v0.0.6

• Removing a source now removes its indexed memory without deleting the original files.
• The running API now picks up chat, embedding, image, and provider changes made from the CLI.
• Updating models now unloads previous OpenMind models(from the model provider) before loading their replacements.
• Ask responses now format answers, links, and local sources as Markdown.
• Sessions have been added to the api: Chat now uses faster provider-backed sessions, keeps API sources separate, and lets CLI and API users enable model reasoning when needed while keeping it off by default.