Feature/ollama llm integration

[focusedbrain]

Ollama and Mistral 7b integrated in the installer and UI of the sidebar.

---

Note

Adds local LLM support (Ollama/Mistral) with Electron services and HTTP/IPCs, a first-run setup wizard and model management UI, plus extension-side agent execution and routing wired to the new LLM API.

• Electron Backend:
  • Add electron/main/llm/* services: hardware checks, config, Ollama lifecycle, provider-agnostic client, and IPC.
  • Extend electron/main.ts: initialize LLM services, auto-start Ollama, and expose HTTP endpoints (/api/llm/*: status, hardware, start, download, delete model, list models, chat, config).
  • Update preload.ts: expose window.llm API (check/status/start/download/chat/config/progress).
• Renderer/UI:
  • Add setup UI: LlmSetupWizard and FirstRunWizard; integrate in App.tsx (first-run flow, reopen from Settings).
  • Add BackendConfiguration with LLM tab: hardware/status view, install/switch/delete models, progress.
• Extension (Chromium):
  • Add coordination + LLM client: InputCoordinator, OutputCoordinator, AgentExecutor, llm/LlmClient, and shared types.
  • Wire command chat and agent boxes to Electron LLM HTTP API; ensure system agents; update sidepanel to route input and update box output.
  • Update provider/model selectors in agent box UIs (grid-script*.js, content-script.tsx).
  • Extend Backend Switcher to manage local LLM (status, models, install) via Electron API.
• Docs:
  • Add comprehensive guides: LLM_INTEGRATION.md, LLM_WIZARD_GUIDE.md, implementation and testing docs for agents and coordination layer.

^{Written by Cursor Bugbot for commit f682c89. This will update automatically on new commits. Configure here.}

[cursor]

This PR is being reviewed by Cursor Bugbot

Details

You are on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.

[cursor]

Bug: useEffect dependency missing causes memory leak

The useEffect hook for listening to download progress has downloadDetails in the dependency array, causing the effect to re-run every time downloadDetails state updates. This creates a new event listener every time progress updates, accumulating listeners without cleanup. The listener should be set up only once on mount. The dependency array should be empty [] or the effect should include proper cleanup to remove the previous listener.

 

[cursor]

Bug: LLM config update doesn't persist to disk

The HTTP endpoint for updating LLM configuration calls llmConfigService.update() instead of llmConfigService.save(). According to the config.ts file, update() only modifies the config in memory and does not persist changes to the file. The configuration changes will be lost when the app restarts. The IPC handler in ipc.ts correctly uses save(), but the HTTP endpoint does not. This should call await llmConfigService.save({ modelId }) to ensure changes are persisted.

 

[cursor]

Bug: Awaited async config operations incomplete before client initialization

The llmConfigService.update({ modelId }) call in the HTTP POST /api/llm/config endpoint is awaited, but llmConfigService.load() is not. This creates a race condition: load() returns a promise that's awaited, but subsequent code may execute before the config file is actually read from disk. In the /api/llm/config endpoint (line 1195), update() is called but save() is never invoked to persist changes. This means configuration updates are lost after app restart since only save() persists to disk. The correct flow should be: call update(), then await save().