repository relies on Ollama for processing text workflows. While Ollama is ideal for entirely local setups, it requires substantial hardware resources (CPU/GPU/VRAM) to run effectively. Users operating on low-spec hardware or headless environments need an option to offload processing to external cloud API providers. ### Proposed Solution Add native support for external cloud LLM APIs alongside the existing Ollama implementation. Ideally, the application should allow switching to an API mode where users can input custom endpoint credentials. ### Feature Requirements * **Custom API Base URL:** Allow setting a custom endpoint (e.g., to connect to `://openai.com`, `://anthropic.com`, or OpenAI-compatible aggregators like OpenRouter). * **API Key Authentication:** Provide a secure configuration setting or environment variable field to input bearer tokens/API keys. * **Model Selection String:** Provide a text input field to manually specify the target model string (e.g., `gpt-4o`, `claude-3-5-sonnet`, etc.). ### Alternatives Considered Forcing all installations to run large multi-gigabyte models locally via Ollama, which restricts usage on machines with lower performance thresholds or limited storage capacity.