Connect an AI provider

Red Candle's AI is "bring your own". The app calculates trend, levels and statistics; the AI explains them. Everything except the AI features works with no AI at all.

Open Settings → AI Provider and choose a Provider.

Local (Ollama / LM Studio)

Free and private — everything stays on this computer.

1. Install Ollama or LM Studio and get a vision model, for example ollama pull qwen2.5vl:3b. New to this? Follow Install Ollama and Qwen step by step.
2. Keep it running.
3. In Red Candle: Server URL http://localhost:11434/v1 (Ollama's default; LM Studio uses http://localhost:1234/v1) and Model qwen2.5vl:3b (or the model you pulled).

The first request loads the model into memory and can take 1–2 minutes. After that, replies take seconds. Small models around 3B are fine for checklist rules, but they can misread charts and mix up levels, so check the numbers yourself. For better analysis, use a cloud provider with your own API key (below).

ChatGPT (OpenAI API)

1. Create an API key in your OpenAI API project.
2. Select ChatGPT (OpenAI API) and enter a vision-capable Model ID available to that project. Red Candle leaves this field empty so you can choose an accessible model.
3. Enter your API key, then click Save.

The official OpenAI endpoint is set automatically; there is no Server URL field for this choice. API usage is billed separately from your ChatGPT plan.

Gemini (Google AI API)

1. Create a Gemini API key in Google AI Studio.
2. Select Gemini (Google AI API) and enter a vision-capable Model ID available to your Google AI project.
3. Enter your API key, then click Save.

The official Gemini endpoint is set automatically. Gemini API quotas and billing apply; a consumer Gemini subscription does not replace the API key.

OpenAI-compatible (OpenAI, OpenRouter…)

Enter the Server URL (OpenAI, OpenRouter or any compatible endpoint), a vision-capable Model name and your API key.

Anthropic (Claude)

Enter your API key and a Model, for example claude-opus-5-5.

Test Connection

Fill in Model and save the required API key to enable Test Connection. Click it to send a small test prompt; normal provider API charges or quotas can apply. "✓ Connected · responded in 2.1s" means you're ready; Red Candle rechecks every minute and shows the status.

AI depth

• Auto — Full for cloud APIs, Basic for local models.
• Basic — the AI explains the levels the app found.
• Full — the AI also picks key zones, a trend line and a pattern from the screenshots, and describes the levels above and below price. Needs a capable vision model.

Your API key and your data

• Each cloud choice stores its own key: switching from ChatGPT to Gemini or OpenAI-compatible does not reuse another provider's key. Remove clears the selected key; Erase All Data clears all AI keys.
• API keys are stored in your computer's keychain (macOS Keychain or Windows Credential Manager) — never in the app database — and are only sent over https (or to localhost).
• With a cloud provider, the screenshots and notes you analyze are sent to it — nothing is sent until you click Analyze. Choose Local to keep everything on this computer.
• AI replies follow the app language.

Not working? See AI isn't connected or is slow.