hermes - 💡(How to fix) Fix [Feature]: a Hermes-Canvas (Text Editing & Hybrid Completion Environment) on top of a well-built GUI for Hermes Agent

Problem or Use Case

Request Description: Implement Hermes-Canvas, an interactive, text-first collaborative editing workspace integrated directly within Hermes Agent. The core mechanic allows users to highlight specific sections of a prompt or assistant output, apply natural language edit instructions, and re-run only that isolated text block.

To create a premier open-source workspace superior to commercial alternatives, Hermes-Canvas will support three distinct editing behaviors driven by a dual-backend infrastructure:

• Targeted Inline Edits: Highlight a paragraph or sentence and issue a natural language instruction to modify only that section. The system passes the entire canvas as background context to preserve document consistency, but only alters and streams back the highlighted region. • Triggered Text Completion: Inside the editing space, typing a partial sentence and pressing double spaces will instantly auto-complete the text inline, utilizing the surrounding canvas document as its reference context. • Batch Editing (Industry-First and a clear product differentiator): Apply a single edit instruction (e.g., "Change tone to formal" or "Update variable name") across multiple disconnected, highlighted text selections simultaneously. {Suggestions for editing can also be generated via agents and confirmed via user.}

Infrastructure & UI Requirements: • Split API Architecture: Separate standard chat tasks from text completion workflows by allowing users to configure different API endpoints/models for each purpose (e.g., a fast, cheap model for double-space inline completions and a frontier model for canvas editing). • Version Control & Replayability: A native version-history timeline or back-button interface to instantly undo/redo edits. • Quick-Action Preset Shortcuts: Floating quick-access buttons for common document tasks (Adjust Length, Polish Grammar, Shift Reading Level). • Smart Invocation: Auto-trigger the Canvas UI layout when Hermes Agent detects long-form text generation or explicit user formatting commands. User Intentions • Granular Document Control: Shifting away from full-context chat regenerations toward a highly targeted, interactive document workbench. • Architecture Flexibility: Optimizing token usage and performance by routing chat interactions and background autocomplete requests to completely different, specialized backend APIs. • Unprecedented Parallelism: Executing multi-region parallel edits at scale across massive texts—a feature entirely absent from hosted alternatives like ChatGPT and Gemini.

Key Benefits • Substantial Token & Cost Savings: Minimizing context payloads by regenerating only targeted text blocks instead of re-submitting and re-writing entire long documents. • Frictionless Writing Flow: Merging visual text selection with predictive double-space completion eliminates the constant copy-pasting between the chat window and external text editors. • Provider-Agnostic Sovereignty: Provides the sophisticated canvas workflows of closed systems (Claude/Gemini) while remaining entirely local, provider-neutral, and embedded within the private Hermes Agent ecosystem. • Auditability & Automation: Every canvas edit is represented in the chat transcript as a discrete tool-action card, making document generation histories reproducible, undoable, and capable of being automated via Hermes skills.

Proposed Solution

First build a well-balanced and rich GUI experience and assets like Hermes-Canvas by studying Gemini-Canvas. Additional features that other canvases do not have are given in the feature requested. like batch editing and text completion in Canvas. For this, a GUI web app experience natively in the Hermes agent JS is necessary.

Alternatives Considered

No response

Feature Type

New tool

Scope

Large (new module or significant refactor)

Contribution

• I'd like to implement this myself and submit a PR

Debug Report (optional)