insight-lens

Setting Up an AI Provider

Ask InsightLens is an optional feature that lets you ask questions about your survey data in plain English. InsightLens does not bundle an AI provider — you bring your own. Supported services:

• OpenAI (ChatGPT family)
• Anthropic (Claude family)
• Google (Gemini family)
• OpenRouter (aggregator for many models)
• Groq (fast hosted inference)
• Ollama running locally for fully offline use
• Any OpenAI-compatible HTTP endpoint (LM Studio, LocalAI, vLLM, etc.)

You choose which one to use from Settings → AI Assistant. There is no preselected default — the drop-down starts blank so you have to make an explicit choice.

Why this matters for privacy

• If you don’t configure a provider, InsightLens makes no AI-related network calls. The feature is simply unavailable.
• If you configure a cloud provider (OpenAI, Anthropic, Google, OpenRouter, Groq), your questions and relevant context from your database are sent to that provider’s API. Read their terms before sending sensitive data.
• If you configure Ollama or another local server, nothing leaves your machine.

Configuring a cloud provider

1. Open Settings → AI Assistant.
2. Pick a service from the AI Service drop-down.
3. Paste your API key into the Secret key field. InsightLens stores keys in the OS-level Electron secure store.
4. Click Test Connection. You should see a success toast.
5. Use Fetch Models to populate the model list, then pick one (for example gpt-4o-mini, claude-3-5-sonnet-latest, or gemini-2.0-flash).
6. Save.

If you already have the relevant environment variable set — OPENAI_API_KEY, ANTHROPIC_API_KEY, GOOGLE_API_KEY / GEMINI_API_KEY, OPENROUTER_API_KEY, or GROQ_API_KEY — InsightLens will pick that up automatically and the key field becomes optional.

Configuring Ollama (local)

1. Install Ollama from ollama.com and make sure it’s running.
2. Pull a model that fits your hardware:

    ollama pull llama3.1
   

3. In InsightLens, pick Local AI (Ollama) from the AI Service drop-down. The URL will be http://localhost:11434.
4. Leave the secret key blank — Ollama doesn’t need one.
5. Click Test Connection, then Fetch Models and pick the model you pulled.

Using a custom OpenAI-compatible endpoint

Pick Custom from the drop-down and paste the full base URL (for example http://localhost:1234/v1 for LM Studio). InsightLens speaks the OpenAI chat-completions protocol, so anything that implements that should work.

Where the call actually happens

All AI calls are made from the main Electron process over IPC, not directly from the renderer. That’s a deliberate choice — it avoids CORS issues with providers that don’t set permissive headers, and it keeps API keys out of renderer memory.

Changing providers

Switching providers is just a matter of picking a different option from the drop-down, pasting the relevant key, and saving. Your question history isn’t tied to a specific provider.

This site is open source. Improve this page.