Features:
• Local models: import GGUF from Files, download from Hugging Face, or via a direct link.
• Model catalog: search, “GGUF only” filter, curated recommendations.
• OpenRouter (optional): browse/search models, including vision-capable models. Requires an API key.
• Attachments: photos, videos, and PDFs — OCR/text extraction is added to your prompt context.
• Voice input: dictate prompts with speech-to-text.
• Generation controls: system prompt, context size, temperature, top‑p, max tokens.
• Chat history: multiple threads with rename & delete.