Doc Guardian (VS Code Extension)

Doc Guardian is a coding-side assistant powered by your ElevenLabs ConvAI agent.

1. Ask while coding in a dedicated chat panel.
2. Search docs for APIs/libraries from the command palette.
3. Live code monitoring for risky patterns and diagnostics.
4. Voice mode with dictation + spoken responses.
5. Background docs watch that alerts automatically when active-file issues appear.

Features

• Doc Guardian: Open Assistant
  • Opens the chat panel.
• Doc Guardian: Search Docs
  • Searches web docs (Firecrawl) and shows top references.
• Doc Guardian: Ask About Selection
  • Uses current code selection as context.
• Doc Guardian: Set Project Context
  • Auto-derives repo hints (name/stack/summary), lets you add docs URLs, then saves all for this workspace.
• Doc Guardian: New Session
  • Starts a fresh saved session while keeping project context.
• Doc Guardian: Switch Session
  • Re-open previous saved sessions for this workspace.
• Voice in panel:
  • Dictate button captures voice-to-text (when supported in runtime)
  • assistant replies are spoken automatically (ElevenLabs TTS + local fallback)
• Background docs watch:
  • continuously monitors active file diagnostics/findings
  • low-priority monitor for changed files across workspace
  • auto-searches docs relevant to file imports + active errors
  • shows alerts and pushes top docs without asking

ConvAI runtime flow:

• Extension opens a real text-only ConvAI session to your dashboard agent.
• Agent calls these client tools from the extension:
  • get_code_context
  • search_docs
  • get_monitor_findings_v5
• Tool results are returned to the agent, and agent replies are shown in panel + spoken.
• File-aware context is injected on asks:
  • active file + visible editors
  • files you mention in prompt (for example src/app.ts, @server/index.ts, package.json)
• Session memory is persisted:
  • project context per workspace
  • reference docs URLs per workspace context
  • turns history per session
  • docs lookups per session

Live monitor checks:

• any type usage
• @ts-ignore
• console.log
• TODO/FIXME
• eval(...)
• existing VS Code diagnostics (errors/warnings)

Setup

1. Open this folder in VS Code:
   • vscode-doc-guardian
2. Install dependencies:
   • npm install
3. Build extension:
   • npm run compile
4. Press F5 in VS Code to launch Extension Development Host.

Configuration

In VS Code settings, you can override:

• docGuardian.firecrawlApiKey: Firecrawl API key
• docGuardian.elevenLabsApiKey: Optional ElevenLabs key override
• docGuardian.elevenLabsVoiceId: ElevenLabs voice id for speech output
• docGuardian.elevenLabsAgentId: ElevenLabs ConvAI agent id
• docGuardian.autoSpeakResponses: Enable/disable assistant auto speech
• docGuardian.enableLiveMonitoring: Enable/disable monitor
• docGuardian.monitorDebounceMs: Monitor debounce timing
• docGuardian.showPopupsForAlerts: Popup alerts for high findings
• docGuardian.enableBackgroundDocWatch: Enable background docs monitoring on active file
• docGuardian.backgroundDocWatchDebounceMs: Delay before each background docs check
• docGuardian.backgroundDocWatchMinIntervalMs: Minimum time between repeated alerts for same issue
• docGuardian.enableWorkspaceBackgroundDocWatch: Enable changed-files workspace background monitoring
• docGuardian.workspaceBackgroundDocWatchDebounceMs: Delay before workspace changed-file checks
• docGuardian.workspaceBackgroundDocWatchMinIntervalMs: Minimum time between repeated workspace alerts for same issue

Notes

• Firecrawl and ElevenLabs are pre-wired with built-in defaults for quick install-and-run.
• Ask flow uses ElevenLabs ConvAI directly (not local mock responses).
• Firecrawl search endpoint used: https://api.firecrawl.dev/v2/search.
• ConvAI session endpoint used via SDK (@elevenlabs/client) with websocket text-only mode.
• ElevenLabs TTS endpoint used: https://api.elevenlabs.io/v1/text-to-speech/{voice_id}.