Engine operator guide

ContextMint Engine is the desktop tray app that manages the Python API server, Ollama, indexing jobs, and operator diagnostics. Developers use VS Code for daily chat; operators and leads use Engine for health and capacity.

Tray & lifecycle

  • Start all — API server + Ollama (local-first)
  • Stop all — graceful shutdown; extension shows offline
  • Autostart options in Settings (OS service / login — platform-dependent)

Overview tab

At-a-glance cards:

  • Readiness — server, Ollama, Redis (if used), data layer notes
  • Workspaces — registered repos, indexing state, drill-down to Indexing / Quality
  • Chat lanes (session) — Repo / Work / Hybrid stream counts and Work-lane suggest metrics

Server tab

  • Port and bind address (default loopback)
  • Restart server without quitting Engine
  • Link to /api/health for raw health JSON

Ollama tab

  • Runtime status and URL
  • Restart Ollama process
  • First-run: install link if binary not found

Models tab

  • Installed models vs available pulls
  • Embed model used for indexing (changing it may require reindex)
  • Default chat model for local routing
  • Vision models — pull moondream, llava, or qwen2-vl for screenshot Q&A and ARGUS audits

Indexing tab

  • Per-workspace progress, file counts, errors
  • Manual reindex / activate workspace actions
  • Drill-down for file-level status on large repos

Quality tab

  • Run quality scans (clones, structure signals)
  • Findings feed governance drawer in VS Code
  • Scan progress for long-running repos

Knowledge tab

Browse index-time Knowledge Layer registries and tail redacted activity audit events — the operator complement to VS Code governance. See Knowledge & governance for worked examples, persistence (Structure is not time-limited; Activity UI defaults to a 1–24 h window), and screenshot checklist.

  • Structure — routes, services, documents, sections, policies; searchable table + graph preview
  • Activity — audit JSONL tail (governance_fast_path, chat streams, actions); export JSON

Logs tab

  • Tail server logs with time window filter
  • Use when chat or indexing fails intermittently

Settings tab

  • Python path, backend bundle path
  • Server URL override, Ollama URL
  • Operator tunables mirrored from server defaults where exposed

Context budget

Open Context budget in the Engine nav to set max_context_tokens (RAG ship cap) and co-apply max_gen_tokens (answer length / Ollama num_predict) for your hardware tier. Values are saved to ~/.contextmint/server.defaults.yaml; restart server after Apply. Interactive scaling math: Context budget calculator.

About & support bundle

Export a support bundle (versions, config redacted, recent logs) when filing GitHub issues or emailing anis@contextmint.ai.

Problems: Troubleshooting → Engine & server.