AI isn't connected or is slow

Start with Settings → AI Provider → Test Connection.

Local (Ollama / LM Studio)

• Not connected — make sure Ollama or LM Studio is running, and the Server URL matches (Ollama http://localhost:11434/v1, LM Studio http://localhost:1234/v1).
• Server found · waiting for the model — the model is loading. The first request can take 1–2 minutes; later replies take seconds.
• Model must be a vision model you've pulled, for example qwen2.5vl:3b. Setting up from scratch: Install Ollama and Qwen.
• On a computer with little memory, small models can be slow when other apps are open. Close some apps, or use a cloud provider.

Cloud (ChatGPT, Gemini, OpenAI-compatible, Anthropic)

• If Test Connection is disabled, fill in Model and save an API key for the selected provider. Keys are separate for each cloud choice.
• Check the API key and that your account with the provider has credit.
• Use a vision-capable model.
• ChatGPT and Gemini use official endpoints automatically. Check that API access and quota are enabled; a ChatGPT/Gemini app subscription alone does not provide API credit.
• For OpenAI-compatible, the Server URL must be https (or localhost) — Red Candle never sends keys over plain http to another computer.

Results look off

• Small local models sometimes reply in English; AI Analysis and Similar Patterns still show the calculated facts in your language.
• For drawing, use AI depth → Full only with a capable vision model.
• Compare the reading with the chart: the structure and key levels should match what you see. If they often don't, try a more capable model.