#1255: Please add a lightweight, text-based Canvas editing inside the WebUI

Add a lightweight, text-based Canvas editing inside the WebUI that lets users select a portion of a prompt / output (a localized region of text or structured content), apply an edit instruction in natural language, and re-run only that region through the chosen model. This is a text-first "Canvas" (not an image editor) for fine-grained, interactive editing of assistant outputs and local content.

Why this benefits hermes-webui

Token & cost savings: Resending and regenerating only a selected text region avoids re-submitting entire contexts, reducing tokens and cost. Faster, clearer editing flow: users can iteratively refine specific parts of an answer (or generated document) with a visual-edit surface instead of crafting full prompts or pasting large context blocks. Better provenance and replayability: edits become discrete tool-like actions saved in session history (prompt region + edit instruction + model), making edits reproducible and auditable. Mobile- and touch-friendly: localized text edits are easier on phones than retyping whole prompts. Product differentiator

Self-hosted, provider-agnostic Canvas: provides the convenience of hosted Canvas features (Claude/Gemini/ChatGPT) while staying provider-neutral and running through whatever model(s) the user configures. Session-integrated artifacts: edits are first-class in the chat transcript (tool/action cards) enabling regenerations, undo, and automation via Hermes skills. Token-aware UX: the UI can prioritize minimal-region edits and expose estimated savings, which will be attractive for users running on paid APIs or limited local models.