- Add real-time token usage and cost tracking from Claude Code responses - New UsageStats type with tokens, cost, and context window - Context window tooltip shows detailed token breakdown - Cost displayed in main panel header and session list - Improve input area UX with $ prefix for terminal mode - Add Cmd+L shortcut to insert markdown checkbox in Scratchpad - Add terminalOnly flag for slash commands (/jump is terminal-only) - Disable LLM Settings panel behind feature flag (documented in BACKBURNER.md) - Fix AgentSessionsBrowser to use layer stack for Escape handling - Update README with cost tracking feature and keyboard shortcut - Update CLAUDE.md with process:usage event and usageStats session field
1.6 KiB
Backburner Features
This document tracks dormant features that are implemented but disabled via feature flags. These features may be re-enabled in future releases.
LLM Settings Panel
Status: Disabled
Feature Flag: FEATURE_FLAGS.LLM_SETTINGS in src/renderer/components/SettingsModal.tsx
Disabled Date: 2024-11-26
Description
The LLM Settings panel provides configuration options for connecting to various LLM providers directly from Maestro. This feature was designed to enable a built-in AI assistant for the scratchpad or other future AI-powered features within the application.
Supported Providers
- OpenRouter - API proxy supporting multiple models
- Anthropic - Direct Claude API access
- Ollama - Local LLM inference
Configuration Options
- LLM Provider selection
- Model slug/identifier
- API key (stored locally)
- Connection test functionality
Files Involved
src/renderer/components/SettingsModal.tsx- Main settings UI with LLM tab- Settings stored in electron-store:
llmProvider,modelSlug,apiKey
Re-enabling
To re-enable this feature:
- Open
src/renderer/components/SettingsModal.tsx - Find the
FEATURE_FLAGSconstant at the top of the file - Set
LLM_SETTINGS: true
const FEATURE_FLAGS = {
LLM_SETTINGS: true, // LLM provider configuration (OpenRouter, Anthropic, Ollama)
};
Reason for Disabling
Currently not in use as Maestro focuses on managing external AI coding agents (Claude Code, etc.) rather than providing built-in LLM functionality. May be re-enabled when there's a use case for direct LLM integration within Maestro.